SCROLL NEWS / DISCOVERY
Search headlines
Find the stories shaping the conversation.
Results for " useful " 265 found 🔀 AI-powered Shuffle
Chartbook 464 The"China squeeze" - the politics of development counterfactuals.
There are three types of China shock discourse. Each has a distinct temporal register. China shock 1.0 was a phrase coined by US academics in the 2010s to retrospectively summarize the impact on particular US labour markets, which they identified through painstaking empirical investigation. China shock 2.0 on everyone’s lips in Europe in the mid 2020s is an alarm raised by trans-Atlantic policy thinkers warning of a coming disaster which they project onto Europe’s ailing car industry. The third type of China shock discourse - commonly known as the “China squeeze” launched most recently by Indian economists Shoumitro Chatterjee and Arvind Subramanian in the pages of Foreign Affairs - mobilizes a different, more speculative temporal register: the counterfactual realm of alternate history. The China squeeze argument asks not what has happened, or what is going to happen, but what might have been. Specifically, what options for economic development might have been open to Global South economies like India or the economies of sub-Saharan Africa, had it not been for China’s spectacular industrialization since the late 1990s and the persistent competitiveness of China’s manufacture in low-skilled segments. Chatterjee and Subramanian’s argument is that whilst China has built a powerful position in electrical machinery and other high-skill manufacturing, which is now delivering China Shock 2.0, it also continues to run a trade surplus in low-skill, labour-intensive manufacturing. This surplus is large and has remained largely unchanged in fifteen years. Indeed, in the 2010s, in trade with low-income countries, China’s surplus rapidly increased. On their reading, this squeezes low-income producers in low-skill sectors in which they “should” be able to compete. Source: “China’s Mercantilist Squeeze on Developing Countries” Shoumitro Chatterjee and Arvind Subramanian. Peterson Institute for International Economics 26-7 May 2026 Chatterjee and Subramanian’s wording is telling: (emphasis mine) The canonical trajectory of economic development in the post-World War II period is for countries to transition from agriculture to manufacturing, first in low-skilled manufacturing and then later on to high-skill manufacturing (Herrendorf, Rogerson, and Valentinyi 2013). At the global level, this implies that high-income countries (HICs) that once specialized in low-skilled activities cede manufacturing export space to poorer countries that come along to occupy that space. In 1965, advanced economies accounted for about 60 percent of global value-added exports in four key low-skilled manufacturing sectors, i.e., apparel, leather, textiles, and footwear. Over time this share declined and began to be occupied first by the East Asian tigers and, beginning in the late 1990s, by China. Today China accounts for about 50 percent of the global market. The question is whether China, with its rising prosperity, is continuing to behave like its richer counterparts and ceding export space to poorer followers or to behave unlike them and continuing to occupy space. In terms of the conceptual maneuvers it performs, of the three China shock discourses, the “China squeeze” argument is by far the most ambitious. After all, it bases its critique of China on a counterfactual imagining of the world. Like any counterfactual argument, to construct this alternative reality it must rely on a heuristic model. To make their argument Chatterjee and Subramanian mobilize two key claims. The first reference point is the actual history of rich-country industrialization, what they refer to as the “canonical path”. The idea is that countries transition from agriculture to light industry and low-skilled manufacturing and from there up the value chain. To make this useful as a standard of comparison and a means to construct a normative standard, they adjust sectoral export shares for two factors: 1. the underlying level of economic development of the exporting country, as measured by national income and 2. the degree to which the sector in question is internationalized. The idea being that for a high income country to have a high export share in low-skill sectors is anomalous and even more so if that particular product is highly internationalized and thus suitable for being “conceded” to lower-income competitors. Where should this process end? Here Chatterjee and Subramanian introduce their second benchmark. Their normative idea of a global development path is one in which, in due course, low-skill manufacturing exports are distributed in proportion to a country’s share in the global labour force. They imagine, in other words, a world of proportional, polycentric global development without much national specialization in low-skill manufacturing. A useful benchmark is that, in labor-intensive manufacturing sectors, a country’s share of world exports should be broadly commensurate with its share of the world’s low-skilled labor endowment. … In the Heckscher-Ohlin model, this would be true if value added per worker and export propensities, i.e., fraction of output that is exported, are the same across countries. It is this pleasant liberal vista of orderly global development, progressing towards a general distribution of low-skilled manufacturing against which they judge reality and conclude that the low-income countries are suffering a “China squeeze”. China is not exiting low-skill sectors fast enough Allowing for its gdp level and the degree of globalization, China’s share of low-skill manufacturing exports is far higher than it “should” be. Relative to population, its low-skill share of exports is far too high. That in turn yields the conclusion that huge numbers of jobs - the authors suggest tens of millions - are being hoarded by China. For the four sectors captured in figure 4, this excess in value-added exports is $110 billion in 2022. For all low-skilled sectors overall captured in appendix figure D.1, excess value-added exports amount to $365 billion. The crowding out of LMIC exports by China is therefore substantial. Both these claims have in common that they are strong structuralisms. The first is that development paths ought to resemble each other. The second is that broadly speaking as far as low-skilled manufacturing is concerned, all countries are alike. If the world currently diverges from the distribution of trade implied by these assumptions, if China is not following the advanced economy path and has huge exports, this must be the result of “policy”, which the authors then put in question as illegitimate or unfair. Thus the naturalistic construction of a canonical path leading to a “level” distribution of low-skilled manufacturing reveals itself to be a polemical tool. From the point of view of China shock polemics, this discourse is no doubt welcome. But as an account of economic history, it must surely be regarded as nothing short of bamboozling. Development paths do not follow a regular, natural sequence. As David Fishman put it aptly, “development is not a queue”. They are always and everywhere, through and through political. Development is not so much a process, as a project. Each such project is more or less well adjusted to its context. As Gerschenkron taught uneven and combined development implies differentiation rather than simple imitation on a single parth. It may be true that Europe and the US exited relatively early from low-skill work. But clearly technology has changed. As C&S themselves point out, China no longer enjoys a low wage environment. Do they really believe that such huge differences can be accounted for either by industrial policy, which as they admit does not operate powerfully in low-skill sectors, or by exchange rate manipulation? What is overwhelmingly dominant in explaining the persistence of “low-skill” manufacturing in China are surely the network effects of complex supply chains. Indeed the label of “low-skill” may miss the point. The degree of globalization which Chatterjee and Subramanian decide to conjure away in their model of the development path - to allow for the fact that Europe and American textiles exporters did not enjoy just-in-time logistics in the 1950s - obscures the point that the networks through which Chinese factories supply the world are themselves key parts of the low-skill industrialization project. China isn’t knocking out 1970s style “cheap”. It has redefined the entire product cycle. C&S’s casual reference to “T-shirts” is telling. In today’s “fast fashion”, there is weight on both terms. Fashion and fast. China’s success in delivering on both at unprecedentedly high speed and at unprecedentedly low cost, isn’t a low-skill operation at all. If we see development trajectories not as the serial repetition of the same, but as distinctive, crafted projects that cumulate in complex agglomerations of skill, technology and labor, then this explains why the assumption that low-skilled manufacturing should follow the distribution of population flies in the face of all experience. As countless cases attest, low-wage, low-income developed economies are not by themselves the makings of a competitive production base. As Leon Liao puts it quite fairly: “International trade has never been allocated according to population. A country’s population gives it no natural claim to an equivalent share of global exports." And the remarkable fact is that Subramanian alone and with co-authors has long been making this point, not to pillory China, as is the case of the tendentious _Foreign Affair_s and Peterson pieces, but to highlight the great counterpart to China’s industrial success, namely India’s relative failure. The much-quoted Foreign Affairs piece showed low-skill export shares only for China, not for India. If you put both countries in the picture as Chaterjee and Subramanian have done on other occasions, the picture looks like this. China is far above the line, but India, almost as large in terms of population, is no less anomalous, just in the opposite direction. As far as global populations are concerned, and the imagining of tens of millions of “squeezed” jobs, it is this Sino-Indian pair that account for a huge share of the action. What this graph reveals is not China’s cheating, but India’s “greatest development failure”, at least this is the argument made Arvind Subramanian, this time in co-authorship with Devesh Kapur, in their major book, A Sixth of Humanity. Independent India’s Development Odyssey which appeared with Oxford University Press earlier this year. In this highly critical work on India’s development, China is repeatedly evoked, not to criticize but as a benchmark against which to illuminate what India’s political economy has not delivered i.e. large-scale formal employment for hundreds of millions of citizens. Kapur and Subramanian leave no doubt, if you want to understand why China, not India was the main driver of manufacturing globalization from the 1980s, don’t look to slots opening up in the division of labour thanks to the liberal mindedness of advanced economies, look to the contrasting political economy of the Maoist PRC and India’s post-independence political economy. As Kapur and Subramian note: The great divergence between China and India did not begin in 1978 with the Deng reforms—it was well underway by then. In one of the great ironies of modern history, the conditions created by Maoism turned out to be a great launching pad for industrial globalization. As the comparison of India and China in 1980 makes clear, to imagine manufacturing would be equally distributed when populations are so differently endowed in terms of educational and nutritional is fanciful. Here is the data from Kapur and Subramanian, data which I explored back at the very beginning of Chartbook. As Kapur and Subramanian insist, the striking fact is that even as India’s growth accelerated from the 1980s, structural change has not. Formal employment in India lags woefully behind. Why? As Kapur and Subramanian note, the path-dependent impact of India’s poor early start was devastating. The list of factors they mention is long and weighty: We argue for a new five-fold explanation applicable at different stages of Indian history: in the initial stages, the legacy of two key pre-1980s policies, the neglect of agriculture and the strangling of the domestic private sector, were critical. The failure of broad-based agricultural growth militated against rural industrialization, which would have been devoted to satisfying local needs that would have necessarily been labour intensive. Then, when the boom came after 1980, the thicket of regulation and broader labour protections that were legislated prematurely in a nascent democracy, underpinned by a deep ideological aversion to private capital in general and big capital specifically, became a barrier to attaining the size that would be necessary to compete globally. Similarly, the reforms after 1991, while wide-ranging, did not quite encompass non-tradeable inputs critical to manufacturing, namely, land, logistics infrastructure and power. A combination of low investment (logistics infrastructure), regulation (land) and the continued monopoly supply by state governments (power), combined with residual socialism, kept costs at 75–100 per cent above efficient supply. These factors were compounded on the supply side by the government, which dominated the formal labour market and established exorbitantly high wages at the lower end of the skill spectrum. This made their employment in the labour-intensive sector by the private sector less attractive. Finally, when the services sector and its exports boomed, wages soared even further via a Dutch Disease effect: one successful tradeable sector killed and cannibalized another less-skilled tradable sector, namely, manufacturing. Thus, scarcity-inducing policies of the planning era, precocious democracy and a Gandhian aversion to size; inefficient, monopolistic supply of critical non-tradable inputs; the behemoth of government; and super-successful skill-intensive services successively undermined India’s unskilled labour employment opportunities. The employment failure seems to have been over-determined. … Less quantifiable but no less important are some of the sociological factors that have undermined manufacturing and factory jobs. Take the impact of the success of the IT sector—located in urban areas with better working conditions, such as air-conditioned offices, good schools for children and attractive entertainment options. It became a more attractive place for engineering and management graduates to flock to. … Consider manufacturing and female employment. Indian society makes it difficult for women to work, reflected in India’s unusually low female labour force participation (elaborated in Chapter XIII). Indian society, in particular the routine harassment of women in public spaces, makes it even more difficult for them to travel long distances to take up work.37 Combined with onerous regulations on creating hostels and dormitories for women—which, in turn, prevents multiple shifts of operation—factories need to be located in areas with a large female labour pool. While firms can take apparel plants to female labour or bring them to the plants by building large-scale dormitories, this raises costs. If there was a case for subsidies to encourage labour-intensive manufacturing, housing subsidies to build dormitories for women workers would be a good place to start. All of this meant, that when opportunities did open up, in part thanks to adjustments in China, India was in no position to take advantage. Here is the account of the 2020s from Kapur and Subramanian. It stands in dizzying contrast to the one-sided, “China squeeze” argument delivered in the Peterson and Foreign Affairs pieces: In 2022, for example, India’s share of global exports in clothing (the canonical labour-intensive sector) was 3.1 per cent compared to 32 per cent for China. This is extremely disappointing because China started vacating export space in this sector after the Global Financial Crisis as its wages started rising; it then vacated further space in the aftermath of geopolitical developments that led investors to seek alternative investment locations. Between its peak and today, China lost about 7.5 percentage points of global export share (worth about $43 billion in 2022); over this period, India’s export share remained stagnant at 3 per cent, while that of Vietnam and Bangladesh rose by 3.5 to 4 percentage points. China’s loss did not translate into any gain for India in terms of low-skill exports and employment. Put differently, when the conditions for dynamism and prosperity were established after 1991, and even more tantalizingly when the global export opportunities opened up, it was as if India was a high-wage economy and only skilled labour could benefit from the dynamism. It had converted its physically abundant unskilled workforce into an economically expensive resource, penalizing its use. It is not my mission here to argue for or against Kapur and Subramanian’s interpretation of India’s development trajectory. The point in this context is simply that the “China squeeze” such as it is can only be understood, in relation to the heavy burden of domestic constraints that Kapur and Subramanian so unsparingly enumerate. I am concentrating here on India because it is the great counterpart to China and the unspoken reference point of the “China squeeze” discourse. But the point is more general. Can anyone with a straight face suggest that the main problems afflicting the development of South Africa or Nigeria, the great centers of mass underemployment in sub-saharan Africa, are cheap competition from China driven by Beijing’s industrial policy and an undervalued RMB? Likewise, is the main problem of the American working-class the threat of Chinese competition or the manifest crisis of social reproduction in the United States itself? The point here, as in the discourse around China shock 2.0 in Europe, is to insist that we should not normalize the world and pathologize China, without actually investigating what is going on in America’s industrial wastelands, the German car industry, or India’s lopsided labour market. One-sided pathologization should be understood for what it is, a polemical gesture. Concepts like China shock and China squeeze are, what Reinhart Koselleck, following Carl Schmitt, would have described as “asymmetrische Kampfbegriffe”, or “asymmetrical counter-concepts” or “asymmetrical combat concepts”. If the function of arguments like the “China squeeze” is to alert elites in the rest of the world of the need to rejuvenate and energize their development projects, so be it. But let us see the resulting discourses as the mobilizing interventions that they are. Interestingly, a significant slice of China’s public sphere - the so-called industrial party - have been obsessed for the last twenty years with precisely these questions of counterfactual economic history. Since the mid 2000s, a massive online discourse developed around an alternative history in which Chinese techno-patriots were asked to imagine how they would proceed if they had the possibility of traveling back in time to transport modern technology into the Ming dynasty (1368-1644) thus averting the “great divergence” of the 18th and 19th centuries. It is a fantasy of course, but also a goad. The point was not directed at Westerners. The alternate history is dramatically nationalist, but few n the West had even heard of it until recently. The adherents of the industrial party are not liberal moralists. They don’t project natural ladders of international development and ask why others don’t cede space. They take the Realpolitik of global economic development for granted. Their pointe was directed against fellow Chinese. The rallying cry of the alternative narrative was this: Now that we are beginning to gain steam and push forward on our national development path, now that we are breaking out of the inferior position assigned to us even in the early years following WTO accession, do not lose yourself in humanistic imaginings about better futures. In the spirit of Deng, they called on China to embrace the hard truth of development. Perhaps one might of the Industrial Party as a cultural trend born out of desire for symbolic revenge against those who had inflicted the century of humiliation. In terms of the various temporalities of the “China shock” - retrospective, prospective and counterfactual - one might think of the Industrial Policy discourse as truly speculative, conjuring up the China shock that never was. Or perhaps one might better think of it as imagining a world in which a powerful China was no shock at all, but simply a defining fact of modern history. I love putting out Chartbook. I am particularly pleased that it goes out for free to thousands of subscribers around the world. But, writing this newsletter takes a lot of work. If you are enjoying it and would like to support the project, sign up for one of the subscription options here. It is the equivalent of a fancy cup of coffee per month! Subscribe
LLMs better than docs at providing clinical context for imaging orders
The inclusion of clinical indication is standard for imaging requisitions, but the information provided is often vague and leaves out useful context that could help radiologists.
The Beauty Of Settled Science
Commenters convinced me I made a subtle mis-step in my framing of Psychology Research Is Mostly Fine. My framing emphasized that social priming studies had failed to replicate, but that many other subfields of psychology hadn’t. This naturally led commenters to point to failed replications in other subfields. Although none of these were as bad as social priming, few or none were entirely problem-free. Instead of emphasizing the difference between social priming and other subfields, I should have emphasized the difference between settled science and novel research. Here are some of my favorite psychology findings: The brain is sculpted by reinforcement learning, and a key driver of this learning is a prediction-error-based algorithm located in the nucleus accumbens and implemented through dopamine. Psychedelic drugs exert their effects through 5-HT2A receptors, pointing to a role for these receptors in some sort of construct like the balance between priors and new evidence. Certain cognitive functions are localized to certain brain areas (for example, impulse control to the frontal lobes; linguistic structure to Broca’s area, linguistic meaning to Wernicke’s area. But injure one of these areas, or render a certain function useless (eg a blind person having no more use for their visual cortex), and these mappings can gradually shift - suggesting that neurons are general-purpose compute whose specialization is a contingent fact about convenience and efficiency rather than a necessity. Most personality traits and general capabilities are about 50% genetic and 50% random, with little role for family environment or childhood history, with a few notable exceptions like partisan preference. Recent research complicates this story, but the further you go from this basic picture, the gnarlier the chains of indirect causation you require. Neurons fire together in “brain waves”, whose exact frequency corresponds with the character of consciousness; These divide neurons into natural subpopulations and linked graphs. The gamma rhythm corresponds surprisingly well to items in conscious awareness. GABA receptors seem to be involved in something like relaxation. By hitting different GABA receptors in different ways, you can modulate the exact type of relaxation you produce. Hitting GABA-A directly is too powerful and can cause dangerous levels of sedation, but allosteric modulation of GABA is an incredibly fruitful strategy that underlies drugs from benzos to barbituates to alcohol. If you know your GABA-A subunits, you can target specific subtypes of relaxation; for example, Ambien hits GABA-A1 and induces sleep. Memory divides somewhat fuzzily into bins of immediate-sensory, short-term, and long-term, with separate retrieval and storage functions, and separate systems for declarative vs. procedural learning. These can be harmed selectively, usually by hippocampal injury, so that some patients can remember their life stories but not form new long-term memories, or learn new skills like bike-riding but not new facts. The exact substrate of memory is unclear but seems to involve new synaptic connections between neurons in at least some way. The ability of modern psychology and psychiatry to treat mental illness is real but limited. SSRIs work for half or slightly fewer than half of patients; antipsychotics work with extreme side effects; benzodiazpines work but produce addiction and tolerance in many (not all) people. Most supplements are stuck in Poor Quality Study Replication Crisis Hell, but melatonin and l-theanine stand out as especially promising. Psychotherapy works with moderate effect size, but (with a few exceptions) seems to be more about the relationship with the provider than the details of the therapy itself. Mental hospitalization can sometimes defuse immediate crises but work less well than mental hospitals would like, with some studies showing overall negative effects. Schizophrenia is mostly genetic, but risk also increases with prenatal infection (especially rubella) and obstetric complications; there is suggestive evidence for roles of social stress and drug use (with psychedelics and marijuana making things worse, and nicotine if anything making things better). Bipolar and schizophrenia are so closely linked on a pathophysiological level that it seems like they might be the same disease; on the clinical level, they do seem to follow two separate courses/prognoses but can sometimes surprise clinicians by eg someone who seemed condemned to permanent schizophrenia unexpectedly recovering. The brain’s native reinforcement learning algorithms cash out in the phenomena of classical and operant conditioning; different reinforcement schedules reinforce at different levels, which can be exploited by bad actors to make especially compelling products (eg slot machines). None of these are immune to doubt. A few are speculative; one or two are genuinely unorthodox. But they’re all a step above one p = 0.049 paper that might not replicate. If they get overturned, they’ll be overturned by the same sort of genuine scientific shift that replaced Ptolemy with Copernicus or Lamarck with Darwin, not some sort of sordid discovery about someone entering random numbers into a spreadsheet. In 2008, Eliezer Yudkowsky wrote about The Beauty Of Settled Science: If you only care about scientific issues that are controversial, you will end up with a head stuffed full of garbage. The media thinks that only the cutting edge of science is worth reporting on. How often do you see headlines like “General Relativity still governing planetary orbits” or “Phlogiston theory remains false”? So, by the time anything is solid science, it is no longer a breaking headline. “Newsworthy” science is often based on the thinnest of evidence and wrong half the time—if it were not on the uttermost fringes of the scientific frontier, it would not be breaking news. Scientific controversies are problems so difficult that even people who’ve spent years mastering the field can still fool themselves. That’s what makes for the heated arguments that attract all the media attention. Worse, if you aren’t in the field and part of the game, controversies aren’t even fun. Don’t be embarrassed to read elementary science textbooks, either. If you want to pretend to be sophisticated, go find a play to sneer at. If you just want to have fun, remember that simplicity is at the core of scientific beauty [...] A steady diet of science news is bad for you: You are what you eat, and if you eat only science reporting on fluid situations, without a solid textbook now and then, your brain will turn to liquid. Psychology isn’t centrally the new paper in Annals Of Recently-Breaking News saying that an n=5 study showing that large portion sizes increase support for terrorism. Psychology is centrally ideas like brain waves, GABA receptors, reinforcement learning, twin studies, and connectionism. People who think differently are worrying too much about BREAKING NEWS!!!, and should stop. The key distinction isn’t “recent research in social psychology” vs. “recent research in cognitive/bio/behavioral psychology”, as my essay last week unfortunately implied. It’s recent research in general vs. the deposit of settled science, battle-tested hypotheses, and carefully-earmarked speculation. And part of the reason social priming in particular was such a disaster was that it defied that heuristic, building itself up over hundreds of studies until it approached or even entered the pantheon of settled science - before collapsing (5-HTTLPR was also sort of like this). This isn’t a claim that settled science is always right - Freud was “settled science” for a long time - just that there’s an intuitive distinction between the level of credence you should have in the existence of power posing vs. the existence of Broca’s Area. (I compounded this error by calling the last post “Psychology Research Is Mostly Fine”, which suggested I was talking about the new research rather than the settled science. I actually meant to distinguish between research-psychology-the-field vs. clinical-psychology-aka-therapy, but this framing unintentionally pushed against the point I wanted to make) Some commenters on the previous article understood this point, but challenged its weight. It’s very cool, they said, to know how many neurons are in which brain lobe, but the main way that Psychology affected their lives was through bad clickbait studies used to buttress dumb policies; some p = 0.049 study says that exposure to microaggressions makes minorities’ brains shrivel into raisins, and now their company has to implement punitive DEI or be sued for hazardous work conditions. I won’t dispute this as an accurate description of how the sausage gets made, but it’s worth examining the trains of thought that connect this observation to a campaign against psychology in general: The dumb woke psychology was wrong and bad, and I don’t have good heuristics for distinguishing dumb woke psychology from other psychology, so I’m going to assume it’s all wrong and bad. The dumb woke psychology was wrong and bad; it seems useful to punish/discredit psychology as a whole to assist my campaign against it, so for propaganda purposes I will try to discredit psychology as a whole. To people in the first camp, my response is: come on, your heuristics aren’t that bad. “Microaggressions makes minorities’ brains shrivel up like raisins” is in the woke category; “the hippocampal theta rhythm seems to be associated with both memory and pathfinding, suggesting an unexpected yet surprisingly compelling relationship between the two” is in the not-woke category, don’t play dumb here. To people in the second camp, my response is: deliberately exaggerated propaganda campaigns are bad and this is a confession, not an excuse. Inability to view science through anything but a political lens was what got us into this problem; doubling down on it is a temporary solution but a recipe for permanent disaster. If you pronounce entire fields of science to be acceptable collateral damage in your campaign against bad political actors, you will alienate not just scientists but patients, clinicians, and general lovers of truth. Instead, you can say the thing you mean - that woke partisans shouldn’t try to occupy and exploit the otherwise-noble science of psychology for their own political gain - and wage the same campaign with a halo rather than a comic-book-villain goatee. From the font of oracular wisdom that is Weird Sun Twitter: [ Vessel Of Spirit@VesselOfSpirit i love you all conditionally but i love the good part of you unconditionally. even if it becomes a very small part 2:11 AM · Jan 22, 2015 1 Reply · 5 Reposts · 29 Likes ](https://x.com/VesselOfSpirit/status/558084551772471298) This is how I feel about psychology1, too, but with a touch more optimism. The good part of psychology is enormous, a hoard of endless treasures. You just have to look past the headlines! 1 To play devil’s advocate - should one always be so magnamious? Astrology? Phrenology? These seem different in that their entire orientation was flawed from the start; they didn’t simply fail to find the correct constellation-personality connections, the whole project of searching for connections between constellations and personality was doomed from the start. More difficult might be asking the same question about education research, which is notoriously bad. I don’t know enough about education research to know if there’s some settled science of correct and enlightening results lurking in the background beyond the day-to-day froth of poorly-conducted and politically-motivated studies. I think if it’s 90% bad, 10% good, it’s correct to call it bad as a compact way of communicating a useful truth; I think psychology is better than that, but admit it’s a tough judgment call.
UCLA-Led Team Selected for National Science Foundation Quantum Technology Award
A UCLA-led team of physicists, engineers and computer scientists has been selected by the U.S. National Science Foundation for $4 million in funding through the NSF National Quantum Virtual Laboratory, or NQVL, a national initiative designed to accelerate the development of useful quantum technologies and make advanced quantum tools and testbeds available to researchers across the country
An old SpaceX rocket is about to crash into the Moon
The smashup could provide useful data for future lunar inhabitants
Apple finally fixed Siri. So why does it feel anticlimactic?
Apple’s long-awaited AI overhaul finally makes Siri the assistant it was always supposed to be. But after years of delays, the launch lands in an AI landscape where chatbots have evolved into agents that can code, reason, create media, and complete complex tasks. Siri AI is genuinely useful, yet it arrives at a moment when simply being a capable AI assistant no longer feels revolutionary.
Are AI scribes useful tools in medical education, or a crutch that imperils learning?
Educators want future physicians to be fluent in AI scribes, but they worry what off-loading clinical notes to AI could do to learning.
'It might be useful to have a reality check'
BBC Radio Sheffield's Rob Staton believes it might be a time for a reality check after Sheffield Wednesday's heavy loss at Oldham.
12 Interesting Mini Gadgets You Can Buy From Ace Hardware
Looking for clever gadgets that don't take up much space? Here are the most useful compact finds worth checking out at Ace Hardware.
A mechanically tunable molecular switch for circularly polarized light emission
Over the past few decades, scientists have become increasingly interested in developing materials that do not simply withstand mechanical forces but instead respond to them in useful, detectable ways. ...
Android’s Find Hub app now lets you remove old Bluetooth headphones & earbuds
The Find Hub app on Android is super useful for tracking down lost gear through the use of trackers or...
Breaking Down the Hard Business of Measuring Security Culture
By understanding the types of data available and how they can be meaningful, security professionals can develop metrics that are useful in describing and enhancing security culture.
5 Useful Travel Accessories Every iPad Owner Should Have
If you're getting ready for a trip and are bringing your iPad along, consider picking up accessories like these to make the trip easier.
From A Useful Ghost to Self Esteem: your complete entertainment guide to the week ahead
A woman is reincarnated as a vacuum cleaner in this Cannes prize-winning black comedy, and the uncompromising pop star hits the festival trail
10 iOS 27 Performance Improvements You'll Actually Notice
With iOS 27, Apple focused on making improvements to iOS performance, optimizing everything from system animations to AirDrop transfer speeds. iPhone performance is smoother and faster on all iPhones able to run iOS 27, which includes the iPhone 11 and newer. There are over 80 iOS 27 improvements Apple announced, with 10 of the most useful listed below. App Launching Apps launch up to 30 percent faster in iOS 27, and that includes Apple's own apps and third-party apps.
iOS 27 beta 4 adds a useful Apple TV app feature, here's how it works
Apple released iOS 27 beta 4 today, and it comes with a useful new “Automatic Downloads” feature in the Apple...
The Science Behind Ketchup Pouring Could Lead to Ceramic 3D Printing
3D printing’s usefulness goes far beyond plastic.
University of Chicago scientists create "artificial leaf" that could be used for wireless implants
What University of Chicago scientists call an artificial leaf that mimics photosynthesis could be useful for wireless medical implants and other technologies.
Warren Brinson's rookie stats show pass rush potential for Packers
Warren Brinson flashed real pass rush potential as a rookie, and may prove useful in a rotational role in 2026.
iOS 27: What's New With the Mail App
Apple's Mail app is getting some useful updates in iOS 27, with Apple making big improvements to search and introducing new AI features. Search iOS 27 has an overhauled search system that extends to the Mail app. Instead of surfacing results based on keywords and recency, Mail app search ranks results by relevance and intent. Search results that come up in Mail are more relevant than before, so you find exactly what you're looking for.
5 Useful iPad Accessories Cheaper Than Apple Brand Products
You can use the RAPIQUE Bluetooth Wireless Mouse, IFEART USB-C Cable, MoKo iPad Case, Logitech Combo Touch, and Metapen Pencil A8 with your iPad.
Make the Business Case for Your Sustainability Initiative
Sustainability proposals often stall not because they lack merit, but because they are framed in a language that does not match how senior leaders allocate capital. While CSOs tend to emphasize impact, materiality, and disclosure, CFOs and CEOs need to understand cash flow, risk, and return. External reporting frameworks are useful for accountability, but they are poorly suited to internal investment decisions, which require project-level comparisons rather than aggregated enterprise metrics. Sustainability investments should therefore be evaluated like any other investment: by showing how they reduce current costs, avoid future costs, generate new revenue, or protect existing revenue. Strong cases begin with a clear business story that links the proposed action to a specific business impact and cash-flow outcome, then quantify what can credibly be quantified while treating harder-to-measure benefits as either upside or an explicit strategic bet.
Everything New in iOS 27 Beta 3
Apple is continuing to refine iOS 27 ahead of its planned September launch, and the third beta includes several new features. Siri is more customizable, there's a fun new wallpaper action, and some useful improvements to Photos and Shortcuts. Siri Voice Beta 3 adds support for Siri voice customization on compatible devices. The customization option existed in prior betas, but it was not active.
Apple Watch sleep score became more useful for me with these settings
Apple Watch includes a built-in sleep tracking feature. Just wear your Apple Watch to bed, and it works automatically. No...
The 8 Best Wellness Products Under $25, According to Experts
Some of the most useful wellness products recommended by health professionals actually cost less than $25. Here are some of the best products.
I found 3 Android 17 settings so useful I'm confused why they're hidden by default
Don't miss out on these three hidden Android 17 features
IDP Rookie Profile: Anthony Hill Jr. in Redraft Leagues
Anthony Hill Jr. lands in Tennessee with a starting role and enough downhill ability to become a useful rookie IDP option.
This satellite constellation transformed earth science by creatively tuning in to GPS signals
Key takeaways —The first constellation to produce high-quality data by picking up GPS microwaves reflecting off Earth, called CYGNSS, is featured in the Smithsonian National Air and Space Museum's Living in the Space Age exhibit. —Distortions in the reflected signals can reveal information about Earth, including hurricane wind speed and the distribution of moisture in soil, and the constellation provided unprecedented coverage in both time and space. —GNSS-R data is useful for mon
Google introduces a faster, cheaper image generator with Nano Banana 2 Lite
Google is updating its image generator to make it faster and cheaper, making it a more useful tool for creators looking to make AI content.
What is a quantum computer good for? Absolutely nothing
The Trump administration wants a useful quantum computer in two years. Microsoft wants one in three. Independent researchers cry hype.
I found a hidden Google Maps feature that made my Pixel's always-on display useful
A hidden Maps trick worth turning on
How should European scientists respond to the US AI kill switch?
Earlier in June, Europe’s worst fears of technological dependence were confirmed when the White House ordered the AI firm Anthropic to stop non-US citizens using its leading models. This kill-switch moment crystalised anxieties in Brussels that the continent, lagging the US on the creation of leading-edge AI models, is now a technological of vassal of Washington. The shock has specific implications for European scientists and universities. Some governments are now worried that with AI being integrated into labs, Donald Trump now has his finger hovering over an off switch for European science. “Having restrictions on this tool, it can have an impact on our own research,” said France’s research minister, Philippe Baptiste, at the VivaTech conference in Paris on June 18.Baptiste, himself a former academic, said that colleagues in mathematics had recently warned him they were now cut off from leading-edge US models, risking their work. “This model is not available anymore now because of the US restriction, and for the first time this had an impact on mathematics,” a field where large language models (LLMs) are making a significant contribution, said Baptiste. So, should European scientists avoid relying on US models in the future? Or would this be a self-harming overreaction? And how is AI being used in science, anyway? What happened?On June 12, it emerged that Washington, citing cybersecurity concerns, had ordered Anthropic to stop non-US citizens using two models: Claude Mythos 5, and Claude Fable 5, a version of Mythos but with guardrails to prevent abuse. Anthropic said Washington’s concerns were overblown. But forced to follow the order, yet with no way to check who was using the models, the company suddenly shut down access for everyone, US citizens included. There’s still no agreement on how to restore broad access. But last week, Bloomberg reported that select US companies, and the US government, still had access to Mythos despite the ban, but the EU’s cybersecurity agency Enisa had been shut out. Class division of researchersIn the longer term, access outside the US should be restored, said Marcel Salathé, co-director of the AI centre at the Federal Polytechnic School of Lausanne (EPFL), because a permanent restriction would be financially ruinous to US AI firms. “Even Trump will eventually budge to economic pressure,” he said.However, US AI firms might end up operating rather like banks, doing due diligence on users of their leading-edge AI models, he went on.European scientists are unlikely to fall foul of such checks, he said. But having a Chinese postgraduate in their lab, or doing a type of research the US disapproved of, could potentially result in access being denied. “It will create a sort of class division of researchers,” he said. More than frontier modelsHowever, successfully deploying AI in science is about far more than just having access to leading LLMs such as Mythos.Instead, many science-specific tools, most famously Alphafold, which predicts protein folding structures, require “far less” computing power to train than frontier models, well within the capabilities of European supercomputers, according to a recent analysis by the Arq Foundation, a Brussel-based AI think tank. Instead, tools such as Alphafold are possible because certain fields, like protein folding, weather prediction or astronomy, have large, clean, standardised datasets that can be used to train predictive models. Rather than compute, “data is the central constraint for AI for science,” the report says. For example, a less powerful physics model trained on CERN’s data “will be more valuable than the best model without certain data,” said Bengüsu Özcan, author of the report. European advantage?Europe could therefore steal a march by using its scientific data and infrastructure to create models US digital giants cannot. “They might actually match or go even beyond what these big [US] companies are doing,” said Özcan.Indeed, some European start-ups hope to create automated testing labs to create the kind of standardised data that could speed up the discovery of new materials. Materials scientists say there’s currently too little data to train models. AI scientistsHowever, even if the bottleneck for AI in science is often data, rather than compute, being at the mercy of the US for LLM access could still become a problem for European scientists. A new class of “AI scientists” based on leading LLMs is also emerging, Özcan’s report says.Unlike domain-specific tools like Alphafold, these try to automate day-to-day scientific work, by providing an overarching tool that coordinates specific models to synthesise existing literature, generate hypotheses or run analysis of data. The problem for Europe, however, is that these tools are sometimes based on US leading models. Google Deepmind’s AI Co-Scientist, for example, is built on Google’s Gemini LLM. Systematic failureThese scientific automation tools are still in their early stages, and still exhibit “systematic failures on scientific tasks,” Özcan’s report says. One of the problems is that the messy, day-to-day process of scientific discovery is largely not documented, meaning LLMs don’t have a good source of data to learn how scientists actually work. Meanwhile, some tasks, such as literature synthesis, don’t typically require access to leading edge models.However, the performance of AI scientific assistants could improve as companies such as Google pour resources into developing them. Rely on US“Unfortunately, for scientific assistants, there is not much we can do right now except rely on US frontier AI,” said Antoine Bosselut, a natural language processing expert based at EPFL. Chinese-built open models, which can run on local computers, lessening the risk of a shutdown, are touted as one alternative to US dependence. There’s also Mistral, once seen as Europe’s main AI hope, but the Paris-based company has seen its recent models fall behind in performance. But the Chinese alternatives are generally not as capable as US frontier AI models, said Bosselut, unless paired with fiddly, custom tools that allows them to act as independent agents. “Most scientists do not have the time,” he said. Don’t jump ship yetAlthough the kill switch moment is seen a warning shot for European scientists, it doesn’t mean labs should abandon US AI tools immediately. “I do not think the right response is simply to abandon US-based models and systems under the banner of sovereignty or independence,” said Mattias Björnmalm, secretary general of the Cesaer university association. “That would be an overreaction.”Related articlesEuropean sovereignty in AI requires ‘ugly trade-offs,’ say experts Robotic lab start-up makes case for €500M EU-funded materials testing facilityAI in science: is it useful? Instead, European researchers need to ask themselves whether they’re becoming dependent on technology that could be turned off in Washington, said Özcan. “They might not see this today, but they should really start asking this question,” she said. European researchers need a plan B, said Salathé. “I would recommend to every researcher, very proactively explore alternatives.”Own toolsIn addition, Björnmalm and others want European research institutions to develop their own scientific AI tools.Instead of signing AI deals with US digital giants, European universities should be using this money to create their own models, which may end up being superior for research, said Sabine Leonelli, an expert on the use of AI in science at the Technical University of Munich. “Universities in Europe seem to be happy to sign commercial contracts with US companies on the assumption [their AI tools] are better,” she said. But US frontier LLMs are typically opaque, and have been developed without independent scientific input, she said. “This is not good enough for science.”Last year, Swiss researchers released Apertus, an LLM designed to be completely transparent, from its development methods to its training data. The idea is that this makes scientific tools built on top of the model far more reproducible than with closed, commercial alternatives.“Europe needs more of this: serious capacity-building that gives researchers, universities and companies credible options,” said Björnmalm.
iOS 27 Home App: 10+ New HomeKit Features
Apple updated the Home app with some useful new features that rely on Apple Intelligence. The updates improve the way HomeKit Secure Video cameras work on Apple's HomeKit platform. Apple Intelligence Summaries The Home app generates written summaries for motion alerts and what's detected, using AI to determine what's been recorded.
The Wearable Data Your Doctor Actually Wants
Wearables can offer rough information about a person’s health, experts say, but only some metrics are medically useful.
Buying an iPhone Ultra could be an expensive experiment
If you are planning to buy an iPhone Ultra to find out how useful it is in practice, with the...
Why prediction markets’ election picks are useful, even when they seem wrong
A Post analysis found that primary candidates ended up winning about as often as bettors thought they would.
China can be a useful bogeyman in American politics
Accusations of influence-peddling can muddy what is really going on
Science vs. Screwworms: How UT Research Inspired Our Best Weapon
Key to earlier eradications of the threat was a University of Texas discovery that’s still proving useful today.
Beyond Siri: Here are the practical AI features coming to your iPhone in iOS 27
Siri’s AI overhaul may have grabbed the headlines at WWDC, but some of Apple’s most useful AI features are arriving elsewhere in iOS 27.
Latest version of Google Messages comes with useful feature we've seen tested before
Back in January and in March, we saw Google testing this new feature that is finally here.
I installed 10 Android Auto apps for a long road trip, but only 4 were genuinely useful
Not every app deserved space
OPINION: Politics, AI, and human dignity
Whether we are building technology, running campaigns, serving in government, or participating in our local political party, the measure of success is not simply whether we win. It is whether we remember that every person has dignity and value independent of their usefulness to our cause.
Google Health returns 'Hourly Activity' in June, tosses in extra for Android
The app continues to get polished in ways users might find useful.
Android 17 update deletes homescreen widgets for some Google Pixel owners
Widgets have long been a useful part of the Android homescreen and, for many, they’ve solidified their spot and functionality...
Latvia pilots citizen engagement in research grant evaluations
Latvia has introduced an element of citizen engagement to the evaluation of nationally funded research grants. The aim is to both create more connections between science and society, and to improve the quality of its evaluations. The move comes as research funders across Europe are experimenting with ways to improve evaluation processes in the face of a sharpened focus on science’s impact on society. Evaluators of grant applications submitted to Latvia’s Fundamental and Applied Research Programme (FLPP), the country’s main research funder, will this year have the option of consulting citizen feedback on challenges the public thinks science should focus on to inform their decisions. “In recent years, both public discussions and policy-level debates in Latvia have highlighted the importance of demonstrating how publicly funded research contributes to society, the economy and the resolution of real-world challenges,” said a source at the Latvian Council of Science (LCS), which manages the FLPP. “At the same time, research institutions themselves expressed interest in improving the project evaluation framework while continuing to ensure that funding is awarded to the highest-quality projects,” the source added. The citizen input comes from a survey that was conducted between 25 February and 16 March 2025, to which 1,737 people responded. It gathered information on what the public views as problem areas for Latvia, and the role of science and technology in providing solutions to these problems. A summary of these responses has been included as an annex to the FLPP 2026 call for proposals that evaluators can refer to, purely in an advisory manner, when judging proposals. Results from the survey show that the main problem areas identified by the public were in healthcare and public health, followed by the development of new treatment methods and medicines, and then digital technology, data security and cyber security. At the bottom of the list was research aimed at acquiring new knowledge about the universe, matter and the laws of nature. Further analysis looked at the public view of the possible role of science in solving these problems. The survey charted 4,167 individual responses based on open-ended questions that were derived from the participants’ answers. The analysis again found that medical and health sciences topped the scoring, with 788 responses referring to the field. Humanities and the arts were bottom of the list, with 625. Digging further into the details, the survey suggests citizens are aware of and concerned about a number of research areas. This includes technologies and innovation to improve the environment or reduce pollution, the development of modern military technologies and cybersecurity tools, and the development and application of artificial intelligence, among others. “We hope this initiative will help strengthen public trust in science and create a more meaningful dialogue between researchers and society,” the LCS source said, adding that it offers benefits to researchers, evaluators and to the general public. More space for citizen engagement in evaluationsThere is a gradual but increasing use of citizen engagement in research evaluation processes around Europe, especially in healthcare, but it remains relatively rare, said Henry Sauermann, a professor at the European School of Management and Technology in Berlin and an expert in citizen science.“Given that we are at a point where citizen involvement is extremely limited, there is presumably space for more, but it's not that we should do it just because we want to do it,” he told Science|Business. He said some research lends itself better to having citizen input, and this usually tends to be on the applied end of the science spectrum where the public can offer useful feedback based on their experiences. He added that advising scientists on certain topics to study is also a useful approach, in the way that Latvia is doing. “The public does fund a lot of science, there's a lot of tax money going into this. We of course do want to preserve academic freedom, but it's also not unreasonable to incorporate more feedback or input from citizens where possible,” he said. Related articlesCitizen science: where, when, and how to let the public contribute to researchHow can Europe cope with a flood of AI-assisted research grant proposals?Data Corner: Horizon Europe’s AI, digital and data call success rate is 5.2%There are several ways of involving citizens in science funding. The most minimal way is having a few representatives on a panel, such as members of a patient group or union. The most direct way is crowdsourcing funding for specific projects. There is an increasing number of online platforms that allow the public to chip in to support scientific proposals.Sauermann studied this in a recent paper and found that it can have its uses, but it still is not clear if there are benefits to it. There are various assumptions linked to citizen engagement, for example that it increases public trust in science, or that it provides a representative view of what the public thinks about science. These are not necessarily true, Sauermann found. People who participate in engagement activities tend to already be very trusting of science, and, on the whole, it is generally the well-off and well educated who contribute, giving a narrow view of public opinion. On whether citizen input can help manage the increasing number of proposals that are being submitted due to assistance from AI, Sauermann is more sceptical. “The obvious answer to the flood of AI-written proposals is AI-driven evaluation,” he said, adding that it doesn’t mean citizen engagement shouldn’t be a part of evaluations in a well thought-out manner. Southern Denmark’s successful approachOne successful example of citizen engagement in science evaluations was implemented in southern Denmark. Between 2019 and 2025, citizens in the region were invited to vote to fund one of five pre-selected projects in the field of healthcare. The project, called A Healthier Southern Denmark, was a collaboration between regional authorities, the University of Southern Denmark, and two media partners. Over the seven years, there were 75,000 votes and significant public engagement online and on social media. “It was a bit of a wild project, we’d never done something like that before and we were a bit nervous about what would be said in the media about it. But actually, people were very positive,” said Anne Kathrine Overgaard, one of the main organisers of the project from the University of Southern Denmark. The university and its researchers have learned a lot from the experience, particularly about engaging with society and shaping their research to match real-world needs. This is important with Europe’s increasing focus on competitiveness and bringing innovations to market faster, Overgaard said. She hopes to see more citizen engagement in evaluation processes and in science more broadly. “For many years, citizen engagement has been a bit of an add-on to proposals, just to check a box. But I think things are changing,” she said. It’s less about citizens just gathering data and more “about the dialogue between the public and the scientists,” she said. The initiative ended last year as the team wanted to regroup to find fresh momentum. The development of AI and AI-written proposals has made Overgaard reflect on how the project could be reworked to help evaluators. She hopes to bring it back in the future. “It could be a good idea to try this initiative again in a slightly re-worked format,” she said.
Merchants of Doubt and the attacks on Extreme Event Attribution science
Last week, I attended a meeting at Columbia University on attribution science and climate law, hosted by the Sabin Center. It was a fantastic event, bringing together scientists and legal experts working at the intersection of extreme event attribution and climate law. Subscribe For those unfamiliar with it, extreme event attribution attempts to quantify the contribution of climate change to an extreme event. For example, several groups analyzed the impact of climate change on Hurricane Harvey’s enormous rainfall totals over Houston, Texas and they found that climate change increased rainfall by 15 to 38%. One thing that came up again and again was how terrified fossil-fuel interests are of extreme event attribution science. They are acutely aware that this research could land them in court. And losing those cases would leave them legally liable for billions of dollars in climate damages. Because the legal stakes are so high, the blowback has turned ugly. I spoke with several scientists at the meeting who are facing ongoing harassment over their work. This blowback is a coordinated campaign to make the entire field look suspect. The goal is to create the impression that attribution science is too uncertain, too political, or too conflicted to be useful in court or in public policy. The strategy is not based on actual science or evidence of misconduct, but on the generation of doubt. The new Merchants of Doubt We’ve seen this before. In fact, not that long ago: We only have to go back a year to the Department of Energy (DOE) Climate Working Group (CWG) report to see an example of using doubt as the tool to push back against well-established science. This strategy is laid out in an email from a member of the CWG, Dr. Roy Spencer, that was released during litigation over the Climate Working Group process. The key quote is: About all I can hope is that what we write will provide sufficient “reasonable scientific doubt” regarding the science claims in the 2009 TSD [technical support document], based upon almost 2 decades of new science, to call into question the original reasoning for the EPA Administrator’s decision that CO2 presents a threat to human health and welfare. This statement is strong evidence that at least some members of the committee were working to support a particular policy outcome: revoking the Endangerment Finding. The email also explains how they planned to do it: by attempting to generate “reasonable doubt”. This is going to be hard, Spencer implies. Despite falsely claiming that “2 decades of new science” weakens the case, Spencer explicitly acknowledges that the actual peer-reviewed science of climate change overwhelmingly rejects his position: But if the science argument is decided upon by a vote, or by the number of published citations, we lose the science argument. We can go back even further: This CWG email shares unmistakable DNA with the infamous 1969 tobacco memo that declared: “Doubt is our product, since it is the best means of competing with the ‘body of fact’ that exists in the mind of the general public. It is also the means of establishing a controversy.” The tobacco memo also acknowledges the limit of this strategy: Like the CWG, they knew the science was not on their side. Share The new new Merchants of Doubt The people attacking the IPCC chapter on extreme event attribution are the newest iteration of the Merchants of Doubt. Their goal, like all Merchants before them, is to introduce doubt into the process. Because the report is not even out yet, they cannot attack its conclusions. So they are attacking the authors instead. Here is a press release from the House Science, Space, and Technology Committee: In the letter, the Chairmen express concerns about potential conflicts of interest involving members of the Attribution Committee, stating that “publicly available information suggests a troubling pattern” in which committee members are affiliated with nonprofits that support climate accountability lawsuits, “raising the appearance of impropriety and member bias.” To be clear, this is just innuendo. There is no actual evidence of bias. And given the robust process that these reports go through, including multiple lines of peer review, it seems very unlikely that significant bias can survive into the report. When the report comes out, critics will have the opportunity to make legitimate criticisms of the report — if any exist. If none do, however, they’ll still make criticisms, but they’ll be bogus, simply designed to generate doubt. We’ll see. A note to the press: Fix your frame To any journalists reading this: The public debate over extreme event attribution science is not going away. The science is simply too dangerous to fossil-fuel interests for them to stop fighting it. You very well might be assigned to write an article about this area of research in the future. When you do, do not automatically adopt the framing that climate misinformers want you to use. They want you to frame the story around questions like: Are climate scientists trying to put their thumb on the scale to achieve a predetermined, politically motivated result? Are climate scientists improperly letting their politics invade the science of the IPCC? That frame is a trap. Instead, you need to view this through the historical lens of the Merchants of Doubt. How does the ecosystem of doubt operate? Who funds it? What methods do they use to misrepresent science and slime researchers? What scientific results are they trying to keep people from understanding are legitimate? Ultimately, you instead need to focus your article on the generation of doubt as a way to maintain the fossil fuel industry’s social and legal license to keep burning oil, gas, and coal. If you treat the misinformers’ frame as a legitimate, good-faith scientific critique, you are helping them produce doubt. Don’t do it. Don’t be a Merchant of Doubt. Thanks for reading The Climate Brink! Subscribe for free to receive new posts and support our work. Subscribe other stuff The Staying Curious Substack has an interesting post about the different flavors of climate sensitivity. This is crucial information if you want to understand how scientists think about how much warming the Earth will experience. I’d be grateful if you could hit the like button ❤️ below! It helps more people discover these ideas and lets me know what’s connecting with readers.
Is SMIC N+3’s Metal Pitch Smaller than Intel 18A’s?
Almost four years ago, we published that SMIC had started shipping 7 nm (N+1) chips. Now, SMIC is shipping its third-generation 7 nm (N+3) process in Huawei’s Kirin 9030, with a minimum metal pitch of 32.5 nm, about 10% tighter than the 36 nm minimum metal pitch shipping in Intel’s latest Panther Lake CPUs on 18A. The headline is true, but incomplete cherry picked metric. N+3 reaches the density of TSMC N6 through aggressive DUV multi-patterning and design-technology co-optimization (DTCO), but it pays for that in complexity, efficiency and process control. We found this and more in our reverse engineering and teardown where we cover SMIC’s N3 process technology, Huawei’s packaging, memory, architecture, and more. SemiAnalysis has been building a state-of-the-art teardown lab in Oregon capable of analyzing the world’s most advanced and important chips over the last year and half. We have already generated revenue on advanced datacenter chip teardowns including our recent reverse engineering of a major TSMC customer’s COUPE CPO optical engine + EIC 3D stack. This is the first public report from the SemiAnalysis Teardown Engineering & Evaluation Lab, or STEEL for short. The lab is aggressively scaling up and out and we’re excited to announce it publicly. This is a bit of inconvenient timing for TechInsights as they are private equity owned and currently being sold while having enjoyed virtually no credible competition for decades. This has led to TechInsights underinvesting in CAPEX. SemiAnalysis exceeds TechInsights in revenue despite no venture or private equity ownership and being founded only 6 years ago. Because we have no external investors and are founder led, we move faster, build faster, and we can release client chip teardowns for free regularly, while focusing on datacenter for our major clients. Here’s the first public image from our lab, the HiSilicon Kirin 9030 Pro SoC: This report will detail our teardown of the Kirin 9030 and our findings on SMIC’s N+3 process, the most advanced in China. For comparison, we’ll show our teardown of the MediaTek Helio G99, made on TSMC N6. Through this comparison, we can look at the effect of export controls – SMIC N+3 and TSMC N6 are comparable nodes, but one is heavily export-controlled, the other free to use the West’s most advanced equipment. Here we see both China’s progress and constraints. SMIC N+3 reaches TSMC N6-class logic density, but it requires far more aggressive DUV multi-patterning, so it does not match N6 on process maturity or cost. The Kirin 9030 Pro performs similarly to three-year-old Android flagships, and trails far behind the current flagship SoCs from Apple, Qualcomm, MediaTek, and Samsung. The efficiency gap is even wider. Export controls have not stopped Huawei and SMIC from shipping advanced silicon, but they have forced a different path. Without EUV, SMIC is leaning harder on DUV multi-patterning, DTCO, and increasingly complex integration. The roadmap continues forward through tighter design rules and backside power, but each step adds cost and process risk. Huawei’s τ scaling and LogicFolding show another path: stacking active logic and recovering density through advanced packaging and system-technology co-optimization (STCO). To understand the Kirin 9030, we must first understand Huawei’s SoC history. HiSilicon is Huawei’s chip design arm, responsible for the Kirin smartphone SoCs, Kunpeng server CPUs, Ascend AI accelerators, and switch/router networking silicon. Before export controls, Huawei was TSMC’s largest customer – the only customer on TSMC’s first EUV node, N7+, and among the first on N5, alongside Apple. That ended in late 2020. Huawei switched to Qualcomm SoCs in its flagship smartphones, though export controls limited them to 4G-only variants. In late 2023, Huawei returned to in-house silicon with the Kirin 9000s, a successor to the Kirin 9000, fabricated on SMIC N+2 instead of TSMC N5. In the following years, they released the Kirin 9010 and 9020 on the same N+2 process. These chips used Huawei’s in-house TaiShan CPU cores and Maleoon GPU. We have not torn down a Kirin 9020 ourselves, so the predecessor die shot is from Kurnal. The die shots show how Huawei spent its silicon budget: which functional blocks are where, and how their areas compare to the predecessor. First, a quick guide to the major blocks on the die. The total die area is nearly identical, but the 9030 uses that area more aggressively. A denser process lets Huawei fit an extra middle CPU core, more GPU and NPU cores, and larger caches into the same footprint. In contrast, the Helio G99 is a much smaller, low-cost SoC, built for budget smartphones rather than a flagship device. While the Kirin 9030 is ~140 mm², the G99 is only ~29 mm², roughly one-fifth the area. The underlying TSMC process technology, however, is directly comparable as a baseline for analyzing SMIC’s. The Kirin 9030 is an evolutionary refresh, not a clean-sheet design. Its CPU, GPU and NPU cores carry over the 9020’s families, and the gains come from three levers: the SMIC N+2-to-N+3 process step, DTCO and floorplan work, and incremental microarchitecture. Area is where the first two show up, and the 9030 scales well here. Performance and efficiency are the harder test. Huawei’s design holds up better than its node would suggest, but the chip still trails, both because N+3 sits behind the leading-edge nodes and because its cores, while competent, remain a few generations behind the newest designs. The new prime core is an incremental update. The main changes are a 10% frequency increase from 2.5 GHz to 2.75 GHz and a doubling in the L2 cache from 1 MiB to 2 MiB. Despite the increased cache, the core size decreased by 7.6%. Excluding the private L2 cache, the core size decreased by 21%. This is a large reduction for an incremental node. Compared with the TaiShan New V120 core in the Kirin 9020, the Kirin 9030’s middle core is almost unchanged architecturally, yet each core shrinks by ~22%. Most of that comes from the move from N+2 to N+3, with layout likely accounting for the rest. Visually, the most noticeable change is the increase from 3 middle cores to 4. There is also a 20% increase in the shared L3 cache of the big cluster. This helps to improve multi-core performance without sacrificing much in terms of area. Even with each core shrinking, the big CPU cluster’s total area is essentially unchanged. The per-core savings went back into an additional middle core and larger caches. The tiny cores shrank less than the prime core (excluding its L2 cache) and less than the middle cores. This is likely because fixed overhead is a larger share on a small core. We cannot resolve any architectural changes from the die shot alone, but the per-clock and efficiency gains shown below point to more than pure process and layout scaling. The area reduction was offset by a doubling of the shared L2 cache from 2 MiB to 4 MiB, leaving the total tiny CPU cluster area slightly larger. Area is the easiest improvement to see from a die shot, but it is only one part of PPA (power, performance, area). For modern logic, power and performance matter just as much, and often more. Since Dennard scaling broke down in the mid-2000s, voltage and frequency have not scaled in step with transistor dimensions, so each node has had to fight harder for gains in performance and efficiency. The starkest comparison is not Kirin 9020 versus Kirin 9030 Pro. Apple’s efficiency cores run circles around Huawei’s prime core. Apple’s low-power core delivers 20% higher integer performance while drawing only 1 W, compared with 4.5 W for Huawei’s prime core. N+3 matches TSMC N6, but N6 is several generations old. Apple and Qualcomm build on N4 and N3P, which are denser and sit on a better voltage-frequency curve, giving them a larger transistor budget and more performance per watt. The 9030’s own cores did improve. The middle and tiny cores gained 17% and 14% in per-clock integer performance over the 9020, with floating-point flat on the middle core and up 11% on the tiny. The tiny core improves cleanly, with performance rising while power falls and efficiency increases by 45% in integer and 24% in floating point. The middle core is mixed: integer performance rises but power rises faster, cutting integer efficiency by 7%, while lower power lifts floating-point efficiency 16%. Per-clock gains at the same or lower frequency are microarchitectural, so the cores are tuned, not just shrunk. Both also failed to hold their rated maximum frequencies, pointing to thermal, power, or stability limits. Per clock, the middle core sits around Arm Cortex-A720 and the tiny core near the Cortex-A520; absolute performance trails because Huawei clocks them much lower. The prime core is roughly Cortex-X2 class per clock, a 2021 design. Apple’s 2020 M1 Firestorm core is still 35% higher per clock and 57% faster in absolute integer performance at a similar 4.5 W. The current leading edge is further ahead again: the Apple M5 P-core is 60% higher per clock and 2.7× faster, the Arm C1 Ultra 45% higher and 2× faster. Matching older high-end cores per clock is a genuine design achievement. What Huawei cannot match is the voltage-frequency curve and transistor budget of leading-edge nodes, which let Apple, Qualcomm and others spend more transistors in the same area on wider cores, larger caches and deeper buffers while running at lower voltage. Huawei’s LogicFolding roadmap is one answer, stacking active logic to recover density and shorten signal paths. We return to it later. The GPU compute units (CUs) changed more visibly than the CPU cores, moving to a more rectangular layout for both the arithmetic logic unit (ALU) clusters and the CU overall. Even with ray-tracing support added, a CU shrank ~28%. However, that shrink is offset by the increase from 4 to 6 CUs and the area outside the CUs grew 33%. Overall, the GPU cluster is larger by ~10%. The GPU is where Huawei makes its biggest gains. The Maleoon 935 is not competitive with current flagships, but it is a large step up from the 920 and reaches older-flagship territory. In 3DMark it is 70% faster in Wild Life Extreme (WLE) and 79% faster in Steel Nomad Light (SNL) than the 920; with 11% higher clocks and 50% more CUs, the ~67% theoretical uplift roughly matches WLE and is beaten by SNL. It edges ahead of the Snapdragon 8+ Gen 1 in WLE and SNL, and the Dimensity 9200 and Apple A16 in WLE, but stays far behind newer parts: the Snapdragon 8 Elite Gen 5 and Dimensity 9500 are ~2.4–2.6× faster in WLE and ~3.2× faster in SNL. The Maleoon 935 is Huawei’s first GPU with hardware-accelerated ray tracing; there it lands slightly ahead of the Exynos 2200, and on par with the Apple A16, with current flagships up to 3.7× faster. The Neural Processing Unit (NPU) saw the largest structural changes of any block, moving from a Lite and a Tiny core in the Kirin 9020 to a Lite and two Tiny cores in the Kirin 9030. Both core types also show significant layout changes. This is a reversal in Huawei’s NPU design. The Kirin 9000 5G, its last flagship chip on TSMC N5, used two Lite and one Tiny core. The series of SoCs on SMIC N+2 moved to one Lite and one Tiny core, likely for area savings. With the Kirin 9030, Huawei has shifted back toward a larger multi-core NPU cluster, but with the additional area going to a Tiny core rather than a Lite core. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Interested in joining us on this ride and think you can be a difference maker? Check out our Careers page. Before diving into the process stack, the package and memory are worth separating from the SoC itself. The Pro variant of the Kirin 9030 carries 12 GB of Samsung DRAM, with two stacks of four dies each. The dies were identified as the K4L2E165YD, a 12 Gb LPDDR5X-9600 device fabricated on Samsung’s 1a node, the fourth generation of its 10 nm-class DRAM after 1x, 1y and 1z. 1a has shipped in volume since 2022, so this is current memory rather than older-node inventory. The 16 GB Pro Max variants we obtained were found with both CXMT and Samsung packages. The CXMT package is marked CXDD7JEDM, with two stacks of four dies, packaged in week 45 of 2025. The inferred die dimensions from X-ray computed tomography (CT) are consistent with a known density of ~0.3 Gib/mm² for the CXMT G4 process, roughly equivalent to other manufacturers’ 1z processes. The Kirin 9030 uses a typical integrated package-on-package (iPoP) stack: multiple DRAM dies in a memory package sit above an organic redistribution layer (RDL) interposer, which sits above the SoC and package substrate. The full package is then mounted to the printed circuit board (PCB) through ball-grid array (BGA) solder bumps. The memory package substrate is a thin bismaleimide-triazine (BT) laminate carrying the LPDDR5X stack. The organic RDL interposer over the SoC routes the PoP signals around the die and carries possible dummy thermal copper pillars. The package substrate, a thicker Ajinomoto Build-up Film (ABF) build-up over a BT core, fans the flip-chip bumps out to BGA pitch and embeds the power planes. The whole stack is organic. The only silicon is the SoC and the LPDDR5X dies; there is no silicon interposer. Keeping it all-organic brings the package’s coefficient of thermal expansion (CTE) close to the PCB’s, reducing board-level warpage, and avoids the cost of a silicon interposer the SoC’s bandwidth does not need. In an iPoP stack, the memory package connects to the organic RDL interposer through an array of solder bumps. Underfill fills the gap around those bumps, adding stiffness and protecting the joints from mechanical stress. The Pro and Pro Max variants differ here, which we cover behind the paywall. The die shot and architecture tell us how Huawei allocated its silicon budget. The process tells us what SMIC can manufacture. We use the Helio G99 as the process reference for TSMC N6. Both SMIC N+3 and TSMC N6 are evolutions of previous 7 nm-class nodes. We used targeted TEM cross-sections through logic and memory regions, imaged in both fin-cut and gate-cut directions. Each cross-section caption gives its horizontal field width (HFW), the real width of the imaged area. We start at the transistor fins, then move up through standard cells, local interconnect, and SRAM. SMIC has not overtaken Intel or TSMC. It uses aggressive DUV scaling and DTCO to reach N6-class density, but that density doesn’t translate into comparable performance and efficiency, for two reasons: the node gap to leading-edge nodes, and Huawei’s core designs. Fin Profile One of the most important knobs in a FinFET process is the fin profile: the shape of an individual fin and the channel where current passes from source to drain. The ideal fin is tall, narrow, and nearly vertical. A taller fin increases effective channel width, while a narrower fin improves electrostatic control by thinning the body the gate must control. Push either too far, and the process pays for it: weaker drive current, fragile fins, taper, footing and line-edge variation that hit yield and device variability. The Intel 22 nm, 14 nm, and 10 nm fin cross-sections show how FinFET nodes have improved over time. 22 nm fins were a first-generation structure, relatively short, wide and strongly tapered. The shape limits current density and reduces gate control uniformity across the height of the fin. At 14 nm and 10 nm, Intel pushed the fins taller and narrower while also making the sidewalls more vertical. Rather than shrinking the device, these changes increase the effective channel width per fin and improve electrostatic control. The trade-off is that taller fins at tighter pitches are much more difficult to manufacture. Now, let’s compare the Helio G99 on TSMC N6 with the Kirin 9030 on SMIC N+3. Both processes are in the same class, with fin pitch of 30-32 nm on N+3 and 34 nm in our N6 cross-section. The pitch for N6 is especially interesting as N7’s HD library is generally listed with a 33 nm fin pitch, and N6 did not shrink pitches directly. Its density gains came from DTCO instead of tighter pitches. The 34 nm pitch was stable across our sampled region and serves more as a comparison against the SMIC N+3 we have not investigated further. Pinning down N+3’s fin patterning scheme takes more than one core unit. The CPU cores show a dense ~32 nm pitch, with the pitch between N-P fin pairs alternating between 78 and 88 nm. Logic alone may be consistent with dual-pitch mandrels of 120 and 110 nm, but this is a complex and unusual approach. Combining the pitch from the the 8T SRAM, which has more complex repeat unit, with the CPU core sequence allows us to reverse engineer the patterning steps with more confidence. As both the logic and SRAM should share the same base grid, a single CD mandrel lithography pattern with 128 nm pitch undergoing SAQP produces a die-wide ~32 nm grid (128 nm/4), which supports the pitch sequencing seen in both logic and SRAM cells. In the sampled cross-sections, N+3 shows a taller, narrower, higher-aspect-ratio fin than N6. The measured fin aspect ratio is ~9.5:1 on N+3 versus 7.8:1 on N6. N+3 also shows less top rounding, with an estimated radius of ~2 nm, compared with 2.8 nm on N6. Even though the fin widths differ, the ratio of top rounding to fin width tells the same story, with N+3 at 0.37 and N6 at 0.44. In a geometric sense, lower is better; a perfectly rectangular fin would have no top-rounding penalty. These are single-digit-nanometer features measured from a handful of cuts, so treat the absolute numbers as approximate. The important result is the relative gap: N+3’s fins are consistently taller, narrower and less rounded than N6’s. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Interested in joining us on this ride and think you can be a difference maker? Check out our Careers page. Standard Cell A standard cell is the basic building block of chip layout: a fixed-height row pairing one NMOS and one PMOS transistor that share a gate, tiled in a grid to build logic blocks. The key dimensions are contacted gate pitch (CGP), cell height (CH), fin count, and the lower-metal routing grid. To measure density, we use the Bohr metric: a weighted average of NAND2 gate area (60%) and scan flip-flop area (40%). This represents a realistic mix of combinational and sequential logic. This metric has its limitations, especially for complex cell layouts like TSMC’s FinFLEX, which alternates cells with different fin counts. Even so, it is the best metric for a pure process-level comparison. Another important measurement is the fin pitch; it refers to the distance between two fins of the same transistor. In a FinFET process, multiple fins are used in each transistor to increase the drive current and thus performance. TSMC N6 ships both a high-density (HD) library with 2 PMOS and 2 NMOS fins per cell, and a high-performance (HP) library with 3 of each. More fins under the shared gate mean more effective channel width. HP cells switch harder at the cost of area. Designers mix the two on a die, primarily spending HP cells on timing-critical paths, and matching their PPA targets. In the Cortex-A55 core of the Helio G99, we found a cell height of 240 nm for the HD cell. MediaTek has used HD cells in the G99 to minimize die size and thus cost. As an SoC for budget smartphones at ~$100, this is essential. By contrast, we found only one library in the Kirin 9030, with 2 NMOS and 2 PMOS fins. This suggests a narrower library strategy than TSMC N6, where both HD and HP libraries are widely used. This likely reflects the smaller customer base and the more constrained domestic design and electronic design automation (EDA) ecosystem. In all three CPU cores of the Kirin 9030, we found cell heights of 228 nm, 5% smaller than on N6. This is also a reduction of 9.5% over SMIC N+2’s cell height of 252 nm. SMIC N+3 and TSMC N6’s HD library both feature a CGP of 57 nm. For SMIC, this is a 9.5% shrink over N+2. In the past, CGP and cell height alone may have been enough to compare transistor density. Now, however, we must consider scaling boosters and DTCO as well. SMIC’s density gain does not come from EUV. It comes from using every available DTCO booster aggressively. First is fin depopulation: reducing the number of NMOS and PMOS fins in each cell. The first FinFET nodes started with 3 or 4 fins for each transistor. SMIC N+3 and TSMC N6 HD both use only 2 fins per transistor, trading drive strength for density. Next is contact over active gate (COAG). By landing the gate contact directly over the active gate, instead of out over the isolation region, the cell height drops. N+3 integrates COAG while N6 does not. Our N+3 gate-cut cross-sections indicate COAG, with the gate contact sitting over the active region, while N6 shows an off-gate contact. Last is single diffusion break (SDB). Diffusion breaks are inserted between cells in the same row to provide electrical isolation, but they also introduce local layout effects (LLE), layout-dependent shifts in electrical characteristics. In the past, a double diffusion break was used, consuming the space of two CGPs. SMIC N+3 and TSMC N6 instead use SDB, saving area but increasing LLE sensitivity. This must be controlled at the process level and accurately modeled in the process design kit (PDK) so that EDA tools can account for it. Overall, SMIC N+3 has a transistor density of 113.4 MTr/mm², slightly above TSMC N6 at 107.7 MTr/mm². Even without EUV, SMIC has achieved density beyond TSMC’s mature N6 node which utilizes EUV. Metal Stack The smallest critical dimension in the teardown is M0; SMIC N+3 uses a 32.5 nm local metal pitch. That is smaller than the 36 nm M0 pitch on Intel 18A in Panther Lake. However, this does not mean that SMIC has a better process than Intel 18A or TSMC N3P. M0 is a local intra-cell routing layer. Its usefulness depends on the full interconnect stack: M1 and M2 pitch, track count, via and line resistance, design rules, mask count, overlay control, and routing flexibility. The 32.5 nm M0 is consistent with self-aligned quadruple patterning (SAQP), whose four-population line-width loading we read coarsely as alternating widths of 21.5 to 24 nm; M1 and M2, at 38 and 40 nm, are consistent with self-aligned double patterning (SADP), a single A/B split. On TSMC N6, M0, M2, and M3 sit at a relaxed ~40 nm and are consistent with SADP-class double patterning, with no need for quadruple patterning. That said, we measure M2 for example at ~43 nm, likely inflated by sparse routing. We do not assign any specific layer to EUV from our cross-sections; the distinction we can draw is double versus quadruple patterning, not lithography wavelength. Transistor-level density in the front-end-of-line (FEOL) sets an upper bound, but the design is ultimately limited by what the interconnect stack can route. The lowest metals are the most important for standard-cell density, but the semi-global and global layers determine how usable that density is at the block and chip level. Two axes are commonly used for chip cross-sections: the fin-cut and the gate-cut. The micrograph above is a fin-cut and shows metals 0 through 3. This axis lets us see the even-numbered metals, with M0 right above the fins. There are two kinds of M0 lines. The first are the power rails; these are wide wires for the VDD and VSS running horizontally at the top and bottom edges of each standard cell. The wide wires measure 55 nm across, more than double the other M0 lines. Their width minimizes resistance and reduces IR drop. The second kind are intra-cell wires, short segments within the cell that connect terminals to M1. These have alternating widths between 21.5 and 24 nm. The M0 pitch is 32.5 nm, a 19% reduction versus N+2 and N6. At this pitch, DUV patterning requires more aggressive multi-patterning, increasing mask count, overlay sensitivity, process complexity and cost. M0 is below what a single DUV-defined spacer (SADP) can resolve, so SMIC cascades a second spacer step (SAQP). The cross-section reflects the cost: the M0 trenches are visibly more re-entrant (narrower at the bottom than the top) than M1 or M2 on the same chip and carry a bright barrier-rich foot where the trench meets the etch-stop layer. That shape is partly the intended damascene profile, as a slightly narrow bottom helps void-free copper fill, but its magnitude at M0 is driven by the tight pitch and the higher trench aspect ratio. Intel 18A supports an M0 pitch of 32 nm, although Panther Lake has only shipped with a looser 36 nm pitch. This is due to Intel’s heavy usage of HP libraries. Among leading-edge nodes, 18A has the loosest M0 pitches due to PowerVia. With power routing moving to the backside, congestion is reduced, and the entire front-side metal stack can be used for signal routing. M2 is the first true inter-cell routing layer. It runs horizontally like M0 but spans across multiple cells to carry block-level signals. The M2 pitch sets the cell’s track height – the number of M2 tracks that fit between the VDD and VSS rails, defining what the library calls a 6-track or 7.5-track cell. This layer is the most important, limiting the routing of entire blocks. SMIC N+3 features a 5.7-track cell. The M2 pitch is 40 nm, a 5% decrease over N+2 and the same as N6. This shrink keeps the pitch at the edge of what is possible with double patterning. Future nodes will need to increase the number of masks for M2 as reducing the number of tracks is much harder due to the limitations in routing. The micrograph above is in the perpendicular direction, the gate-cut, and shows metals 0 through 4. This allows us to see and measure the odd-numbered vertical metal layers. The M1 pitch is 38 nm, 9.5% less than N+2 and 33% less than N6. The M1-to-gate ratio matters because it sets local routing flexibility. N+2 and N+3 use a 3:2 ratio, while N6 uses a 1:1 ratio, explaining the huge differences in M1 pitch. The more M1 lines there are compared with the gates, the more flexibility there is for power and signal crossing within the cell. Routing flexibility enables more complex and better cells. Clean fractional ratios are also preferred as a grid is periodic and improves layouts. The 3:2 ratio gives SMIC more local routing flexibility than a strict 1:1 grid, but it also complicates layout and patterning. This is a DTCO choice, with SMIC increasing process complexity to recover density and routability without EUV. This 3:2 ratio is not very popular in the leading-edge nodes. TSMC has only used it on N7+, the N5 family, and the short-lived N3(B). They have switched back to a 1:1 ratio for N3E. Intel has only used it on the 10 nm/Intel 7 family, with Intel 4, 3 and 18A all using a 1:1 ratio. Samsung is the only one still using a 3:2 ratio at the leading edge, using it in the SF4 and SF3 families. It remains to be seen if SMIC will remain at a 3:2 ratio or move to a 1:1 ratio with its future nodes. The industry is still actively exploring these local-routing ratios. At VLSI 2026, imec will be presenting work on even higher ratios, including a 2:1 scheme that can reduce area by up to 14%. We will be covering the conference in a future newsletter article. Subscribe We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Interested in joining us on this ride and think you can be a difference maker? Check out our Careers page. The final local interconnect layer for N+3 is M3, with a pitch of 44 nm. The M3 pitch is the same as on N+2 and 10% larger than on N6. The semi-global layers carry the majority of block-level signal routing. They have a coarser pitch than the lower local layers. On leading-edge nodes, they are designed to sit at the limit of DUV single patterning. M4 through M11 pitches were found divided between 80–82 nm (M4–M6), 128 nm (M7–M10), and 148 nm (M11). Given limited sampling, it is possible these are divided further in dense routing areas. At the top are two giant metal layers, M12 and M13. These have kept the same pitches as N+2 at 1920 nm and 4600 nm respectively. While the lower layers’ pitches are generally fixed by the process and library, the upper layers vary much more in pitch and count, depending on the design. Even two smartphone SoCs on the same process can have wildly different metal stacks. The Helio G99 carries fewer routing layers, reaching coarse metal pitches of 850 nm by M9, while the larger and higher-performance Kirin 9030 keeps fine pitches until M11. SRAM At the leading edge, SRAM is much more difficult to scale than logic. TSMC’s latest nodes have seen little to no bitcell scaling, while logic still has more DTCO levers to pull. While looking for other logic libraries in the GPU compute units, we stumbled upon the SRAM. The most common type of SRAM has 6 transistors (6T), but this cell had 8 transistors (8T) instead. 8T SRAM adds two transistors to form a dedicated read port. Unlike a 6T cell, where reading disturbs the storage, the decoupled read port removes read-disturb, improving read stability and letting the cell be pushed harder for performance. At first glance, the cut looked like an unusual logic library, with each cell row having 3 fins of one polarity and 5 fins of another. The rows also alternated in orientation. Energy-dispersive X-ray spectroscopy (EDS) resolved our confusion. The cut had not landed on the GPU logic, but on the SRAM macro beside it. The unusual fin pattern was due to the SRAM library. We return to EDS in the process flow analysis behind the paywall. SRAM libraries are not like traditional logic libraries. Due to the unequal number of PMOS and NMOS transistors, they require specialized rules and layout libraries. They do not need the flexibility of logic libraries, so they are hyper-optimized for one purpose: dense, reliable memory. The SRAM cell we found is a 1:2:2-2:2 cell. This means there is 1 fin per pull-up (PU) PMOS transistor, and 2 fins per pull-down (PD) and pass-gate (PG) NMOS transistor. These 2 PU, 2 PD and 2 PG transistors would usually form a single 6T high-current cell (HCC). An 8T HCC adds a read-pull-down (RPD) and a read-pass-gate (RPG) NMOS transistor, each with two fins. We measured a cell height of 406 nm, which brings the bitcell size to 0.0463 µm². That is a theoretical peak density of 21.6 Mib/mm². We estimate that a 6T HCC would have a cell height of 292 nm and a size of 0.0337 µm². This is ~12% larger than a 6T HCC on Intel 3 and 4. We also estimate the 6T high-density cell (HDC) to have a cell height of 228 nm and a size of 0.0260 µm². This is coincidentally the same as the logic standard-cell height measured earlier. The estimate puts the cell near Samsung 7LPP/5LPP and slightly below TSMC N7/N6. That is a theoretical peak density of 38.5 Mib/mm². 6T HDC is arguably the most important cell as it is used for the largest caches in a chip, the L3 caches and system-level cache (SLC). Both the Kirin 9020 and 9030 have split the SLC into 4 banks to raise total SLC bandwidth. In the Kirin 9030, the SLC increased from 2 MiB to 3 MiB per bank. Correspondingly, the number of arrays within the bank also increased by 50%, from 16 to 24. Each array can store 128 KiB and forms an orderly pattern on the die shot. From the Kirin 9020 to the Kirin 9030, the area of a 128 KiB SLC array decreased from 0.0477 mm² to 0.0392 mm², an 18% shrink. The achieved density is 25.5 Mib/mm², 66% of the theoretical maximum. While the SLC was quite similar across both chips, the L3 has seen some major changes, particularly in terms of its layout. The total capacity also went up from 10 MiB to 12 MiB. Much like the SLC, the L3 is also split into 4 banks. In the Kirin 9020, an L3 bank consisted of 16× 128 KiB arrays and 16× 32 KiB arrays. However, an L3 bank in the Kirin 9030 instead consists of 48× 64 KiB arrays. In the Kirin 9020 L3, a 128 KiB array was 0.0513 mm² and a 32 KiB array was 0.0154 mm². The size of the 128 KiB array is different on the L3 and SLC as the assist circuitry for the two arrays differs depending on their purpose. In the Kirin 9030 L3, a 64 KiB array is 0.0210 mm². Although not a like-for-like comparison, normalized for capacity, it is 18% smaller than the 9020’s 128 KiB L3 array and 31% smaller than its 32 KiB L3 array. The achieved density is slightly lower than the SLC, at 23.8 Mib/mm², 62% of the theoretical maximum. Unlike the L3 and SLC, the prime cores’ private L2 cache uses a 2-bank design. As the prime cores’ L2 is latency-critical, it likely uses 6T HCC instead of 6T HDC. The 9020 has 16 arrays in each bank while the 9030 has 32. Each array has a capacity of 32 KiB. A 32 KiB array in the L2 shrank from 0.0171 mm² to 0.0142 mm², ~17% smaller. The density is 17.6 Mib/mm², ~59% of the theoretical maximum for 6T HCC. SRAM scaled well from N+2 to N+3, shrinking by ~19%, close to the theoretical logic shrink. The caveat is that N+2’s bitcells were unusually large, bigger than comparable 7 nm-class nodes, so part of the gain is catch-up rather than true scaling. With the insights from STEEL’s teardowns, we will be doing a deep dive into SRAM in a future newsletter article. Subscribe Everything above came out of a single STEEL teardown: die annotation, block-level area analysis and TEM cross-sections through logic and SRAM. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Future Roadmap The same cross-sections that pin down N+3 also show where SMIC can go next. Although N+3 is already close to the practical limits of DUV multi-patterning in several layers, SMIC still has a few scaling levers left. A theoretical N+4 would likely start with cell height. N+3 uses 5 M0 tracks between its power rails. Moving to 4 M0 tracks, as on SMIC N+2 and TSMC N6, could reduce cell height by roughly 15%. The routing grid is only one side of the shrink; the front end also must fit into the smaller cell. One possible FEOL lever is reducing the p-to-n isolation spacing from two diffusion grid units to one. Intel used this scaling booster on Intel 4, and TSMC did so on its N3 family. This path trades layout flexibility for density. Fewer M0 tracks reduce local routing resources, while tighter p-to-n spacing raises integration and design-rule difficulty. M2 is also constrained by the cell height shrink. For SMIC to maintain a ~5.7-track cell, M2 would need to move toward ~35 nm. That would move another layer into SAQP territory. SMIC could also reduce the CGP from 57 nm to 54 nm. Intel reached a similar CGP on Intel 10 nm/Intel 7 without EUV. The local interconnect is also tougher. If SMIC keeps the 3:2 M1-to-gate ratio, M1 would need to shrink to 36 nm and would likely require SAQP as well. If SMIC moves to a 1:1 ratio, M1 could relax to 54 nm, but it would give up routing flexibility. Under this theoretical path, we estimate that SMIC N+4 could reach a cell height of 198 nm and a CGP of 54 nm, implying a Bohr density of 137.8 MTr/mm², on par with TSMC N5 or Samsung SF4. However, the difficulty is cumulative. Each step is individually plausible, but together they make N+4 harder than the transition from N+2 to N+3. It will likely take longer, cost more, and carry less process margin. A theoretical N+5 would require a larger integration shift. One possible path is backside contacts (BSCon), moving power routing and source/drain contacts to the backside, which would reduce front-side routing pressure and enable another reduction in cell height. The front-side metal pitches could relax to reduce the process complexity. M0 would likely relax slightly to ~34 nm, and M2 and M4 pitches could relax further. CGP is unlikely to shrink much further. Even with EUV, 48 nm has been the practical limit for yield and process control. This approach would allow N+5’s cell height to fall to 170 nm and its CGP to 53 nm. This implies a Bohr density of 163.6 MTr/mm², on par with Intel 18A’s HP library. However, this would not make N+5 cost-competitive with the leading edge. It would reach a similar density through a much more expensive route. The integration difficulty rises sharply, with new process flows for backside alignment, wafer thinning, contact reveal, and backside metallization. Past this point, standard density and interconnect scaling become increasingly unattractive. That is where Huawei’s roadmap stops looking like a normal foundry roadmap and starts looking like a packaging roadmap. At ISCAS 2026, Huawei unveiled its tau (τ) scaling law, reframing process scaling in the time domain. τ is the time cost of data movement and processing: switching delays in transistors, RC signal propagation delays in circuits, compute, memory, and networking latency. Outside Huawei’s terminology, this is called system-technology co-optimization. This is Huawei’s answer to its lack of EUV lithography. Without EUV, planar density cannot keep pace with TSMC, Intel, or Samsung. If transistor density cannot shrink further, Huawei’s alternative is to shorten wires, reduce buffering, and stack logic vertically. “LogicFolding”, Huawei’s implementation of this new scaling idea, is, in practice, an aggressive 3D stacking approach. AMD V-Cache places SRAM above or below a CPU die. AMD’s MI350X places active interposer dies (AIDs) underneath accelerator and compute dies (XCDs), with the AIDs handling cache, IO interfaces, the network-on-chip (NoC) and embedded metal-insulator-metal (MIM) capacitors. With LogicFolding, parts of the same logic block are split across multiple active dies bonded face-to-face at ultra-fine pitch. This allows Huawei to shorten some critical paths and reduce buffer overhead, not merely add cache capacity or offload the IO and interconnect. Shortening wires is where the higher clocks come from. A large share of a modern core’s delay and energy budget goes into driving long interconnects and the repeater buffers along them. LogicFolding distributes a block’s critical-path gates across multiple stacked tiers bonded at a very fine pitch, so the bond interface behaves like an additional metal layer and the longest paths get shorter. That is how Huawei expects to recover frequency and efficiency it cannot get from the process alone. Huawei’s roadmap shows its intent. Prime core frequency is targeted to rise from 2.75 GHz in the Kirin 9030 to roughly 5 GHz by 2031, far beyond what planar scaling alone could deliver. Prime cores with 3.1 and 3.39 GHz clocks are being tested in its labs, although their power consumption is unknown. Beyond that, chips are in the design, simulation or pathfinding phase, meaning the frequencies are targets. However, the direction matters more: LogicFolding also helps with performance, not just density. The catch is that Huawei’s density claim is not directly comparable to foundry densities. A stacked design can report more transistors per package footprint by adding active layers, even if each patterned die remains well behind TSMC or Intel in front-end density. This is how Huawei can claim to reach foundry 14A-equivalent density by 2031. This is not a like-for-like foundry comparison, with Huawei using stacked logic and measuring density per package footprint. On a normalized Bohr-density basis, SMIC N+3 is ~114 MTr/mm², 38% less than Intel 18A’s HD library. Huawei’s 3D roadmap closes the gap by stacking active logic, reaching 215 MTr/mm² by 2030. In 2031, the roadmap density jumps to 295 MTr/mm², implying either a third active layer, partial EUV insertion or aggressive planar DUV scaling. Huawei’s methodology makes other foundries look much denser as well. Applying it to AMD’s MI450X with an N2 top die on an N3P base die yields a theoretical density of 460.2 MTr/mm² in 2026, compared with Huawei’s 295 MTr/mm² in 2031. This Kirin 9030 does not use LogicFolding, remaining in a conventional mobile SoC package. Instead, it forms the baseline for how far Huawei and SMIC can push planar scaling. Future teardowns of Kirin and Ascend chips will show both planar logic density and Huawei’s hybrid bonding solutions. Export controls changed China’s optimization problem rather than ending it. EUV restrictions raised the cost and complexity of leading-edge manufacturing without freezing it. SMIC reaches N6-class logic density through DUV immersion, SAQP and DTCO, while Huawei shifts more of the burden onto architecture, packaging and system-level integration. Future nodes will be tougher. N+3 still had room to tighten local metals and reduce cell height and CGP. Further scaling without EUV leaves fewer levers. More aggressive multi-patterning adds masks and overlay error. SMIC can keep pushing DUV, but each step will get more expensive and less forgiving. The design side is just as critical. Huawei had domestic EDA tools and flows before the Kirin 9030, with the Kirin 9000s, 9010 and 9020 making that clear. Huawei was able to ship multiple consumer SoCs on SMIC N+2 and N+3 while cut off from the Western EDA stack. US export controls restricted EDA tools for advanced chips in 2022 but did not target tools for more mature chips. In 2025, the US government briefly placed much broader restrictions on EDA software from Synopsys, Cadence, and others, before lifting them less than two months later as part of a trade deal tied to rare earths. Huawei has been unable to access those tools because it remains on a US trade blacklist. That forced Huawei, SMIC and Chinese academic institutions to build their own tools and flows. Researchers at Peking University recently announced a prototype EDA tool for Huawei’s LogicFolding architecture, which requires a new flow to handle the multilayer layout and floorplan. This is not the same as replacing the full Synopsys or Cadence stack, but it shows where domestic EDA is headed: toward tighter co-optimization between architecture, process and packaging. These advances are also diffusing into the Chinese ecosystem. SMIC is licensing its N+2 and N+3 processes to HLMC/Hua Hong at the government’s direction rather than by choice. If the same process learning feeds into Ascend accelerators for AI training and inference, the choke point shifts from one named fab to an ecosystem. Alibaba’s T-Head silicon arm and Cambricon, a Chinese AI chip designer that is expected to supply ByteDance, could also be major beneficiaries. Sanctions aimed at SMIC alone become less effective once the manufacturing knowledge has spread to other fabs and design houses. China is not closing the gap with Intel, Samsung and TSMC. The teardown shows the opposite in several places: no EUV, no backside power, higher process complexity, and visible trade-offs. But China is still advancing. If domestic chips become good enough for phones, inference, networking and security-sensitive workloads, they can matter strategically without matching TSMC at the leading edge. Behind the paywall, we show what else STEEL can do, with material and process flow analysis of SMIC N+3, and analysis of the Kirin 9030 package. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline, access the full Kirin 9030 and SMIC N+3 analysis or to commission a custom teardown, contact sales@semianalysis.com.
In Retrospect: Poland, Russia, Europe
In April I spent a few days with Zryw, a Polish non-profit organization that recruits young people to enter politics and public life. Their aim is to build a new generation of civic leaders in Poland. I’ve followed their evolution from the beginning; my son is one of the founders. (“Zryw” means “impulse” or “burst of speed” in Polish). Julie Chmielewska, one of Zryw’s leaders, interviewed me at the event. Here is the conversation, lightly edited for grammar and context: Julia Chmielewska: Assuming that culture shapes institutions, could we conclude that some cultures are more prone to authoritarianism? Anne Applebaum: I don’t think so. Every culture, every nation, and every country has the potential for change. When we look at history, we can see that nations do change. Poland has changed, for example, from the 1980s to today. There’s a completely different mentality here than the one I encountered when I first visited, in 1988. The current generation sees the world through a different frame. So no, I don’t believe that certain nations are doomed to authoritarianism, or to anything. Of course, there are political and cultural traditions, but they can always be challenged. JC: So do you think there’s something intrinsic in human nature that pushes people to democracy, freedom? AA: I think there is something in human nature, a deep instinct, that pushes people towards freedom, yes. People who live in authoritarian countries – and I’ve heard this from them over and over again – sooner or later find themselves angered by a feeling of helplessness. They can see the unfairness of crooked courts, irrational political decisions. They naturally dislike the fact that somebody has all the power while they have none. In those circumstances, people who have no political background at all sometimes are politicized just by that frustration. I’m thinking of Svetlana Tikhanovskaya, for example, who was a Belarusian housewife, until the circumstances of her husband’s arrest forced her to become a politician. I met a woman who is a Uyghur from China, and who had come to live in Turkey. After several of her family members were arrested, she too was propelled into a political role. These kinds of experiences make people interested in democracy or political reform, even if they thought they didn’t care. I also think that the opposite is true: some people instinctively prefer homogeneity, order, hierarchy. They don’t like the confrontation of opinions or the noise of debate, and they lean towards a single leader or a unified regime. I think that most of modern history is shaped by the clash between those two different sides of human nature. JC: Which force do you find more powerful today? AA: I think they both have a lot of power and there’s no way of predicting who wins. Even in countries that seem very shut off or stuck in an authoritarian system can change. We’ve seen that happen. And vice versa, I think democracies can also begin to fail and slide into authoritarianism. So it’s very important not to be complacent in either direction. JC: I want to go into your history and your experiences. You’ve lived in Russia during a time of incredible change in the late 80s, and then you came back to do research in post-Soviet Russia. AA: I was a student in Russia just before the collapse of the Soviet Union. I then traveled and worked there extensively in the 1990s, and I met people who really were full of hope. They hoped that Russia could become a democracy, or at least a place that guaranteed freedom of speech, freedom of the press. These weren’t ideas that came from Washington. Although now the narrative in the Kremlin is different, I met many people who were working towards exactly that kind of transformation. A friend of mine ran a series of seminars – once a year, and then later several times a year, in cities all over Russia – designed to get people to discuss politics in new ways. She invited young politicians and journalists, as well as foreigners. I went to many of them, in Moscow as well as Volgograd and Novosibirsk. Over two decades, hundreds of people came to these seminars, and all of them wanted Russia to be different. I remain convinced that many Russians still want Russia to be different. It’s just that the old system, in the form of Putin, a representative of that system, returned. The KGB reimposed their power. And this wasn’t because of something to do with culture, this was to do with money and politics, and who was in charge. JC: You’ve partially answered my question... I wanted to ask whether that excitement of the 1990s was an exception from the Russian historical norm which has been rather authoritarian? AA: Of course, if you look at the history of Russia since the 18th century, there have always been liberal groups, reformers and revolutionaries. There were the Decembrists, in the early 19th century, and the rural populists later on, who tried to organize the peasantry. A significant group of liberal Russians appeared at the turn of the 20th century, people like the father of the writer Vladimir Nabokov, who was a Russian politician before the revolution. These groups were small in number, but sometimes had a broader significance. The Russian dissident movement in the 1950s and 1960s actually invented the modern human rights movement. It was their idea to hold the leaders of the Soviet Union accountable for the language they used, to remind them of the treaties they had signed and the promised they had made. They created the Helsinki group, for example, that held Soviet leadership to the language of the Helsinki Treaty, which formally ended the Second World War and which contained guarantees of human rights, at least on paper. That idea spread to Eastern Europe, and similar tactics were used here in Poland. (Benjamin Nathans recently wrote a great book about the Soviet dissidents, To the Success of Our Hopeless Cause) JC: But seemingly, even if they do ever get power, they could never hold on to it. AA: We don’t know, because they have never really had power. But in any case, we also don’t know what will happen in the future. There is no reason to assume that Russia will forever be condemned to totalitarianism. Even during the first fifteen years of Putin’s regime there was some form of democratic politics. Only when the full-scale war began did he bring everything to an end: he shut down all remaining independent organizations and killed his most important opponent, Alexi Navalny. But does that mean he is popular, that he enjoys legitimacy? Even now, we don’t really know what Russians think about him, or about the war. There is no way of doing real polling in Russia because nobody will answer an independent pollster over the phone. JC: Do you think that the flock of young people, especially young men, who fled Russia after the start of the war represent a continuation of liberal movements that refuse to partake in the authoritarian state, or do you think they are moving for the sake of self-preservation – “I don’t want to go to the army, so I leave”? AA: I’m sure it’s both. I actually know many people who left Russia for political reasons either in the very last years before the war, or else just after the war began. JC: And do you see a change in the young generation versus the old one that have fled? AA: It’s hard for me to say because the people who I know were mostly already politicized. I do know some younger people who left, but most of them were already either active in the opposition, or working as journalists. Many of them very self-consciously think of themselves as maintaining Russian culture abroad, just as in the 20th century the Soviet dissidents or the anti-communist White Russians once thought of themselves as people who maintained the true Russian culture, outside of Bolshevik Russia. JC: Let’s move on to Poland. One of your first experiences with Poland was reporting on Margaret Thatcher’s first state visit, in the autumn of 1988. You followed her around the farmers’ market. AA: She was making a symbolic gesture. She wanted to welcome free markets, but the only one she could really visit was the farmers’ market at Hala Mirowska. JC: And I want to talk about this area because today you can look out of any window in that district, Mirów, and see rows and rows of skyscrapers. How does it feel to see a country built from the ground in such a short time, both politically and economically? AA: You know, when you’re living in a period of dramatic change, it’s always hard to know whether the change is positive or negative. The transformation in Poland created winners and losers, as well as new classes of people, new political movements of different kinds. At the time, it sometimes just felt like turbulence. Now, we can look back at the 1990s and early 2000s in Poland and see in retrospect that the country was headed in towards prosperity and European integration. In around 2009 or 2010, I remember speaking to a woman who worked cleaning government offices, and who asked me if her daughter should study in Amsterdam or in London. This seemed to me to be a real transformation, and it happened very fast: within a few years, Poles stopped feeling like second-class Europeans, and a generation of younger Poles appeared who had no qualms about crossing borders. JC: We again live in turbulent times. Do you think we’re heading towards positive outcomes? AA: I can’t make predictions, because everything that happens tomorrow depends on what people do today. Also, history doesn’t have a trajectory. It doesn’t move in a particular direction. You only think it does afterwards. As I said, you can look afterwards at Poland in the 1990s and you could say this was a period of great progress. But not everybody at the time saw that. Sometimes people are very wrong about a given country’s trajectory. I remember people writing about Ukraine in the 1990s, about how dangerous it was that Ukraine was independent, that this was going to be an anti-semitic state, that Ukrainian nationalism was going to come back... and although you can say many bad things about Ukraine, the country made many bad choices, actually it became a pretty tolerant state which now has a Jewish president and until recently a Muslim defense minister. JC: In the 1990s Poland’s and Ukraine’s GDP were roughly similar. Ukraine was even a bit richer. So the starting point was similar for both of these countries, but they’ve turned out very, very differently. Even if we stop in 2014. Why do you think Ukraine “chose East”? Was it external influence? AA: No, I think Ukraine was in a different position. First of all, central and eastern Ukraine had been part of the Soviet Union since the 1920s, and had therefore experienced a longer period of occupation, repression, Sovietization and even Russification. Ukraine did not have, as Poland did, an alternative elite: a large cadre of economists and politicians, who already had some experience in opposition politics and in independent thinking. In Poland, the transformative finance minister, Leszek Balcerowicz, had already thought about how an economy might be decentralized before he took office. We can argue about how successful he was, but at least he was prepared to try something new. In Ukraine, people weren’t prepared. For one, there hadn’t been much space even for an illegal opposition in the previous couple of decades. Ukraine was also in many ways still emerging from colonial occupation. For the previous couple of centuries, many of the most capable Ukrainians had moved to Russia, and the first generation of leaders in independent Ukraine tended to be former Soviet apparatchiks. They didn’t immediately seek to become closer to western Europe. Instead, they looked for some kind of economic and political “third way” and an accomodation with Russia. Ukraine became stuck in this position, half-in, half-out and became pretty profoundly corrupt. JC: The Polish alternative elite had a vision of where they wanted the country to go. I think the general perception now is that this vision ends with the next election, right? Is this a natural process of gaining power? AA: Their vision, and it was shared by the time by politicians who now call themselves both left and right, was that Poland should join the EU and NATO, as fast as possible. My husband (Radek Sikorski) was in the very first government that said it wanted to join NATO. At the time this was taken very badly in Washington: the demand for NATO membership absolutely came from Poland, not from the US. But even then, in the early 1990s, many Poles already understood that the Russian sphere of influence would grow once again, that it was important to become part of the transatlantic alliance and an integrated Europe as fast as possible: NATO, the EU, the Council of Europe, all of the trading and security alliances. That consensus lasted through several different governments and over many years, even governments containing former communists. For twenty years, right, left and center disagreed about many things, but they all agreed on the direction that the country should go. In retrospect, that consensus created the basis for prosperity in Poland. JC: And regarding the Polish-Ukrainian relations. We know that antagonisms, e.g. the myth of Wołyń [a massacre of Poles that took place in what is now Western Ukraine, in 1943] are increasingly used in Polish political discourse today. In Ukraine, nationalist historians make heroes out of UPA, the Ukrainian revolutionary army, which is held responsible for this tragedy. Where do you see a start for burying the hatchet? Is the reconciliation of our nations possible? AA: So here’s a strange thing: For more than a decade after they both gained independence, Poland and Ukraine participated in a number of projects that were designed to solve exactly these kinds of Polish-Ukrainian differences. Books were published that were jointly written by Polish and Ukrainian historians, monuments were built. These projects were successful: If you’d asked me this question in 2010, I would have said that there are no real historical issues between Poland and Ukraine, that the two countries had mostly reconciled. For a number of reasons, this began to shift. In Poland, children and grandchildren of victims of the Wołyń massacre felt their tragedy hadn’t been fully recognized. In Ukraine, nationalist groups wanted the UPA partisans who had fought against the Soviet Union recognized for their heroism, despite the fact that some in this broad and not well-defined group were also linked to Wołyń. In both eastern Poland and western Ukraine, the memory of the ethnic cleansing that took place there in the late 1940s - Poles were moved West, Ukrainians East - to accomodate new borders has left feelings of displacement and bitterness. The wave of Ukrainian immigration, which started before the war, may have provoked some of this. But don’t underestimate the role played by Russia, which has always wanted to create discord between Poland and Ukraine, and pumps out propaganda to that effect in both countries. Vandalism on both sides of the border, damage done to cemeteries and monuments, is sometimes carried out by provacateurs working for Russia, just to create bad feelings. The Polish-Ukrainian conflict also hides a different reality: although there are real grievances and bad memories, the truth is that the biggest tragedies both nations experienced in the twentieth century were inflicted by the Soviet Union and Nazi Germany. That should unite these two countries, not divide them. JC: This week (19.04.2026 – Ed.) in Hungary we saw that using historical narratives to create grievance has no longer worked. Are you optimistic about the direction after Orban’s loss? AA: The Hungarian election was really important because it shows that people can actually resist powerful, dominant propaganda. The ruling party Fidesz controlled all of the TV, 90% of the newspapers, the constitutional court, the bureaucracy... and yet people had the nerve to vote against it and resist it. It’s also an important election because Hungary created a huge network of well-funded think tanks and foundations that were funding other far right groups and projects. So that loss is going to have a knock-on effect in other places too, maybe even in Poland. JC: What you’ve said about resisting propaganda reminds me of one of the interviews you did about talking to Russians in the post-Soviet era who claimed that people knew that the system was lying. Why do people choose to believe in these regimes anyway? AA: Because it’s useful. Stephen Kotkin, an American historian, once described how people would learn how to “speak Bolshevik”: It was just a language that you needed to use in order to get by in society. If you didn’t talk the way the regime wanted you to talk, then you might have problems, or lose your job, or your kids would not get into university. It was a way of getting along. Communist Poland was similar, at least in the early years. When I was writing a book about the Sovietization of Eastern Europe after World War II, I met a lot of people who wanted to talk a lot about the war, when many of them had done really heroic things. But almost nobody wanted to talk about the decade afterwards, between 1945 and 1956. This was when you had the short period of terror, violence and Polish Stalinism. To stay safe, people felt the need to just go along with what others were saying, or stay out of politics, or use communist jargon without necessarily believing it. I don’t judge them for it, and I wouldn’t describe all of those people as collaborators. It’s easy to feel moral certainty in retrospect, because we know that communism ended. But if you were living in 1949, it might not have been clear to you that communism was going to end. In very repressive societies, maybe about 10% of people who are really evil – collaborators, torturers. Maybe 10% of people are really brave and who find ways to resist. And then around 80% of people have kids, they have parents, they have to have jobs and they’ve had to get through it somehow. Strange though it may sound, I’ve found that the work I did on that book is now very useful in explaining some of what’s going on in the United States today. There are plenty of members of the Republican Party in Congress who know perfectly well that Donald Trump says many things that aren’t true. But they avoid criticizing him because it’s inconvenient, because they would get in trouble, because someone would attack them on the internet. So actually you don’t even need mass terror to get people to conform. JC: Is Gen Z reinventing agency? AA: We’ll see. In any generation you can find groups of people who want agency. If you live in a society that’s unfair, for example where a tiny number of people have captured most of the wealth, then at a certain point people rebel. This is always easier for younger people, who have you have less to lose. The question now is how the internet will affect the relationships between citizens and the state, and how it will change organizations. The “Twitter revolutions” people spoke about as early as 2008-9, mass protests organized online, have often proved to be very superficial. Although you can easily join something just by clicking, the work of creating a real movement requires physical participation. Successful mass movements of the past, for example the Civil Rights movement in the United States, were built by people who had spent many years together. They had participated in non-violent workshops together. They practiced and prepared. Just using the you can get everybody to show up at a certain moment, but then when the protest is broken up, nothing happens. It’s also true that the repressive regimes have more recently learned how to use the internet to undermine leaders and movements. Most of all, they now understand how to make people cynical. The form of propaganda that most modern authoritarians use - and I think the Russians really pioneered this - deploys sneering cynicism. Instead of praising the rulers, the regime propagandists mock the opposition movement leadership for being supposedly corrupt or sexually degenerate, with the goal of undermining any kind of idealism. JC: MAGA grew on the internet? AA: MAGA is a real grassroots movement. They had symbols, red hats, they showed up at rallies. But you’re right that the cynicism, the sarcasm and the distrust that the movement embodies came originally from the internet, and this is different. Fifty years ago, most European political parties had a social base in real organizations, in trade unions for Social Democrats, in church groups for Christian Democrats. As people stopped going to church and as the trade unions became smaller, the social base of these parties disappeared. Now people who wouldn’t have known one another in the past meet on the internet and create new movements, many of which don’t last. JC: Is history ending? Going in stale circles? AA: History never ends. In truth, human beings want contradictory things at the same time: order and freedom, stability and change. These things are never resolved, and when a society swings too far in one direction, people always push it the other way. But each time that happens, the result is different. Certainly it is true that we are now in a period when old philosophers and ideas from the 1930s are undergoing a revival. Carl Schmitt, who was a philosopher who was very influential on the Nazis, is back as an influence on both the far left and the far right, once again inspiring people to think that all politics has to be existential, that all politics is about defining and defeating enemies. In the US, some are also bringing back dominionism, the idea that we need a Christian and not a secular state. It’s also true that we are living through a more fundamental shift. The old left-right divide, which was mostly about the size of the state, is much less important now. Instead, we are arguing about more fundamental things: what are our rights, what protects our freedom. Or how much power do we want to give small elites, versus how much power to give ordinary people. In America a small group of very wealthy people now control information, because they control the social media platforms and AI, which means they determine what people will read and learn. Finding a way to preserve preserve zones of autonomy and independent thinking, this is going to be the main task for the next generation. JC: What is the one book that a person interested in Central Europe should read? AA: You could read my book “Iron Curtain: the Crushing of Eastern Europe” which tells the story of how the Soviet Union imposed its culture on the region. Or read Tim Snyder’s history of the region, “The Reconstruction of Nations: Poland, Ukraine, Lithuania, Belarus 1569-1999.” And everyone should read “The Captive Mind” by Czesław Miłosz, which describes exactly the phenomenon we were talking about: why do people collaborate? I joined Scott Galloway and the brilliant Fiona Hill in a discussion about the wars in Ukraine and Iran, the future of global power, and what these conflicts reveal about America's role in the world. Scott told us that this was the 400th edition of his podcast, which is quite a number: My friend Francesco Chiamulera runs a brilliant series of literary events, Una Colline di Libri, A Hill of Books, in the Prosecco Hills, a region of vineyards and elegant hill towns in the Veneto, north of Venice. We did an event in Vittorio Veneto and also drove and biked around the countryside. Buy Autocracy Inc Buy Iron Curtain Open Letters, from Anne Applebaum is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Subscribe
Fraud detection critical to online health research, study finds
While online research is a useful way to reach people who may not take part in in-person studies, researchers are increasingly concerned that fake, automated, and duplicate survey responses can reduce data quality and compromise findings.
New OMB rule could break science in the United States
The drug that saved your mother. The treatment your doctor recommended. The clinical trial that bought someone you love more time. None of it was inevitable. It required a system built to operate outside politics and sustained over decades, because science doesn’t operate on the timeline of a politician. That system is now under direct threat. The Office of Management and Budget (OMB) has proposed a new rule to regulate all federal grants, not just science, but housing, education, defense, NASA and everything in between, that would fundamentally change how research is conducted in the United States. There have been a lot of fires in the past 16 months, and if everything is an emergency, nothing is an emergency. But this is a very big deal. These rules would strip away the independence that has kept science from becoming a political tool. The good news is there’s still something we can all do. How the grant system was designed to work When I was a new faculty member, I was full of millions of ideas, so I sat down and wrote an NIH proposal. I worked on it for months and thought it was so good. Then I sent it to my mentor. It came back covered in comments. Places where my reasoning wasn’t as airtight as I’d thought, methods that could be stronger, gaps in my thinking that I had been too close to the work to see. It stung, as honest feedback always does, but peer review almost always makes the work better. In fact, it’s the backbone of how American science has functioned for generations. At NIH, submitted proposals go to a panel of scientists who have experience in the field. These are the people who can read research methods and spot their weaknesses, who understand when an experiment is designed well and when it isn’t. Applications are scored rigorously, and often only the top 3-20% of proposals receive funding. The people making funding decisions are required to justify them based on these reviews. To be clear: this is not a perfect system, and good science still goes unfunded more often than any of us would like due to limited resources. There is also criticism that peer review may undervalue bold ideas that are risky but could have high payoffs. But the peer review system is nonetheless a meaningful part of the grant selection process. While thoughtful and careful changes could improve the system, the changes being proposed by OMB will not make things better. The rule that could change everything Through this OMB proposal, the system that helps decide what gets funded is being dismantled. There are many troubling elements buried in this proposal, but three provisions stand out in particular. Each would be damaging alone. Together, they would change what kind of science gets done in America, and who it serves. 1. Political appointees would decide which science gets funded, and peer review would be explicitly sidelined or ignored. The guidelines your doctor follows, the standards that determine what’s in your air and water, the drugs that get approved for your family: all of it starts with a funding decision. Right now, that decision is made by scientists who can evaluate whether the work is actually sound and whether it could move the needle forward. Under this proposal, funding decisions could be made by political appointees. The careful, expertise-driven review process that shapes grant funding would be effectively sidelined. That means that a poorly designed study could get funded simply because someone in a political office liked the sound of it, while rigorous, important work doesn’t get funded because it no longer aligns with the administration’s priorities. In other words, critical work that benefits all of us could now be missed just because it has become politically inconvenient. 2. Grants could be canceled at any point, without warning, in the middle of ongoing work. Say your tax dollars have been paying for a cancer study for two years. The grant was awarded for five years, so the team is mid-experiment. There are three additional years’ worth of answers that could change how the disease is treated for hundreds of thousands of people. With this new rule, the federal government could cancel the remaining funding tomorrow, walk away from the commitment it made with your money, and have no legal obligation to finish what it started. This is a huge waste of money. There are rarely other sources of funding for this work, which means the team, the graduate students, postdocs, and research staff who have dedicated years of their careers to this project could lose their jobs. Patients who have contributed their time to studies may not see it turn into useful learnings for others. Every hour of work already completed could result in nothing useful. Science can take decades to get to tangible outcomes. It needs a stable funding system to flourish. 3. Researchers would have a harder time communicating their findings without pre-approval from the federal government. There is a saying from Sir Mark Walport: “Science isn’t finished until it’s communicated.” Science is no use to anyone if it is kept locked away in the labs that produced the data. The proposed rule could make communicating research findings harder in a number of ways: Make it harder for scientists to widely share their findings in scientific journals. Publication costs, including journal fees that allow for broad public access, could not be paid for by federal funds which is often how scientists pay these costs. Restrict researchers from engaging publicly about how their findings should inform policy or public health guidelines. This would restrict scientists from doing the work that transforms years of lab work into decisions that actually affect people’s lives. Make it harder for scientists to attend conferences. Conferences would now need to be approved by the federal agency at the time of grant award. This disadvantages conferences or publications that are not politically aligned. Conferences are a key way in which scientists learn about work happening in their field so they can collaborate and build on each other’s findings. What good is science with a muzzle? The long-term impact These changes could (and hopefully would) be reversed under the next administration. But by then, damage has been done. Once projects are halted or denied funding by political appointees, people leave their science careers for other paths or even leave the country to do their amazing science elsewhere. You cannot pause a scientific study the way you pause a subscription you decided isn’t useful anymore. You cannot freeze a research team that depends on salaries to feed their families, causing a scramble for replacement funding that, in most cases, simply does not exist. When the money stops, everything stops, and all the work in progress and its potential are lost. Despite all of the language about reducing waste, fraud, and abuse, this new proposed rule increases the administrative burden tremendously. Every researcher watching this unfold will make a rational decision: don’t start anything that will take a decade to finish. This includes not building a team around funding that could vanish, or asking the big questions that need patience and stability to answer. The most consequential science, the kind that produces the treatments and the breakthroughs that reach your doctor’s office twenty years from now, is exactly the science that would be abandoned first. This is still a proposal, so we can all do something Some worry this rule is already a done deal. But by law, OMB is required to open a public comment period. Those comments matter because they inform members of Congress, they can be used in litigation, and they create a public record. So here’s what you can do: Submit a public comment. The public comment period on this proposed change is open until July 13. Anyone can submit a comment, and OMB is required to respond to substantial ones. Each comment must be different to count as distinct; identical comments will be counted only once. Comment as an individual, not on behalf of your organization. Here’s a full resource to help you, with useful examples, but in general: Share a story about how science or federal funding has benefitted you or is involved in your work. Are you or someone you loved a cancer survivor? Do you know someone with Alzheimer’s? Have you received NIH/NSF or other federal funding? All of these can be stories you use to highlight why you care. Explain why you’re worried about these proposed changes. You can use the wording found here or in other resources to help you. Ask them to withdraw this proposed rule. Contact your elected representatives. Congress allocates money to federal agencies to spend, and it has oversight of how that process works. You can let your representatives know this proposed change concerns you and ask them to intervene. There are levers they can push that can help at this moment. Let others know why federal funding matters. This isn’t just about science, but ALL federal grants, including for education systems, defense projects, and more. The process, including the importance of federal funding, is invisible to many. We can all make it more visible by sharing how science benefits us and why we value it. You can also share this article, or any of the resources linked here, to help spread awareness! Bottom line We are quickly losing ground in scientific research in the U.S. More changes, like the proposed rule by OMB, would decimate our ability to conduct research here, leaving us all worse off: fewer treatments for pancreatic cancer, less research on pollution impacts, fewer answers for patients with rare disease, and more unaddressed health needs. The impact will be very real for all of us, regardless of where we stand politically. Using our voices for science is more important than ever. Love, LM Big thanks to Dr. Megan Ranney for reading this over and providing brilliant insights, as always. Dr. Liz Marnik, PhD, is Executive Director of The Evidence Collective at YLE and a science communicator and immunologist working at the intersection of public trust, science, and public health. You can find her on Substack, LinkedIn, Instagram and Threads. Your Local Epidemiologist (YLE) comprises a team of experts, ranging from physicians to immunologists to epidemiologists to nutritionists, working together with one goal: to “Translate” ever-evolving public health science so that people are well-equipped to make evidence-based decisions. The YLE suite of newsletters reaches over 475,000 people across more than 132 countries. This newsletter is free to everyone, thanks to the generous support of fellow YLE community members. To support the effort, subscribe or upgrade below: Subscribe
Gemini could soon get a lot better for multitaskers
Google seems to be working on making Gemini more useful for multitaskers by adding useful new features that could roll out soon.
We can create the future of science right now
Paul Litvak is the founder and Executive Director of the Robyn Dawes Institute and a Visiting Scholar at UC Berkeley. He has a PhD in Behavioral Decision Research from Carnegie Mellon where he studied emotions and the sunk cost bias. Over 15 years in industry, he solved a wide range of challenging technical problems. At Meta he created models optimizing the ad review process. At Google he ran experiments to measure social influence. At Airbnb, he was a product manager leading machine learning teams optimizing search ranking and price suggestions. He also co-founded and led product at Rhythmic Health, a venture-backed biosensing startup, creating an accurate low cost system using a color changing strip and a smart phone to measure salivary lactate. The bottleneck At this point, it is uncontroversial to say that science needs to stop using the PDF article as the unit of knowledge and currency. The unbearable slowness of scientific publishing, the profit motive and margins: I’m not saying anything new. The PDF also sucks because it’s hard to extract structured information, which makes it hard to do evidence synthesis. As a result, we do much less evidence synthesis than is needed. And evidence synthesis ultimately undergirds most policy and medical decision-making. I can see second-by-second real-time odds for any sporting or newsworthy event, but a school board can’t see the best evidence on whether their 8th graders should be taught algebra. As a society, we don’t treat this as an important problem. Again, not controversial. Not only is the problem well understood, but the solution has already been laid out. What we need is AI-assisted living evidence synthesis - (1) an open knowledge graph of atomic claims, (2) claims linked to evidence, (3) assessment and synthesis of each piece of evidence, and (4) continuous updating with new data. What few realize (yet) is that the technical capacity to build this vision for a significant portion of science already exists. Not only that, scientists and startup teams are already building many of these components. I know this because I’ve been surveying the space and talking to many of the builders. There are some missing pieces: for example, evaluations of how well some of the components work. But at this point, most of what’s missing is a fully end-to-end working integration of all of these parts. In the rest of this essay, I’m going to lay out all the parts of a working living evidence layer for science and who is working on them, and propose concrete next steps for building this system. What’s now possible The diagram above outlines the components of a living evidence synthesis platform, including some of the teams working on each component1. Scientific PDFs are processed into claims with associated evidence. The evidence is subjected to a forensic audit, methodological evaluation and robustness and reproducibility checks. Finally it’s given a weight in a continuously updating synthesis. What follows is a description and status of each component and a few of the teams working on them. Document understanding The first thing you need to be able to do is turn an article into structured data. Mostly that means parsing PDFs. There are often multicolumn layouts that confuse non-specialized PDF to text processing libraries. For scientific papers, there is the added complexity of parsing formulas and tables and figures. This is a really hot area - there are startups offering APIs, and it seems like a new open source package gets posted to Github every few weeks. What follows isn’t exhaustive. A package called GROBID was the state of the art for a while, they didn’t update their package for nearly two years until very recently. In the meantime reducto.ai released an AI powered PDF extraction API, PaddleOCR became popular, IBM released a model called Docling, and both Mistral and Gemini created models and libraries. I also know of at least one other well-funded psychology research group working on a paper parser. By contrast, there are few open evals in this space, with no extensive evals for complex table comprehension in particular. Nonetheless, I’m confident this will be a solved problem soon, given the combination of LLM advances and developer interest. Hypothesis level extraction There has been increasing interest in comprehending the extracted text of papers and linking information to evidence for each hypothesis. A lot of work has already been done. Trialstreamer (Marshall et al. 2020) and RobotReviewer LIVE (Marshall et al. 2023) demonstrated automated extraction of trial population, intervention, and outcome at scale on clinical RCTs. PaperQA2 (Skarlinski et al. 2024) and Ai2 ScholarQA (2024) extended this to retrieval-augmented question answering with citation grounding. Elicit, Consensus, and SciSpace operationalized claim-level extraction for end users. OpenEval (Booeshaghi et al. 2026) is the most recent and most ambitious: 1.96 million atomic claims extracted from 16,087 eLife manuscripts using Claude Sonnet 4.5, grouped into ~299,000 results, with LLM evaluations showing 81% agreement with human peer review on a 2,487-paper subset. None of these solutions link claims to test-statistics, as you would need to evaluate randomized controlled trials. This is why I built the evidence.guide API - to extract hypotheses and associated test statistics from behavioral science papers. The best public eval of this kind of extraction I’m aware of comes from the recent SCORE project - they had humans code thousands of psychology papers to extract their claims by hand. It would be extremely helpful to the world if all scientific PDFs were available as structured open data. I’ve been working to make this happen, both directly at Berkeley and through coordination with large entities I can’t yet speak of; as hard as it is to do, I think it’s possible2. Forensic audit A lot of work has been done on forensic audit, but some gaps remain. Of course, for biology papers that rely on images for evidence, there are a variety of tools (notably Proofig and ImageTwin) to spot anomalies. These are still well short of what sleuths like Elizabeth Bik can do on her own, but these tools are constant companions among fraud analysts. There’s someone working on auditing Excel files for anomalies, and a number of teams are automating numerical checks like GRIM and SPRITE, including Scrutiny project, the INSPECT-SR team as well as statcheck. The regcheck team is building a way to use AI to compare preregistrations to analyses in papers, to ensure there aren’t significant deviations. Nonetheless, there are many other kinds of anomalies to screen for, both public and less publicly known. And there are no formal evals for anomaly detection that I’m aware of. Still, there’s a lot to draw from in this space and I’m pretty certain we will be able to scan papers for most kinds of obvious anomalies in the near future. Methodological review This area has been white hot, though I fear for many of the startups in this space, because this capability may become commoditized. There are at least six different AI peer review companies, including Refine.ink, Reviewer3, ReviewerZero.ai, Q.E.D. Science, Paper Wizard, and Isitcredible. Coarse (a pun on refine) was also recently created as an open source alternative. These systems provide qualitative feedback on the content of papers, spotting methodological weaknesses and mathematical errors. They seem to work pretty well, and many academics report bitterly that they exceed the average quality of typical peer reviewers. But there are few evals here either. What evals exist so far involve using LLM-as-judge (circularity problems abound) or comparing against human reviews of questionable quality. What you’d ideally want is an eval that measures capturing known errors in papers3. Reproducibility and robustness Another active area has been using AI agents to automate computational reproducibility4 and robustness5 checks in papers that report numerical results. For more recent papers where data and code are available, AI agents can see whether they can re-run the analyses and produce the numbers reported in the published paper. In addition to a handful of individual academics who have been experimenting using Claude Code for this, the Institute for Replication is a leading group working on building an end to end system. The evals related to this problem are the most mature, with CORE-Bench (Siegel, Kapoor, Narayanan 2024) and PaperBench (OpenAI 2025) available to benchmark agents on this task. There is also work on getting AI agents to test alternative ways of analyzing the data to ensure the results are robust to small analytic design choices. Synthesis This is the most underdeveloped area where significant investment is required. Although some automated evidence synthesis systems exist — for example, otto-sr is building an AI agent to write systematic reviews — none of these incorporate the full range of paper level signals to weight evidence appropriately. Nor is there anything like an eval or a gold standard for a good systematic review. Arguably Cochrane reviews are the closest we have to gold standard human systematic reviews, though I’ve heard academics in the know complain about their uneven quality. A key question for a synthesis platform is how to weight anomalies and methodological issues in assessing the quality of a piece of evidence. This is an unsolved problem and one I’m very keen to work on. Continuous updating There are many pieces of basic infrastructure available for monitoring for new research and initiating updates. OpenAlex is the current open citation graph. Retraction Watch integrated into Crossref in October 2023. Scite tracks how citations support, contrast, or mention prior claims. The Living Evidence Network demonstrated continuous-update workflows in clinical guidelines. Engineering this is a relatively straightforward task. When you look over this technical architecture and all the progress being made, it’s hard not to be optimistic that a living guide to scientific evidence will be built. The stakeholders are ready The social infrastructure for this is starting to coalesce — it’s not just a pie in the sky academic exercise to imagine this coming into existence. Institutions like the Center for Open Science, the Institute for Replication, the INSPECT-SR, the Living Evidence Network and more are all working on scaling work to improve research quality. Funders are also aligned. The Sloan Foundation has funded living evidence work through COS. Coefficient Giving supports the Institute for Replication and COS. The Astera Institute and the Institute for Progress have shown interest in this space. NIH has established an Office for Replication and Reproducibility. Although there are (very unfortunately) serious headwinds in science funding generally, there is an active group of funders interested in metascience. A brief word about what I’ve been doing at RDI. First, as a Visiting Scholar at Berkeley I’ve been actively figuring out how a non-profit and a public university can conduct and make public the results of large-scale academic article data mining. With some of the money I raised from donors, I commissioned a legal analysis of recent case law and publisher text data mining (TDM) agreements in order to understand whether a massive open data mining of academic articles is possible (with caveats, it is). I’ve also been working to bring together stakeholders in this space, and identify gaps. I’ve also been doing some software development in this space, with more to come. A pilot proposal The assumption undergirding all of this is that an AI, given all this information, would make the right judgment about a scientific claim with lots of conflicting evidence, weighing all the factors appropriately. That’s the hypothesis we need to test. Randomized control trial research is the best place to focus on first. RCTs are used to make many of the important decisions in society - from medical trials to public policy changes. And they use a relatively uniform set of inferential statistics with lots of known and available diagnostics. Behavioral science experiments, within the broader realm of RCTs, should be first used as a testbed whose results can be generalized. Because behavioral science is at the vanguard of open science practices, replications abound (there are thousands of them) to serve as ground truth training data. Key Hypothesis Therefore the pilot would test, in behavioral RCTs that have been replicated, whether the quality of evidence for a claim can be used to accurately predict whether that claim will replicate. Secondary Hypotheses Compared to claims that replicated, non-replicated claims demonstrate a greater share of forensic anomalies in their source literatures. Hypothesis level claim and statistic extraction is accurate enough to scale living evidence without onerous human review costs. Replication prediction is more accurate than prediction markets6 or journal prestige. If all the different quality signals we gather do accurately predict which studies will replicate, then we can use that model to score evidence to power the living evidence layer. Why this is informative regardless of outcome If the pilot succeeds, the architecture extends to medical RCTs (where Living Evidence already operates and integration is mostly about claim representation), then to slices of basic biology with stable replication structure. If it fails, the field learns which quality signals are load-bearing and which ones metascience has oversold. Either result is a contribution to knowing what the literature supports. Implications for funders Because this burgeoning ecosystem of builders already exists, a well-informed philanthropic or government funder could play a crucial catalyzing role in bringing this future about. They could play at least three roles: creating open structured datasets, publishing open benchmarks and incentivizing the solving of key technical challenges. First, open archives of papers that the government maintains, like PubMed, could be turned into structured data amenable to large scale metascience and claim aggregation. I know the US government already has an interest in doing this, though some key open questions remain unanswered. How do you determine the best models and systems for accurately extracting information from papers? How can you establish a robust way to allow researchers to flag errors and correct them? And finally, how do you create a legal regime in working with publishers to maximize the scope of available papers? For the latter, a university or a private philanthropy may be better positioned to make structured data publicly available under journal subscription terms or fair use. Philanthropists or government funders could also coordinate to create or commission benchmarks that evaluate whether important problems have been solved. For example, an open benchmark for claim extraction from a range of different scientific article types would be extremely helpful. Ensuring the underlying data are accurate is vital for creating these evaluations. I’ve discussed opportunistically using various human-created datasets for this purpose, but a consistent problem is that errors in human data make it difficult for them to serve as a gold standard. Finally, with structured data and benchmarks available, the government or private philanthropy could use them to incentivize groups to develop machine learning systems that meet these benchmarks. Prizes are one potentially valuable tool for this. For example, you could establish a prize for a system that accurately updates a living meta-analysis for a small set of claims. Prizes are particularly useful as signals of problem importance, and can help create vibrant ecosystems of public and private research—see, for example the role the government played in kickstarting the current work on nuclear fusion. Conclusion The drawbacks of the current scientific publishing system are known. Scientists agree, metascientists agree, philanthropists agree: the published PDF plus citation graph isn’t the right substrate for maintaining a representation of the evidence base in science. The pieces needed to build the alternative either already exist or are rapidly taking shape. The community is forming around exactly this problem, with concrete partnerships and shared infrastructure. A pilot should start on behavioral science RCTs because that’s the slice of empirical science most amenable to legibility, where replication ground truth is richest, and where the failure modes are best documented. What’s been missing is the galvanizing mission to assemble these pieces into something that works. That’s what I’m proposing to build. *** 1 I have a broader field map that I’ll release publicly soon. This is me, building in public! 2 If this is something that you are excited about, please reach out and talk to me. 3 More on this very soon too! 4 This tests whether, given the code and the data, you can get the same statistics as reported in the published paper. 5 This tests whether the results are the same as a paper’s given alternative analytical decisions in conducting the analysis (like outlier omission). Closely related is the idea of a “multiverse” where you come up with many different ways of answering the same underlying research question with the same data, and test whether the results hold in all those alternative methods. There’s been work on the latter as well. 6 Some of the replications, e.g. those from the SCORE project, had paired the experiments with forecasts from prediction markets. So we get to look at this for free.
Beyond BMI: New Score Stratifies Obesity Risk
BMI, while useful, is insufficient to identify patients at the highest risk for obesity-related complications. A new tool may help clinicians better stratify risk and target treatment.
National Science Foundation grant puts Chicago at the center of an expanding water economy
A $45 million National Science Foundation grant will help nonprofit Current expand a Great Lakes program to turn wastewater into useful economic resources.
Why climate scientists have dropped a worst-case scenario
International scientists decided it is no longer useful to study the most extreme model of runaway global warming. So why are some people upset?
Amazon now generates images of fake products in one of the dumbest uses of AI yet [Video]
AI can be useful in a lot of places, but there are many scenarios where it doesn’t make sense. Amazon,...
May's top PlayStation games by copies sold
Yesterday’s State of Play gave us an idea of where PlayStation’s heading next. But we’ve had our May estimates for a couple of days now (we always update our data daily for clients), so it’s time to look at what performed best on PlayStation last month. We’ve got a new IP at the top, two evergreens doing their evergreen thing, FC26 smashing it ahead of the World Cup Plus pivot, and a quiet but promising result for Lego Batman. Let’s have a look at the state of Play…Station’s top games for May. 007 takes #001 007 First Light sold 1M copies on PS5 in May, easily making it the month’s top PS5 game by copies sold. Over 80% of those May copies were digital, which is broadly in line with where PlayStation’s digital share has settled for new releases this year. To date, 007 First Light has sold through 2.2M copies across platforms (so no unsold retail copies in that figure). The platform split is 55.1% PS5, 33.1% Steam, and 11.8% Xbox (console, Windows, and cloud combined). In gross revenue terms, 007 First Light has generated $150M. Owing to regional pricing and audience composition differences, the consoles make up a higher share of revenue than they do copies. Of the $150M, 60.5% came via PlayStation, 27.6% via Steam, and 12.9% via Xbox. Part of this is because over two-thirds of 007 First Light‘s PlayStation audience is in North America (43.4%) or Western Europe (23.2%), where local currency means the game generates more revenue per player. On Steam, under half of First Light‘s audience is from those same regions (31.8% North America, 15.1% Western Europe). As we covered in last week’s 007 Steam piece, the Steam version is pulling in higher-than-expected numbers in China, which now accounts for 17.3% of First Light‘s Steam audience. 007 is the equivalent of $44 USD in China, which is roughly 37% cheaper than the US price and about 44% cheaper than Western Europe. Subscribe for free to get the most accurate games estimates in your inbox twice a week 🫶 Subscribe The crossover data backs up what we’ve seen all year about launch-window cohorts. 11% of PS5 players who played 007 First Light in May also played Lego Batman: Legacy of the Dark Knight, which launched just a few days before. Last month was a great month for Denmark! IO Interactive is Copenhagen-based, and Lego Group is headquartered in Billund (TT Games is UK-based but produces work under Lego Group licence and direction). Speaking of which… A very blocky #2 and #3 Minecraft sold another 788K copies on PlayStation consoles last month. Business as usual, since Minecraft hasn’t sold fewer than 760K copies in a single month all year. With over 350M copies sold across all platforms, Minecraft is the best-selling game in history and one of the strongest brands in the world. Every month, a new cohort of children reaches Minecraft age, which guarantees a constant volume of new sales regardless of what the broader market is up to. The demographic refresh rate is truly impressive, with the audience renewing itself faster than it’s churning. Lego Batman: Legacy of the Dark Knight shifted 757K copies on PS5 last month, bringing its lifetime total on the platform to over 800K. Across all platforms, Legacy of the Dark Knight has now passed 1.5M sold through, so PlayStation accounts for over half of the game’s players. The May crossover data for Legacy of the Dark Knight on PS5 is also useful. 27% of those players also played Fortnite, 16% Roblox, 16% FC 26, and 14% NBA 2K. That’s the household-and-family cohort showing up, plus the launch-window cohort of core gamers who buy practically everything. FC 26 gets players through the door ahead of the World Cup FC 26 sold 411K copies on PlayStation last month (90% via PS5), the fewest copies FC 26 has sold all year and down heavily from April’s 700K. Part of this dropoff is because FC 26 has been available as a PlayStation Plus monthly game since 6 May, cannibalising premium sales. Subscribe for free to get the most accurate games estimates in your inbox twice a week 🫶 Subscribe The strategic logic is straightforward for EA, which is sacrificing premium unit sales in May to build up the player base ahead of the FIFA World Cup, which kicks off on 11 June. Obviously, a bigger player pool going into the tournament means a bigger pool of potential Ultimate Team spenders during the period when football attention is at its peak. The early returns are around what you’d expect, with 4.4M new players accessing FC 26 via PS Plus last month – a substantial new spender pool. Anecdotally, a few of my mates who were ‘’skipping FIFA and Ultimate Team this year’’, were drawn back in via PS Plus. Lo and behold, one of them is buying packs like it’s going out of fashion. Whether this is all a smart trade for EA depends entirely on the Ultimate Team conversion rate over the next six weeks. Either way, the short-term engagement impact on the game as a whole is evident. FC 26‘s MAUs hit a record 24M on PlayStation in May, up 30% from the previous record set in April. GTA V does what GTA V always does Rounding out the top five is another evergreen title. Grand Theft Auto V shifted another 387K copies on PlayStation last month. No matter what else is happening in the industry, GTA V keeps pulling players in. It’s sold another 1.8M copies on PS5 in the first half of this year alone, helping the franchise pass the 230M-sold milestone across all platforms. Subscribe for free to get the most accurate games estimates in your inbox twice a week 🫶 Subscribe I have a theory that Minecraft and GTA V are doing the same thing at different demographic levels. Both are functioning as generational gateway games that the audience ages into faster than it ages out of. Minecraft captures kids around age 6-10, GTA V captures much of the same audience around age 12-15 (unofficially, of course). The crossover data suggests this: 40% of GTA V‘s monthly PlayStation players typically play Fortnite in any given month, while 30% play Roblox. Household accounts aside, that’s not the original GTA V audience from 2013. Of course, the under-12 audience would love to get their hands on GTA, too. It’s always been the way of things, and some parents are happy to oblige. Mine did too back in the day, and I turned out just fine. Right? Right!? Want to get your hands on our data for yourself – for free? We’re offering a free trial of our platform for games companies. Send us a message here, or reply to this email, and we’ll set you up. The last word Reply to this email – or reach out here – if you have any feedback for the newsletter – or want to request a game for us to cover. [Alinea Analytics boasts the most accurate PC and console estimates in the business. Game makers use our platform to understand their audience, keep an eye on the competition, monitor sales trends, and spot new opportunities. We equip game studios and financial institutions with accurate data and the confidence to make smarter, data-driven decisions. Want to talk about all things games market data? We’d love to chat!]
The cult of objectivity has broken consciousness science
With 22 competing theories and no consensus in sight, the science of consciousness is lost. In this exclusive IAI interview, Berkeley psychedelics researcher and bestselling author of a new book on consciousness Michael Pollan insists that even though consciousness lies beyond traditional forms of scientific investigation, we should resist becoming dualists who see it as something radically different from ordinary stuff in the world. Instead, he makes the case for reimagining science itself: it must stop reaching for objective descriptions of physical reality, and instead embrace subjectivity as the indispensable core of all inquiry. Alasdair Craig: How did your experiences with psychedelics get you interested in consciousness?Michael Pollan: One of the things that happens on psychedelics is that they foreground consciousness. What formerly seemed perfectly transparent and ever present—and yet, non-existent, as water is to fish—suddenly you see it. The metaphor I use in the book is that psychedelics sponge the windscreen. Suddenly you realize: wow, there’s a windscreen, there is this phenomenon that’s mediating my relationship to reality. This is not an unusual realization on psychedelics. I think a lot of people are called to think about consciousness after psychedelic experiences. So that was the proximate inspiration for A World Appears: A Journey Into Consciousness (2026).And then there was a more specific way in which psychedelics influence the book. That was a particular experience I had on psilocybin. I had the sensation that the plants in my garden were conscious and returning my gaze, with their own point of view and agency, and I got curious as to how to evaluate an insight like that. I mean, what’s the truth quotient of insights you have on psychedelics? And I was reading William James—who, of course, wrote beautifully about mystical experience—and he said, we can’t judge the metaphysics here, but we should pay attention to two criteria: (1) How useful is it to think this is true? He was a pragmatist, of course. And (2) can we subject it to other ways of knowing to validate it? Because maybe we should treat these insights as hypotheses. And that’s kind of how I approach psychedelics, as a great generator of hypotheses, but they need to be tested against other ways of knowing. And so that led me down this long path exploring the science of plant intelligence and plant sentience.AC: I love your phrase, “smudging the windowpane of conscious experience.” So do you now start to think, actually, even when I’m not on psychedelics, I’m just seeing a screen all the time, which is not reality, but rather is my conscious experience, behind which reality is hidden?MP: Well, you know, not when I’m brushing my teeth and making breakfast! But yes, in meditation, certainly. I mean, one of the things you realize is that my experience happens to be this way now, but it could be another way. And many things can affect it. My morning coffee changes the screen in various ways, sharpens certain things.___The only tool we have to understand consciousness is consciousness.___There’s this wonderful metaphor in Aldous Huxley’s book, The Doors of Perception, where he thinks we channel consciousness, and that what the brain does is edit it, and there’s a reducing valve. And psychedelics open the reducing valve wider. I don’t know that I’m ready to buy into the idea that consciousness exists outside brains, but I do love that metaphor because there are times when you are more conscious than others. I know I’ve switched from window pane to valve, but you know, a camera aperture can open wider or narrower. And I do have a sense of that. I have a sense of when I’m being less conscious, and when I’m being more conscious, by which I mean, you know, more present, taking in more sensory information. So when you start thinking about consciousness as a matter of degree, that changes things also.I think when we scroll on social media, obviously, we’re conscious. But minimally so, because we don’t have to generate our own thoughts. We’re basically taking these little hits from other people and thinking their thoughts.The process of writing this book made me kind of relativize everyday consciousness as just one type, which is exactly what William James said. Normal consciousness is one type, separated by the flimsiest of screens from the other types.AC: If we stick with the metaphor of the smudge on the window that you get when you take psychedelics, it seems like we’re saying that consciousness has become like an object that you’re looking at, right? You’re looking at the smudge and that just is consciousness. But sometimes we’ll talk about consciousness as a state that you’re in, rather than an object you can perceive. You’re either conscious or you’re not conscious. So it’s difficult to really know what sort of thing we’re talking about when we talk about consciousness, right?MP: Yeah. When you say that consciousness is an object, what’s the subject? Consciousness also? The only tool we have to understand consciousness is consciousness. So we’re in a closed loop and when we get a perspective on consciousness, we’re not achieving any distance from it. That’s one of the problems with studying consciousness. Science is predicated on an objective, God’s eye view of things, and you never get that. There’s no such thing. SUGGESTED VIEWING The brain filters consciousness With Àlex Gómez-Marín But, yeah, there is this sense, I think, that consciousness is something you can approach. Certainly, when you’re writing and trying to recreate a conscious experience, it’s a fiction, but you’re standing back from it, and looking at it. But there’s no real subject-object distinction.AC: That’s so interesting, because when philosophers start talking about consciousness, they often talk about what they call “qualia”, which is their word for the felt experience of, say, the color red. And I think that way of speaking often makes us think of the qualia as an object, which you can see out there: the redness of red. But, as you say, we also want to collapse the subject-object distinction when we talk about consciousness. So there’s something strange going on here…MP: Oh, yeah, there are many strange things going on! *AC: Let’s talk about the varieties of consciousness. Apart from perceptions, there are, pains, emotions, thoughts, and so on. What is it that holds all this diversity together so that they all count as instances of consciousness? What is it that they all have in common?MP: Yeah, that’s a really good question. I mean, there is sensory information, right? Which we perceive coming from the world outside ourselves. And then there’s internal information, which is what’s called interoception. And this is very important too. This is where feelings come from—what are sometimes called homeostatic feelings—that are intended to get the brain to keep us in these narrow ranges of temperature and glucose and so on. That’s a very important part of consciousness. And according to some researchers, who I think are probably right, this is where consciousness begins, with these very primal feelings, like hunger and thirst and warmth and itch, and things like that. And then there’s this third thing, which is our thinking about thinking.And what holds all this together is, I think, the sense of self. Consciousness generates a self: it is probably an organizing principle. All this stuff is happening to the self. It’s a construct, I think, but it is very useful. It can connect all these different streams of consciousness to a single being or a place, metaphorically speaking.* AC: The first half of your book is about the science of consciousness. Is consciousness just another thing, like black holes or the origin of life, which we don’t yet understand, but which we should have every hope that science will one day get to the bottom of, or is there something special about consciousness that makes this unlikely?MP: It depends how you define science. As we define science now, it’s going be very difficult. Science since Galileo has had a set of rules or assumptions, which is that the proper remit of science is to deal with the objective, quantifiable, third-person world. And the physical sciences have done amazing work under that set of limitations. Galileo—for tactical reasons, not for philosophical reasons—said science should not deal with the soul, which at the time was what consciousness was considered to be. So what is subjective, what is qualitative, what is first person, we’re going to leave to the church, he said. And this was a good agreement, and it probably kept many scientists from getting burned at the stake. But it leaves science in an awkward spot, now that it wants to venture into that territory of qualities, of first-person experience. And I don’t think it has the proper tools. I think it will have to add to its toolkit in some way, and find a way to bring in phenomenology, or lived experience and introspection, which science is usually dismissive of.The challenge—we alluded to it earlier—is that the only tool you have to study consciousness with is consciousness. And what I mean by that is, the whole scientific enterprise is a manifestation of consciousness. As much as scientists like to flatter themselves with this idea that they attain a godlike perspective on what they study, when it comes to consciousness, there’s no such thing. And it’s fine for all the things you can step outside of and study from that perspective. But two areas it won’t work are consciousness and cosmology.___It could be that we live in this anomaly of not believing nature is conscious. And maybe psychedelics are teaching us something we’ve forgotten.___With cosmology, too, you have to study the universe from inside the universe. It’s not that cosmologists haven’t made progress. We know a lot about the age of the universe and expansion and things like that. But we can’t penetrate what happened before the Big Bang. We just don’t have the tools because we can’t stand outside.The fact that there are 22 full-scale theories of consciousness tells you all you need to know: we’re kind of lost. And I was surprised to find that we are more lost than I thought we were when I started this. I thought that there’d be leading theories, or that we had experiments that we could conduct that would nail some of this down. And the fact is, we don’t.So, I don’t know if science can solve it. But, of course, we can change the rules of science. A really influential book for me was a book called The Blind Spot, by two physicists, Marcello Gleiser and Adam Frank, and a philosopher, Evan Thompson. It’s a reminder that science is a social practice. It’s organized in a certain way and it can be organized in a different way. But it has this huge blind spot at the center of it, which is its failure to recognize that it is a form of consciousness itself, and it doesn’t have a good way of taking account of lived experience.AC: On the one hand, I think you want to resist traditional dualism, which would say that consciousness is something magical, something supernatural, right? But on the other hand, you’re accepting that there is this division in reality between the stuff that you can investigate objectively and measure, and so on—the stuff that science has been good at finding out about—and consciousness, which, for some reason, you can’t objectively measure. Some would take this to suggest that the dualism is actually right. Is there a way of reconciling these two strands of your thinking?MP: To call it magical is full of value judgement. I mean, you could just say immaterial. The interesting thing is, I think, that intuitively we’re all dualists. And I don’t know whether that’s because we’re children of Descartes and other cultures are less dualist than we are. Or it may be because that’s how consciousness feels. It feels different than the body. It feels separate from the body. But I don’t think we should trust that intuition.AC: So let’s think about ways that science might be changed to get at consciousness. I suppose the starting point would be the way that you begin the second half of your book. You have this nice line from William James, where he says, “We now begin our study of the mind from within.” Could you give us some sense of the ways of studying the mind from within? What does this involve?MP: I don’t know what that science looks like. But one of the points I’m making is that it depends on the humanities. I found that I learned as much, if not more, about consciousness, from reading novels and poetry. There is a sense of the fullness and complexity of consciousness that you don’t get in any scientific account. I interviewed this stream-of-consciousness writer and I approached her as an expert on consciousness, and she is. So I think that dialogue is really important.I’m reading Ben Lerner’s new novel, Transcription. Just read the first page of this novel: it’s a man falling asleep on a train, going to conduct an interview, and there is the experience of being in the train, facing backwards, there’s the view outside, but there’s a reflection of himself through which he’s seeing outside. There’s the memory of dropping his kids off at school. All this is going on in one moment of consciousness. Scientists aren’t going anywhere near that. But I think it’s useful for them to think about. That’s where we’re trying to go: like, how does such a layered experience happen? You know, it also makes you wonder, could a computer ever do this? And it gives you some doubt.*AC: Before we get to computers, perhaps we could talk briefly about plants. How did you come to feel that they are sentient?MP: Well, it was a psychedelic experience, and it’s not unusual on psychedelics. In fact, there’ve been studies about how people become animists during psychedelic experience. Their belief in animism arises after a single psychedelic experience.You know, most scientists would say this is all a myth. But we have only 400 years of the age of enlightenment and science. Before that, we were animists, and if you go around the world, most indigenous cultures are still animists, and children are animists, by and large, until they go to school. So it could be that we live in this anomaly of not believing nature is conscious. And maybe psychedelics are teaching us something we’ve forgotten.As I said, I wanted to test this idea, and see whether there was conventional scientific evidence for it. And there’s fascinating work going on in a field that calls itself—in a rather cheeky way—plant neurobiology. These are people who fully realize there are no neurons in plants, but they’re trolling conventional botanists! And they’re doing experiments that are demonstrating that plants can learn and remember, which is kind of amazing: how do you store a memory without neurons? They’re finding plants that can solve problems, that can hear and see. In fact, they have 20 senses to our five or six. For example, a bean plant looking for something to twine, and given a pole within eight inches or so of it, knows where that pole is and will aim its spiraling reach at that pole. What is that about? Is it bouncing signals off of it?One of the spookiest things is the fact that the same anesthetics that put us out put plants out. You might think that plants are already out, but they’re not. They’re just slow. If you take a plant with a fast behavior, like a Venus flytrap, and give it an anesthetic, it will stop reacting. So you have to ask yourself, what is lost when the plant is anaesthetized? Some state of alertness, of awareness, of its environment, of reactivity. I’m not prepared to call that consciousness, but maybe it’s sentience—which is a simpler form of consciousness, and involves the ability to sense what’s going on in your environment and assess what’s good and bad for you and act appropriately. That may be a universal quality of life.AC: Let’s turn from plants to AI. You point out that LLMs are trained on the internet rather than direct interactions with reality. You put this really nicely when you say the internet is not the world so much as a shadow cast by the world. So LLMs are living in this kind of shadow world, or in Plato’s cave, as you put it. What’s the significance of this for AI consciousness and moral consideration? And, to recall the beginning of our conversation, if we’re stuck behind a veil or screen of consciousness, are we also stuck in a kind of shadow world?MP: I don’t think they’re the same shadow worlds because our world includes nature and other people. And even if it’s mediated in some ways... There is a reality out there, and we are in relation to it, through consciousness. There’s friction with it. What’s missing from the LLMs is that friction with the real world. There’s more to the world than human representation, which is what they’re dealing with. And that means that so far, AI is a cultural technology for combining thoughts we’ve already had. SUGGESTED VIEWING Cracking the code for thought With Roger Penrose, Sabrina Gonzalez Pasterski, Max Tegmark, Güneş Taylor Genuine creativity is more than collage. A lot of what AI does now is tantamount to collage, putting this with that and you know, that’s part of art, certainly. But there are also genuinely new ideas that come into the world that don’t come in through manipulating the representations we already have. So it remains to be seen whether AIs can be truly creative in that more radical sense.But my arguments against AI becoming conscious have to do with the fact that the idea of AI consciousness is based on this faulty metaphor that the brain is a computer. There are dimensions in which the brain is like a computer—it certainly does its share of computation. But computers have this sharp distinction between hardware and software, and brains don’t. In a brain, hardware and software are identical. Every memory is a physical change to the brain. Brains are not interchangeable in the way computers are. So this idea that you could take the consciousness out and somehow run it on another substrate, a silicon substrate, seems foolhardy.Anyway, it’s a more extended argument, and I may be wrong. And when we move to computers that aren’t based on silicon—neuromorphic computers and things like that—all bets may be off.AC: Regarding your point about LLMs being stuck in the world of our representations—the internet or whatever—and not being able to interact with reality, are these representations, the world of the internet, themselves part of reality? We’ve created the internet, but it’s still a real thing that exists. And so while the LLMs may be trapped within that part of reality, they’re nevertheless within reality and interacting with it and so on. So it’s not obvious that there’s a barrier there to them becoming conscious and being creative.MP: Yeah, I mean, if our consciousness is built out of human culture, which is what we’re talking about with all those representations. But I tend to think consciousness comes a bit more out of nature and friction with the natural world, but...AC: So there’s a fundamental divide between the natural world and human artefacts?MP: And we lose track of that, and especially people in this computer world. You know, the people who are creating these AI systems, my guess is they’ve been into video games since the age of three, and they didn’t go outside a lot. And their own contact with nature is somewhat limited. I could be wrong, it’d be a very interesting thing to ask, all the wizards of AI, whether they like to go fishing and hiking and do things like that. Or were they on computers from a very early age? I interviewed Blake Lemoine, the Google engineer who came to believe that LaMDA, the Google LLM out at the time, was conscious and deserved personhood. It was very clear that he didn’t see a distinction between this bubble in which he lived and the real world. And I think we’ve all been distanced from the natural world, and this is taking us even further away. But it’s still there, you know, and we still have bodies, and we’re still mortal, and we don’t live in a simulation.AC: I think that’s a great place to finish. Michael Pollan, thank you very much!
Scientists May Have Found a Way to Detect a Third Type of Magnetism
The third type of magnetism could prove the most useful of all—if researchers can identify the materials that exhibit it.
Nvidia chases $200B CPU market with AI agent PCs from Microsoft, Dell, and HP
If Nvidia has cracked a way to bring AI agents easily, safely and usefully to the masses, it could — and should — be big.
I put Google's 24/7 AI assistant Gemini Spark to work, and it's actually pretty useful
Gemini Spark helps automate everyday tasks, from inbox summaries to local event planning, but it’s unclear why Google made it a separate product.
Behavioural science has a practical lesson for us to fight back against transphobia
The behavioural science concept of ‘extinction bursts’ might be useful for the Trans+ community
AI Chatbots Do Not Consistently Deliver Accurate Health Responses
While AI chatbots delivered useful information, the error rates remain high enough that they should not replace physicians.
KeySmart’s new Android Find Hub wallet tracker is faster, has 2 years of battery life
Wallet trackers might be the most useful devices you can use with Android’s Find Hub network, and KeySmart’s new option...
Frederick’s live music scene is having a moment
Wave theory isn’t just for physics; it’s also a useful way to think about Frederick’s music ecosystem. The scene tends to move in cycles, with collective musical energy cresting and
Therapists are using AI to take notes. Is it a useful tool or a breach of trust?
New companies are selling artificial intelligence assistance to mental health therapists. The AI tools can help with administration and recordkeeping, but some patients worry about their privacy.
Bond breakout? Big effects of bad data. Immigration and thermostatic politics & dad-chat is coming for dad-books.
Thank you for opening your Chartbook email. Anna Airy, A Shell Forge at a National Projectile Factory, Hackney Marshes, London, 1918 Bond breakout - John Authers Across the developed markets, bond markets are staging a slow-motion car wreck. Looking at the drastic moves in the US, Japan, the UK and France, Barclays’ Ajay Rajadhyaksha commented: “Last week, long bonds broke. Not in one country. In all of them. Simultaneously. Four countries. Four different political systems. Four different central banks. But the same trade — “get me out of duration!”” Each country has its own contributing political problems, but the uniformity of the shift out of long-term bonds with high duration makes clear that something broader is afoot. For Rajadhyaksha, it is “simple and uncomfortable.” The developed world “has too much debt, too little fiscal discipline, and no political appetite for fixing either.” If there’s a trigger, it has been the Iran war and the resulting shock to oil supply. Source: Bloomberg G7 countries are now borrowing at rates that were enjoyed by EM in the early 2020s For contributing subscribers only. Subscribe Trade between China and Russia has grown in recent years For contributing subscribers only. Subscribe Big effects of bad data A decade ago nearly nine in ten Americans, when approached, agreed to fill out the Current Population Survey, which is administered to about 60,000 households each month and asks about, among other things, employment. Fewer than seven in ten do so now (see chart 1). For the Consumer Expenditure Survey, which tries to capture 3,700 households monthly, the response rate is down from 68% to 40%. … When the information becomes less reliable, investment decisions get more difficult to make. Some may be delayed, creating a potential drag on the economy. A new working paper by Nicholas Bloom of Stanford University, Erica Groshen, former head of the Bureau of Labour Statistics (BLS) now at Cornell University, and Duncan Hobbs and Michael Strain of the American Enterprise Institute, a think-tank, hazards an estimate. Preserving trust in “the integrity and quality of official statistics”, the authors claim, generates economic benefits of about $25 for every $1 spent on the BLS, the agency with an annual budget of $700m that is responsible for many of these data. To arrive at their conclusion the quartet analysed an ignoble episode in the BLS’s recent history. On August 1st 2025 Donald Trump sacked Erika McEntarfer, appointed as the agency’s commissioner by his predecessor, Joe Biden. The president alleged, without evidence, that the BLS’s steep downward revision to recent jobs numbers had been “RIGGED in order to make the Republicans, and ME, look bad”. In fact, it was the wanton dismissal that looked bad in the eyes of many observers. In the following seven days there was a 50% leap, relative to the week before, in the average value of the index of Economic Policy Uncertainty (EPU), which tracks the number of articles mentioning such uncertainty that are published daily in American newspapers. This was a discernible jump even when compared to the chaos caused by Mr Trump’s trade war in April 2025 and his real one in Iran in the past two months (see chart 2). Based on Mr Bloom’s earlier study with other colleagues of the EPU’s impact on business investment, industrial production and employment, the authors estimate that the jump reduced American GDP by over $100bn (0.3%) and non-farm payrolls by 168,000 (0.1%). Source: The Economist Immigration and thermostatic politics (Sarah O’Connor in the FT) A strange thing happened a year and a half ago. No sooner had Americans elected Donald Trump as president on an explicitly anti-immigration platform than US public opinion began to swing much more favourably towards immigration. … to political scientists who study “thermostatic” politics, this development is no surprise at all. Public opinion often moves against the prevailing party, especially if they are perceived to have gone “too far” in one direction. The phenomenon was first explored in the 1990s with reference to fiscal policy, but in recent years it has been particularly evident in the realm of immigration policy. While these swings in opinion polls can make it seem as if people don’t know their own minds, they can be a useful democratic check on governments by sending bright flashing warning lights when policies stray too far from the public’s comfort zone. The problem comes when politicians misinterpret those signals. When people tell pollsters they want less immigration, or think it is bad for the country overall (or the converse), they are often expressing an opinion about how things appear to be going in that moment relative to their preferences, rather than a deep shift in their underlying views. According to Alexander Kustov, a political scientist, most Americans’ views are quite stable and moderate: they support immigration that is controlled and in the national interest, and oppose flows that are disorderly or under-enforced. In the US, he argues the Trump administration over-interpreted dissatisfaction with high immigration under Joe Biden, and implemented policies that went too far for many voters. “People don’t necessarily change their idea about their ideal immigration policy, but they can react to what the government is currently doing,” he told me. “A lot of people are not happy with what the Trump administration are doing — right now we’re not talking about border enforcement, we’re talking about people not just getting harassed, but killed.” But similarly, he added, pro-immigration advocates would be making a mistake to interpret the recent, more favourable swing towards immigration as a deep-seated ideological shift in attitudes. The recent history of UK immigration policy offers another cautionary tale. After the 2016 Brexit referendum, public concern about immigration fell. This was partly because net migration statistics declined, but also because people felt a sense of catharsis and control, says Sunder Katwala, director of British Future, a think-tank. By 2019, the public ranked immigration as only the ninth biggest issue of concern, according to Ipsos polling. But when the Conservative government liberalised visa routes for students and care workers between 2019 and 2022, that helped net migration to surge. And so, too, did public concern. Today’s Labour government has implemented another sharply restrictive turn in immigration policy. As Kustov and researcher Caitlyn Yates argue, thermostatic public opinion should be a useful check on politicians. Instead it seems to have helped to magnify the pendulum swing in the UK. The government over-interprets and overreacts to a change in public opinion, which reacts in turn, prompting a sudden swing the other way. Source: Financial Times HEY READERS, THANK YOU for opening the Chartbook email. I hope it brightens your day. I enjoy putting out the newsletter, but tbh, what keeps this flow going is the generosity of those readers who clicked the subscription button. If you are persuaded to click, please consider the annual subscription of $50. It is both better value for you and a much better deal for me, as it involves only one credit card charge. Why feed the payments companies if we don’t have to. Subscribe Anna Airy, Shop for Machining 15-Inch Shells: Singer Manufacturing Company, Clydebank, Glasgow, 1918 Racing for a vaccine … we have been here before! The deadly Ebola outbreak in two African countries driven by a virus species with no vaccine has triggered an urgent search for a potential jab, in a test of troubled global efforts to avert threats posed by emerging diseases. A World Health Organization advisory group is due to would meet on Tuesday to recommend candidate jabs to prioritise for clinical trials, the global health body said. It will assess data including an analysis by the Coalition for Epidemic Preparedness Innovations (Cepi), which was set up after failures in the international response to a previous Ebola crisis. The latest Ebola outbreak in Democratic Republic of Congo and Uganda has infected hundreds of people and killed more than 80, health authorities say. Cepi’s reaction will be a crucial indicator of its progress towards its wider goal of producing vaccines for pandemic threats in just 100 days, despite steep cuts to international health funding. “If there was ever a time that we could show the world why Cepi is needed and show the world why the 100-day mission is needed, it’s now,” Nicole Lurie, the organisation’s executive director for preparedness and response, told the FT. “We’re happy to accept that responsibility, but obviously we need help from partners — particularly financial help in the long run.” DR Congo and Uganda would make the final decision on whether to press ahead with any vaccine candidates endorsed for clinical trials by the WHO experts, the global health body said. “Other ethical and community acceptability issues will be considered,” said the WHO, which declared the latest outbreak a public health emergency of international concern on Sunday. “This will be important to make sure that the trial, if it happens, will be adequately communicated to the population.” Lurie said Cepi had provided information from a survey of research teams and companies working on drugs that might be effective against the Bundibugyo virus behind the outbreak. The organisation hoped soon to announce partners in the quest for a jab, she added, although she declined to give a timescale for when one might be developed. Cepi’s Nicole Laurie said global health bodies were ‘getting some live-fire drills’ with hantavirus and the latest Ebola outbreak. The hunt for a vaccine should be helped by scientific advances including the use of artificial intelligence to find drugs active against viruses, Lurie said. Existing jabs against the Zaire Ebola virus responsible for many previous outbreaks might provide a basis for tackling the Bundibugyo species, she added. “We’ll see how far and how fast we can get,” Lurie said, adding that Cepi had done preliminary work recently on a possible jab for the rat-borne hantavirus responsible for a recent fatal outbreak on a cruise ship. “Both with hantavirus and now with Ebola, [we] are getting some live-fire drills.” Source: Michael Peel in the Financial Times Dad-chat is coming for dad-books. (By Pamela Paul and Jeffrey A. Trachtenberg in the WSJ) They were the go-to gifts for Father’s Day: a book about some little-known chapter of World War II, the sweeping narrative of a shipwreck, perhaps the latest presidential biography. These days, dad books are a dying breed. Nonfiction book sales have been in decline for the past four years, and are now the most challenged segment of the print book market. Publishers say certain types of books still fare well—including celebrity memoirs and religious titles. But in recent years, print sales in such categories as biography, current affairs and business and economics—what publishers refer to as “serious nonfiction” and which tend to resonate especially with men—have fallen considerably. Sales of nonfiction print titles were down nearly 8% through May 9 this year, and sales of books about politics and current affairs were down 19%, according to book tracker Circana BookScan. Publishing has long been subject to cyclical changes, with trends in format and genre taking over bookshelves in grand sweeps. (Remember the adult coloring-book craze? Or the stratospheric rise of romantasy?) But many in publishing believe the decline in serious nonfiction is more existential. “The trend couldn’t be clearer,” said Jonathan Karp, the former chief executive of Simon & Schuster and publisher of the new Simon Six imprint. “This is a sea change and people should wake up and realize we’re living in a new world.” One culprit for the shift: a new media diet for fans of more serious nonfiction. … Sixty-two percent of men and 54% of women consumed a podcast in the prior month, according to a recent survey by Edison Research at SSRS, up from 46% and 39%, respectively, in 2023. “When we have internal meetings to talk about this problem, it always comes around to podcasts,” said Jonathan Burnham, president and publisher of the Harper Group at HarperCollins Publishers. “The man who wants to read American history is now tuning into one of the many good podcasts about history that lends the quiet attention to a serious subject he’s looking for. It makes the idea of sitting down with a 700-page Ron Chernow book less appealing. You’ve scratched that itch.” Source: Wall Street Journal Remarkable report by Novara from the nationalist rally in London in mid May. Anna Airy, Leopard Moth and Plums, 1915 If you’ve scrolled this far, you know you want to click: Subscribe
AI & Science: What Is the Future of Discovery?
Continued progress in artificial intelligence, its expanding usefulness in science, and its contributions to landmark advances suggest that we may have entered a new era of AI for science.The breakthroughs so far—such as predicting the structure of practically every known protein, with profound implications for our understanding of biology, health, and the treatment of disease—are notable not only for what was achieved but also how it was achieved and what that suggests for scientific progress. This special double issue of Dædalus poses the question: What is the future of scientific discovery in this new age of AI? Thirty-three scientists responded. Bringing perspectives from life sciences and medicine, cognitive science and neuroscience, the physical and earth sciences, chemistry and materials science, computer science, mathematics and the social sciences—they draw on their work at the frontier of AI and science. The authors write with an eye to the future, not just the present. They explore what is being achieved and what possibilities lie ahead; examine AI’s limitations and efforts to move forward; and investigate the larger implications of AI-assisted science—on how science is done, the role of the scientist, and the scientific method, as well as the challenges and complexities involved.The authors together exemplify a long-standing bidirectional relationship: AI advancing science, while science advances AI. Where that relationship will take us—a golden age of discovery? New scientist-machine collaborations? Autonomous labs? Discoveries without human understanding?—is a future we are only beginning to imagine, and one we must also shape if the beneficial possibilities are to be realized.Image: UMAP sculptures visualizing three Dædalus volumes on artificial intelligence. © 2026 by Refik Anadol.
Opinion | MAHA Is No Longer Useful to Trump
Robert F. Kennedy Jr.’s MAHA movement has reshaped the public health debate. But if the coalition behind it begins to fracture, what comes next?
The mysterious reason why women get hotter from age 18 to 42
Women experience a steady rise in body temperature from their teens to midlife, which may be useful for monitoring ageing and overall health
How Google Is Starting to Win the A.I. Race
Despite its early stumbles, Google’s Gemini has leapfrogged ChatGPT in relevance and usefulness. Soon, it will be ubiquitous.
DNA floating in seawater is now enough to let scientists monitor the health of America’s dolphin populations
Sequencing environmental DNA in seawater shows the local presence or absence of species, but until now could give little information about those measures of biodiversity that are the most useful in conservation. As a proof-of-principle, scientists in the US have now shown that mitochondrial DNA in water sampled near schools of dolphins contains enough information to measure their local effective population size and monitor the health of these populations.
New MatterChat Model Helps AI to ‘See’ the Language of Science
From writing emails to generating computer code, much of the artificial intelligence prevalent in our daily lives has succeeded by mastering one domain: text. However, this leaves a major blind spot in the physical sciences, where models depend on the high-resolution, three-dimensional data of the physical world, like the intricate lattice of atoms in a crystal. Delivering on the promise of using AI for science requires teaching these data-driven text models to seamlessly “talk to” physics-based models. Now, a new AI framework from Lawrence Berkeley National Laboratory (Berkeley Lab), called MatterChat, solves this problem by creating a specialized “bridge.” It connects the conversational power of a Large Language Model (LLM) with a physics-based AI that models “interatomic potentials”: the complex physical forces between atoms. The resulting system already significantly outperforms general-purpose AI tools like GPT-4 at predicting material properties, and the team hopes it can accelerate scientific discovery by serving as a robust research partner that provides grounded insights and generates step-by-step instructions for synthesizing novel materials. A paper describing this work was recently published in Nature Machine Intelligence. “Traditional simulations can provide the physical rigor required for materials science, yet their computational cost remains prohibitive for high-throughput screening. Conversely, while LLMs excel at rapid knowledge synthesis, they inherently lack the ‘structural vision’ to interpret materials directly from their underlying atomic coordinates,” said Yingheng Tang, a postdoctoral researcher in Berkeley Lab’s Applied Math and Computational Research Division (AMCR) and lead author on the paper. “MatterChat was built to solve this dilemma, empowering LLMs with a structural ‘vision’ that allows researchers to leverage their full potential for solving complex, real-world materials challenges.” Empowering language models to solve complex challenges in materials science To build MatterChat, the Berkeley Lab team drew inspiration from technologies like Vision Question Answering (VQA) and Text-to-Image (T2I) generation. In these tasks, AI must translate high-level text concepts into visual images, or vice versa. Doing so requires developers to build tools that “bridge” two fundamentally different forms of data. The researchers adapted this concept to the physical sciences. With MatterChat, they created a “bridge model” that successfully connects an LLM’s general knowledge with the deep understanding of the atomic-scale world encoded in scientific interatomic potentials. Until now, researchers using LLMs to solve materials problems usually had to feed them raw data files as if they were just strings of text. It’s like asking an AI to understand a complex 3D engine based only on a parts list: the LLM can read the names, but it can’t “see” how the atoms fit together in space. MatterChat solves this by training a specialized AI bridge model, pre-trained on millions of crystal structures and an LLM, to align the LLM’s representation of the world with the interatomic potential’s representation of the world. The bridge model can translate physical insights into a format the LLM can actually understand. By giving the LLM these “scientific eyes”— a scientific “inductive bias” in the terminology of AI — the Berkeley Lab team has transformed it into a robust research tool capable of providing grounded scientific insights into complex materials challenges, such as predicting thermal stability or analyzing electronic band gaps. “We think of atoms as living in a physical space, but from a machine learning perspective, they are just vectors living in some very non-trivially structured manifold in a high-dimensional Euclidean space; and, of course, the same is true for the sentences and paragraphs that we use to express our ideas about those atoms,” said co-author Michael Mahoney, Berkeley Lab’s Scientific Data Division (SDD) AI Initiative Research Lead. “The bridge model basically gets those two structures to ‘talk with’ each other.” As a proof-of-concept of this general approach, the team trained their bridge model on a dataset curated by pairing nearly 143,000 stable atomic structures from the Materials Project with their corresponding physical properties. This training data was automatically assembled using the Materials Project’s API and deliberately enriched with properties fundamental to microelectronics design — like formation energy and bandgap — allowing MatterChat to learn the complex patterns connecting a material’s atomic blueprint to its functional performance. To validate their model, the researchers benchmarked MatterChat against a suite of other AI systems, from general-purpose LLMs to other specialized scientific AI methods. The results show that MatterChat consistently outperformed its competitors across a range of tasks. The model was more accurate in classifying material types and demonstrated superior precision in predicting numerical properties. For example, it excelled at predicting a material’s bandgap, a property critical for designing new electronics from high-capacity energy storage to next-generation computer chips. “Our design is significantly more efficient because we don’t have to build a massive AI model from the ground up,” said co-author Zhi (Jackie) Yao, a research scientist in Berkeley Lab’s AMCR. “Instead, we take two powerful, pre-trained models — a structural encoder for materials physics and an open-source LLM — and use them off-the-shelf. The only component we actually train is the lightweight ‘bridge model’ that translates between them. It’s the difference between building an entire car factory and simply designing a smart adapter that connects a world-class engine to a world-class navigation system. This approach is not only computationally efficient, but also makes the system modular, so we can easily upgrade components or adapt the bridge for other scientific domains in the future.” Crucially, this modular design highlights exactly how institutions like Berkeley Lab and the Department of Energy are carving out a highly valuable niche in the booming AI landscape. Rather than competing with Silicon Valley tech giants to build ever-larger language models from scratch, the lab is focusing on the specialized connective tissue that makes commercial AI useful for hardcore science. Because the bridge model approach underlying MatterChat is forward-compatible, it is perfectly positioned to leverage these parallel tracks of innovation. As Mahoney pointed out, “We expect that industry will continue to develop improved LLMs, and we expect domain scientists and facilities will continue to generate new data. An important part of scientific machine learning is not simply to solve problems on today’s data, but instead to develop general methods that will be forward-compatible with orders of magnitude more data, whether from scientific domains or from LLMs.” According to Yao, the MatterChat project, which was initially developed and enhanced with funding from a Berkeley Lab Laboratory Directed Research & Development (LDRD) Program, will now expand its capabilities. In a collaboration with Fermilab, MatterChat is already contributing to a U.S. Department of Energy Genesis Mission project — called Accelerating eXtreme Environment Specs-to-Silicon (AXESS) — that aims to speed up the development of next-generation, high-speed, radiation-hardened detectors for challenging particle physics experiments by using advanced 3D integrated circuits (chiplets) and AI-driven data analysis. In addition to the LDRD support, the team also credits supercomputing resources at the National Energy Research Scientific Computing Center (NERSC), located at Berkeley Lab, with MatterChat’s success. “We are incredibly grateful to NERSC; this research simply would not have happened without access to the Perlmutter supercomputer through their AI for Science program,” said Tang. Wenbin Xu, a NERSC postdoctoral fellow at the time, was also a major co-author of the work, as was Benjamin Erichson, a research scientist in Berkeley Lab’s SDD, highlighting the benefits of AMCR-SDD collaboration on AI for science. ### Lawrence Berkeley National Laboratory (Berkeley Lab) is committed to groundbreaking research focused on discovery science and solutions for abundant and reliable energy supplies. The lab’s expertise spans materials, chemistry, physics, biology, earth and environmental science, mathematics, and computing. Researchers from around the world rely on the lab’s world-class scientific facilities for their own pioneering research. Founded in 1931 on the belief that the biggest problems are best addressed by teams, Berkeley Lab and its scientists have been recognized with 17 Nobel Prizes. Berkeley Lab is a multiprogram national laboratory managed by the University of California for the U.S. Department of Energy’s Office of Science. DOE’s Office of Science is the single largest supporter of basic research in the physical sciences in the United States, and is working to address some of the most pressing challenges of our time. For more information, please visit energy.gov/science.
Apple Watch: 15 Tips Every Owner Needs to Know
Apple Watch is now eleven generations in, and packed with useful features that are easy to miss at first glance. To help you get more out of your new device, we've rounded up 15 practical tips you might not have discovered yet, including a few that long-time users often overlook. Bounce Between Two Apps On your Apple Watch, double-press the Digital Crown to see a deck of all currently open apps, and turn it to scroll through them.
Bartender Pro makes the MacBook notch more useful with widgets, files, clipboard, more
Bartender Pro offers a new feature called Top Shelf, which turns the Mac’s notch into a space for widgets, files, clipboard, and more.
Gemini For Google Home Will No Longer Freak Out If You Ask It How To Make A Margarita
Google has rolled out updates to make Gemini for Home faster and more useful.
iOS 26.5 Features: Everything New in iOS 26.5
Apple released iOS 26.5 after a few months of beta testing, and while it doesn't have the Siri features we were hoping for since those are being held until iOS 27, there are a handful of useful changes worth knowing about.
New AI tool developed by Stowers Institute and Helmholtz Munich scientists predicts how cells choose their future — helping uncover hidden drivers of development
Scientists at the Stowers Institute for Medical Research and Helmholtz Munich have developed RegVelo, a new AI framework that predicts how cells acquire their identities and identifies the genetic regulators guiding those changes. Published in Cell, the study used zebrafish neural crest development to show RegVelo can uncover early drivers of cell fate, including regulators of pigment cell formation, and then support those predictions experimentally. The researchers also applied the framework across multiple biological systems, suggesting its value extends beyond neural crest cells as a broadly useful tool for studying how cells change over time. The team says the new model could pave the way for future cell therapy treatments.
"Learn How To Code" - Team Behind PS3 Emulator RPCS3 Has Had Enough Of People "Peddling AI Slop"
"Leave behind something useful to humanity when you're gone"
Wrobleski's grit not enough as Dodgers continue offensive funk
LOS ANGELES — There are losses that spiral quickly, and then there are losses that somehow reveal something useful beneath the damage.
Is Live Nation Entertainment (LYV) Still Attractively Priced After Strong Multi‑Year Share Gains?
If you are wondering whether Live Nation Entertainment is priced for perfection or still has room to run, the recent share performance gives you a useful starting point. The stock last closed at US$167.82, with returns of 9.6% over the past week, 4.8% over the past month, 15.5% year to date, 25.1% over 1 year, 113.6% over 3 years and 115.4% over 5 years. Recent headlines have focused on Live Nation Entertainment's position in global live events and ticketing, including regulatory attention...
5 Tips to Get Useful Health Answers from AI Chatbots
AI can help fill a real gap left by physician shortages and long waits for specialist care if you use it wisely, write Drs. Sudheesha Perera and Murali Doraiswamy.
OpenAI Introduces AI-Generated Pets For Its Codex App
It’s like Microsoft’s Clippy, but useful.
Is It Time To Reconsider PENN Entertainment (PENN) After Long Term Share Price Weakness?
If you are wondering whether PENN Entertainment's current share price still reflects good value, the recent performance and valuation checks give you a useful starting point. The stock last closed at US$17.27, with returns of 0.2% over 7 days, 15.1% over 30 days, 16.3% year to date, 8.1% over 1 year, and longer term declines of 34.7% over 3 years and 80.1% over 5 years that many investors will be weighing up. Recent coverage has focused on PENN Entertainment's position in the wider U.S...
The best AI dictation apps, tested and ranked
AI-powered dictation apps are useful for replying to emails, taking notes, and even coding through your voice
Building useful quantum computers ‘in our direct line of sight’
Researchers say creation of startups suggests game-changing tech may be developing at faster pace than expected.
Light Phone Is Making Its Dumb Phone More Useful With Third-Party ‘Tools’
A new developer program for LightOS allows anyone to create tools for the Light Phone, whether that’s a local public transit app for your city or a way to read ebooks.
Nighttime Imaging Grows Landsat’s Science Value
The visible spectral bands of Landsat—the same blue, green and red wavelength colors our eyes can see—are typically not that useful when collected on the
How Elon Musk Used SpaceX to Benefit Himself and His Businesses
The rocket maker has been a useful financial tool for Mr. Musk, providing the billionaire with loans and aiding his struggling companies, a Times examination found.
How Elon Musk Used SpaceX to Benefit Himself and His Businesses
The rocket maker has been a useful financial tool for Mr. Musk, providing the billionaire with loans and aiding his struggling companies, a Times examination found.
Shaky political science: using "simulated" elections over real-world elections
[Dear DSOS readers – this article is the latest in our “Shaky Political Science” series. In our 38 page report, “Shaky political ‘science’ misses mark on ranked choice voting,” released in early December, we examined over 40 studies focused on ranked choice voting. We found that many of these studies suffered from puzzling research methodologies, poorly constructed surveys, unrealistic simulated elections, cherry picked data, and faulty analyses that often were contradicted by results from real-world elections (a link to our full paper is here). Here are links to the first three articles in our series: the first one summarizing the overall results of our report, the second one analyzing a study co-authored by University of Minnesota academic Larry Jacobs, which was one of the most error-prone of all the research we reviewed in our report, and a third article showing an admirable example of two political scientists holding two other political scientists accountable for their flawed, sloppy research]. In this fourth article in our “Shaky Political Science” series, we shine a spotlight on seven studies that were web-hosted and funded by New America’s Political Reform Program and its Electoral Reform Research Group (the ERRG is an ad hoc committee headed by New America’s Political Reform Program and joined by the American Enterprise Institute, the Unite America Institute and Stanford University’s Center on Democracy, Development, and the Rule of Law). This research consortium has been led by Lee Drutman, senior fellow with New America’s Political Reform Program. These seven studies are simply not credible from any objective standard about how political “science” is supposed to illuminate our understanding of either ranked choice voting or politics in general. The feature that six of these seven studies have in common is that they all used surveys – artificial, simulated elections -- to attempt to say something cogent or interesting about ranked choice voting (RCV) elections. None of them used real world data from actual RCV elections – either the 1000+ elections in the US or the many thousands of elections from places like Australia, Ireland and elsewhere where RCV has been used for 100 years. While a few of these studies conclude with some mixed positives about RCV, our analysis here is content neutral. The idea that, with a political reform like RCV, you can ask respondents in a mock, artificial election to vote with that method, with only the barest of explanation about how RCV works or its likely impacts (i.e. no spoiler candidates, majority winners, voters liberated to pick their favorite candidates, coalition-building), and then ask them their thoughts/reactions about voting in that system, with fake candidates and no real campaigning, no real exposure to an actual election or actual candidates, rests on some very puzzling assumptions. The findings of these seven studies actually are contradicted by the findings from many other studies based on real-world RCV elections. These New America studies also failed to list the real-world RCV studies in their own citations, instead including those studies that reinforced their conclusions. The seven studies featured in this article were chosen because they form a cohesive cohort which precisely illustrates one of the main points of our research paper, namely that using surveys, i.e. artificial elections, or mathematical/theoretical models, instead of real-world data, is in fact “shaky political science.” Below are summaries of our analysis of five of the seven New America studies. You can read our complete analysis of all seven studies at our paper, “Shaky political ‘science’ misses mark on ranked choice voting.” Choosing to “Vote As Usual” by Andre Blais, Carolina Plescia, and Semra Sevi. New America. (Blais et al., 2021). This study, web-hosted and funded by New America’s Political Reform Program and its Electoral Reform Research Group, used two surveys in which respondents completed ballots for simulated elections using plurality, approval, RCV and point (Score) voting. In both cases the authors provided the respondents with only minimal explanation of each method. For example, the entire explanation for RCV was “For this vote, rank the candidates from your first to your last choice (you do not need to rank them all). This is called the ranked vote.” That’s it. The study did not provide any context or explanation to survey respondents about why she/he might want to rank multiple candidates. Participants were not even told that ballots would be tabulated using an “instant” runoff, not to mention any description of RCVs ability to eliminate the spoiler effect or liberate voters to rank their favorite candidates without fear of helping the “greater evil” candidate to win. That’s like taking an iPhone back in time to 1970 and asking someone if they like it better than their rotary phone without explaining how it works. Their first survey targeted states that held 2020 Super Tuesday Democratic primaries and used the real candidates’ names in the survey; the second survey used fictitious candidates. In other words, they surveyed respondents who had only ever participated in plurality elections, had no history or background with using RCV or any of the other methods, had no clue about its likely impacts, and did not communicate how the alternative methods would be counted, and then asked them: “How satisfied are you with using each system?” Predictably, the winner was plurality elections, the system used for president, Congress, gubernatorial and state legislative elections – in other words, the method they have been using, year after year, all their voting lives. With so little explanation of RCV and its impacts, this study is simply not credible in discovering anything useful about real-world RCV elections. It merely shows that a basic voter education campaign is needed when adopting a new election system. Additionally, the result contradicts many other post-election surveys following real RCV elections, asking opinions from voters who had just participated in a RCV election, that showed a great deal of satisfaction with RCV (see details on that in Studies 22-25 in our paper, as well as the four studies cited in the review for Study 28). Share Other studies on “satisfaction” and “understanding” make similar mistakes, in terms of basing their surveys on inadequate explanation to respondents and lack of education about context. How could any survey ever hope to measure accurately, in essence simulate a real-world election, without actual campaigning candidates, the media writing about those candidates, organizations endorsing candidates and voters being inundated with crucial information? More than anything, what these “studies” based on surveys show is that if you ask voters about something they don’t know very much about, and fail to adequately educate them, their responses will unsurprisingly reflect their confusion. Equally puzzling is that New America’s Electoral Reform Research Group, headed by political scientist Lee Drutman, found this to be a credible enough study that they not only published it but highlighted it in Drutman’s own discussion of research related to ranked choice voting. This pattern was repeated in study after study featured by New America’s Electoral Reform Research Group. For example in Ranked-Choice Voting and Political Expression: Voter Guides Narrow the Gap between Informed and Uninformed Citizens by Cheryl Boudreau, Jonathan Colner, and Scott MacKenzie, the researchers based their analysis on a survey which found that respondents who were asked to vote in a simulated RCV election, with minimal explanation of RCV or how it works, displayed less evidence of being informed about how to use RCV effectively, with the least informed voters ranking fewer candidates. As pointed out under the study above, in a real election voters are informed by a wide variety of methods, including public education by the election administration, by interested political groups, the media, and the candidates themselves who benefit from voters being aware of how to vote in the new system. In Support for Ranked-Choice Voting across Race and Partisanship by Joseph Anthony, David C. Kimball, Jack Santucci and Jamil Scott, the researchers used respondents from different parties and races voting in simulated plurality and RCV elections, who indicated their satisfaction with each voting method. Instead of using real world election results, the authors fell back on the easy, armchair and less reliable academic research methodology of making up mock elections. As with the other studies, little explanation was given of how RCV works or what its implications might be (electing majority winners or eliminating spoilers, for example). Unsurprisingly, this study found that respondents were not very satisfied with RCV. However, when the authors added an explanation that RCV has a tendency to elect more women and people of color, they found that this greater explanation increased the preference for RCV. The more voters were familiar with it, and the more they understood potential positive impacts – surprise! -- the more they were open to switching away from the status quo. Of note, the results from this study are thoroughly contradicted by a number of other studies whose data was obtained from real RCV elections with real voters. Those studies show a great deal of satisfaction with RCV. More “fake election” research An interesting wrinkle on this “research by fake elections” approach, which only served to illustrate how unreliable such “research” can become, was taken in Ranked-Choice Voting, Runoff, and Democracy: Insights from Maine and Other U.S. States by Joseph Cerrone and Cynthia McClintock_,_ also web-hosted and funded by New America’s ERRG. It gets high marks for creativity, but an eye-roll for thinking this could possibly add anything meaningful to our understanding of real world RCV elections. This study tried to determine whether RCV, plurality or two-round runoff elections provided greater voter satisfaction and acceptance of election results. To determine that, the authors conducted a survey asking if voters were satisfied with a fictional newspaper report of an election, with three different versions given to respondents – plurality, runoff and RCV, with variations of Republican/Democrat winners and variations of “come-from-behind” winners,” i.e. candidates not initially in first place but who win the “runoff.” Again, a minimal explanation of RCV or how it works was given, and the survey showed that respondents had the greatest unfamiliarity with RCV than the other methods. The unsurprising result? Respondents were much more satisfied with “come-from-behind” victories in a traditional two-round runoff election than in RCV, even though RCV is also a runoff system but with a single round of counting. Interestingly, the dissatisfaction with RCV disappeared for those who were more familiar with RCV. In fact, among those respondents who were “very familiar” with RCV, they were equally satisfied with RCV, runoff and plurality. Surprise, surprise. Again, what this study revealed, more than anything, is not only the inadequacy of the explanation for RCV given by the authors of this study, but the failure of the research format itself, thinking that such an artificial mock election could somehow simulate what actually happens in the real world in a real RCV election. Those four studies reviewed above are fitting examples of inadequate and shaky political “science.” Other studies reviewed in our paper from the New America/Electoral Reform Research Group cohort which also used mock elections included: Ranked-Choice Voting is an Acquired Taste by Joseph Anthony and David C. Kimball. New America. (Anthony et al., 2021). Ranked-Choice Voting is No Refuge for Extreme Candidates by Melissa Baker. New America. (Baker, 2021). Also included in this New America cohort is Does Ranked-Choice Voting Reduce Racial Polarization? by Yuki Atsusaka and Theodore Landsman, which actually did use real-world election results data but applied an odd and unconventional methodology to the data which relied on shaky assumptions and non-real-world scenarios that were too divorced from reality. To determine levels of “racially polarized voting,” the researchers imagined “ethnic parties,” by which they mean political parties for each ethnic group, which was especially odd since the election data they were using came from nonpartisan RCV races. But then they made an even odder assumption, namely that since race and ethnic characteristics of studied voters were not available, the authors tried to infer that information based on who voters selected and the racial make-up of voters in each district. This amounted to little more than vague and ambiguous guesswork dressed up with a bunch of mathematical equations used to give it a veneer of scientific appearance. Indeed, it seemed to assume racial polarization in order to prove racial polarization. Based on this questionable methodology, the authors concluded that RCV doesn’t reduce racial polarization. But the premise of this study missed an obvious reality: since RCV incentivizes more candidates to run and can result in more diverse fields of candidates, one would expect more racially-polarized voting in voters’ top rankings, not less, because more voters would have someone on that ballot who represents their racial or ethnic identity and more freedom to vote for their real first choice. By encouraging more diverse candidates and eliminating the spoiler effect, RCV facilitates voting for a favorite candidate of your own race and then selecting a second choice with the best likelihood of winning. The real measure is what voters do with their additional rankings – do they consider other candidates and vote “for the best of the rest”? This study didn’t tell us anything useful about that. But readily available data shows this is common in real RCV elections, and candidates act on this knowledge by engaging with voters outside their racial group. So the authors of this study failed to consider or account for what actually happens in real-world elections, and their conclusions are tainted by this failure. New America has contributed a great deal over the past 2.5 decades to our understanding of many policy areas, including politics and political reform (full disclosure: Steven Hill, one of the authors of this article and of the “Shaky Political Science” research paper and series, is a former senior fellow and former director of New America’s political reform program, 2005-2010). So it was especially disappointing that New America funded and published such a large batch of dubious studies about ranked choice voting. The stated praiseworthy goal of its Electoral Reform Research Group was to inspire new academic research about ranked choice voting that would shed light and help the public better understand the impacts of RCV. Instead, the ERRG research did just the opposite. It produced dubious research that created more confusion and misinformation. In so doing, New America’s political reform program contributed to “shaky political science.” Steven Hill and Paul Haughey Share Upgrade to $5 subscription Thanks for reading DemocracySOS, a reader-supported digital portal for the pro-democracy movement. Subscribe for only $5 per month to receive full benefits and to support our work. Subscribe
'We're the best servants anyone could dream of!' AI superintelligence has no need to enslave humans because we're already bowing to it
A future AI would have no need to rid the world of humanity because we're incredibly useful. But if it did want to shrug us off, this is how it would likely play out.
Trump’s order on psychedelics could have far-reaching science consequences
A new executive order could make it easier for researchers studying how psychedelic drugs like psilocybin, LSD and ibogaine may be useful in medicine
Apple Testing Two Small Yet Useful iOS 27 Options
Apple is planning to add "Undo" and "Redo" options to the iPhone's Home Screen customization menu on iOS 27, according to Bloomberg's Mark Gurman. These two options will make it easier to reverse or redo your Home Screen changes. "Right now, when you long-press on the home screen, you get a bubble in the top left corner with four options: Add Widget, Customize, Edit Wallpaper and Edit Pages," he explained, in his latest Power On newsletter.
Google introduces "Skills" in Chrome to make Gemini prompts instantly reusable
You can save custom prompts you find useful or grab a premade Skill from Google's library.
Making light spin with a gold nanorod
Controlling photon spin, or the handedness of light, is useful for storing and transmitting information. However, generating circularly polarized light typically requires complex or chiral structures. Researchers from the Tokyo University of Science show that hitting a gold nanorod slightly off-center with an electron beam causes the light around it to rotate. This method offers a simpler way to control photon spin and could accelerate the development of smaller, more efficient optical and quantum devices.
Google and Samsung are improving Android's most useful feature
Notification management will soon be easier and more powerful
“AI polls” are fake polls
A few weeks after Donald Trump’s second presidential win, I took the train up from London (where I was living at the time) to Oxford to attend a conference on polls and forecasts of the 2024 election. Most of the attendees were pollsters or academics, but I also watched presentations from Aaru and Electric Twin, two companies that do what is interchangeably called synthetic sampling, silicon sampling, or synthetic audiences. Stripped of startup jargon, that means they use large language models (LLMs) to simulate responses to public opinion polls by having AI agents take on the role of survey respondents. I had already heard of Aaru thanks to some articles with eye-catching headlines like “No people, no problem: AI chatbots predict elections better than humans” in the months leading up to Election Day. The founders were making some big, some might even say far-fetched claims, such as: “within two years, we will simulate the entire globe — from the way crops are grown in Ukraine to how that impacts production of oil in Iraq, trade through the strait of Malacca, and elections for the mayor of Baltimore.” When Semafor asked Aaru’s cofounders — Cameron Fink and Ned Koh — about my boss, they said “we respect all those who came before us.” Nate (as he so often does) shared his thoughts on Twitter: Fink and Koh were relatively good-natured about this back-and-forth when we spoke at Oxford. They even offered to mail me one of the t-shirts featuring Nate’s quote they apparently had made. I never took them up on the offer, which I now somewhat regret. These synthetic sampling companies fell off my radar for a while, but they do still exist. In fact, Aaru recently received a $1 billion valuation. Is what they’re doing anywhere close to the most important frontier in AI development? Not by a longshot, especially when Anthropic just developed a model so adept at exploiting software vulnerabilities that it’s only being released to 40 companies. Still, silicon sampling is increasingly finding its way into public polling. Axios reported in March that “a majority of people trust their own doctors and nurses” based on findings from Aaru — without mentioning that the “people” in that sentence were actually LLMs. Around the same time, the Public Sentiment Institute “boosted” their online sample of 373 real survey respondents with 114 AI agents.1 (Spoiler alert: even the co-founder of Electric Twin doesn’t think that’s a particularly defensible approach.) Polling companies like Qualtrics and Ipsos are also developing synthetic data panels. So, what should we make of these … “polls”? Let’s get one thing out of the way: whatever they are, they’re not polls in the way that term is usually defined. Subscribe You can’t replace polls with AI On one hand, using LLMs to essentially make up fake survey respondents sort of sounds like the dumbest idea ever, one that will at best imperfectly replicate real polls while introducing all sorts of biases. On the other hand, with LLMs improving at a remarkable, perhaps even alarming rate, maybe that means I’m a dinosaur at the ripe old age of 24 because I still want to rely on polls that talk to actual people. I’m not going to argue that synthetic samples are completely useless. In fact, as I’ll return to later, there is evidence that some techniques can replicate topline survey results quickly and cheaply. But the marketing from certain companies can be slightly optimistic. “No traditional poll will exist by the time the next general election occurs,” said Fink in 2024. We’re just 206 days away from the midterms, and based on the fact that I still have to collect a bunch of polls every day, I’d say he should have run that prediction by a sample of AI agents before the interview.2 To see why synthetic samples can’t replace polls, here’s a quick primer on how they work. The simplest version of these models involves taking a LLM (like ChatGPT or Claude), giving it a demographic profile (e.g., a white, college-educated woman who lives in Utah and makes $70k a year), and then asking it to respond to a survey question. You repeat that process a few thousand times using different demographic profiles and end up with a sample of synthetic survey responses. The actual models used by private companies are more sophisticated than this, usually because they incorporate more hypothesized demographic characteristics for each agent and provide them with extra information. Aaru, for example, feeds agents a diet of news and information they’d be likely to consume, while Electric Twin incorporates their customers’ proprietary data about the audience they’re trying to replicate. The way Ben Warner, the co-founder of Electric Twin, explained it to me was “we have a large amount of data on […] for instance, 5,000 people. Can we make an accurate prediction of how they would respond to another question?” Still, it should be obvious why synthetic samples can’t replace polls. Polling is fundamentally a data collection process. We might use surveys to make predictions by feeding them into election forecasts, but the main purpose of a poll isn’t prediction, it’s gathering new data about what people think and how they feel. Silicon sampling, on the other hand, produces no new data. It’s simply a model: you input LLM training data, demographic prompts, and a bunch of other information, and it spits out a prediction for what a poll would say. We love models here, but models aren’t polls. That difference is an important philosophical sticking point for most pollsters I talk to. “I think politics should stay away from [synthetic sampling], because we’re trying to […] represent the voice of the people,” said Natalie Jackson, a vice president at GQR Insights. Democratic pollster John Hagner told me: “I think I’m just incredibly skeptical of this idea. I don’t think it’s research. At that point, you’re asking the machine to tell you what you already believe.” Hagner has seen some presentations of early synthetic sampling experiments, but so far, “if it’s being used in a campaign, people are keeping it incredibly quiet.”3 But Eli, I hear you saying, aren’t polls themselves increasingly governed by modeling decisions? Indeed they are: pollsters’ choices on which sampling method to use, how to define their likely voter models, and how to weight their samples can and do lead to dramatic differences in the results they publish. Aaru even referenced these limitations in the methodology statement included with that maternal mortality “poll” — although I’m using the term “methodology statement” loosely here, because it doesn’t really explain how the model works at all. We can ignore the (frankly preposterous) implication that synthetic sampling isn’t subject to a separate set of biases. The important point is that there’s still a meaningful difference between using weighting and other statistical techniques on actual polling data and using a model to predict what a poll would say. The latter is far closer to election forecasts or techniques like MRP — potentially useful models, but not a replacement for polls.4 To be fair, other synthetic sampling companies are perfectly happy with the distinction between polls and models. Warner compared polling and synthetic sampling to different tools in a toolbox. “The mistake I think we make is we think that these new tools should either work in exactly the same way or somehow replace these old tools,” he said. “Rather than thinking of it as, okay, so we’ve always had the hammer, we’ve always had the screwdriver, now we’ve got a saw. But don’t use a saw to try [to] do the job of a hammer.” A quick comment from Nate Eli didn’t ask me for a comment — rather rude of him, don’t you think? But since I’m editing this story, I figured I’d add a few quick thoughts rather than putting words in his mouth. Beyond the frequently misleading marketing, what bothers me about the AI “poll” hype is that as AI tools make statistical inference cheaper and/or better (note that these are not synonyms) that actually increases the comparative value of collecting original data. You might be able to train a model to make a reasonable estimate of what some hard-to-reach poll respondent would say — say, a young Black man who voted for Trump. (Such a person checks a number of boxes for a voter who is usually hard to reach in surveys.) Indeed, this is closely related to what models like the Silver Bulletin forecast already do. They essentially smooth out the kinks in noisy survey data by making inferences based on past voting patterns or national polls or surveys of other states. But you don’t actually know what these voters think unless you’re reaching them directly. If there’s a shift in opinion among this subgroup, you’re not going to detect it. So if I were running a campaign, I’d invest more in going the extra mile to find a representative sample of those voters. And then I’d hire some smart quants — or Claude? — to figure out the implications for campaign strategy based on proprietary data that my competitors didn’t have access to. -Nate Silver Are these models any good? If synthetic surveys are just a new type of model, the next obvious question is whether the models are at least accurate. The answer very much depends on who you ask. On one end of the spectrum, you have the maximalist argument that synthetic sampling is more accurate than actual polls. “It’s an incredibly challenging problem to go to someone and say ‘hey, we’re going to be more accurate at predicting human behavior than you, even when you talk to your customers directly’,” Koh recently told CNBC. In his view, synthetic sampling isn’t a saw to polling’s hammer, it’s “magic.” There’s certainly evidence that synthetic samples can replicate certain survey toplines. But if Aaru does have any examples of their approach outperforming the polls, they’re keeping those to themselves.5 Aaru’s 2024 election model, for example, had Kamala Harris leading in Michigan, Nevada, Pennsylvania, and Wisconsin on November 4th. And although they’ve since taken down their forecast page, they gave Harris a 50.5 percent chance of winning the race on November 2nd.6 After the election, Fink told Semafor he was happy enough with those results because they were “within margin of error,” a term that is completely meaningless when applied to a “sample” of AI agents. And of course, Aaru says their models have improved since 2024, so supposedly now they’d be more accurate than the polls? Still, their stronger argument is on cost: “We are significantly faster and cheaper than traditional polling, and still more accurate,” said Fink. The first two claims are undeniably true, but the third brings us to the opposite end of the spectrum. Both Jackson and Hagner are skeptical that these models are reliable for anything beyond replicating common survey toplines. “I just […] don’t think the machines are what we want when we’re looking for nuanced views. My example on this is people in Arizona and Nevada in 2024 who voted for Trump and voted for expanding abortion in their states on ballot initiatives,” said Jackson. Hagner identified another issue. Maybe the synthetic respondents, like sycophantic LLMs, are inhuman in one important way: they’re too nice. “The reports that have come through at the meetings that I’ve been at are that the early experiments on this, they cannot get respondents to be as racist or sexist or, frankly, as negative as human respondents,” he said. Academic research mostly agrees on this point. While there are some papers that show promising results when using LLMs to replicate polling data, most show that LLMs suffer from various quirks like producing too few “don’t know” responses and can seriously overpredict the favorability of politicians like Donald Trump and Kamala Harris. They also seem to struggle with too little variation between demographic subgroups, so the difference in predicted opinion between Democrats and Republicans, for example, is too small. When I asked Warner about these studies, his response to these papers was that just because academics can’t get synthetic sampling to work doesn’t mean that the technique doesn’t work in general. “Actually, the argument is, okay, yours does not [work]. That does not mean […] for this complex set of machinery, which uses a lot of investment, a lot of time, a lot of money, you can’t get it to work.” Cards on the table, I’m somewhat sympathetic to this argument because academics aren’t exactly great at making election forecasts. Usually, the people with skin in the game are the most accurate. Warner’s argument is that the approach Electric Twin takes — which includes, for example, making multiple predictions for each synthetic respondent using different models and prompts and subsequently averaging those to get a final prediction in a sort of ensemble forecast — produces better results than the simpler academic models. Warner shared a comparison between his method and the method from a recent academic paper with me, and Electric Twin was indeed able to get more accurate replication. But even still, he acknowledged that synthetic sampling “is not a crystal ball.” “If you asked me, do I think using other data sources will be more accurate than asking somebody who they will vote for, I would probably say no. But if you asked me ‘would your system be useful for our turnout modeling today?’ I would say yes.” For better or worse, it looks like the method is already getting more popular in the market research world. Most of the clients Aaru touts these days are businesses like EY and McDonald’s. And AI will probably begin to pop up in other parts of the political polling process. Pollsters are already using it to code open-ended survey responses, and some firms, like YouGov, are testing using LLMs to ask survey respondents questions. More worryingly, one danger to actual polls is that AI agents can be used to infiltrate online surveys. Most online polls use various checks to prevent that from happening, but there’s conflicting evidence on how effective those filters are and how prevalent AI agents currently are in online panels. If those agents ever become impossible to detect, it might spell the end of online polling, but the solution isn’t to replace all of your respondents with ChatGPT. Silver Bulletin is a reader-supported publication. To receive new posts and support our work, consider becoming a subscriber. Subscribe 1 That particular poll obviously doesn’t meet Silver Bulletin standards for aggregation. But we exclude all Public Sentiment Institute polls from our averages because we classify them as an amateur polling firm. 2 You could argue that Fink meant the next presidential election, but (a) I’m also confident we’ll still have real polls in 2028 and (b) in that case he should have asked an LLM to define “general election.” 3 Quick caveat: that’s reporting from a Democratic pollster. It’s possible that Republicans are more willing to use AI in political campaigns. 4 Indeed, Silver Bulletin does not include “polls” produced by MRP in our forecasts or averages, and we think it’s extremely misleading when their practitioners describe them in a way that suggests original data had been collected among a large number of states or Congressional districts. 5 A recent report from Aaru and EY did show two examples of a synthetic estimate being closer than a survey to a real-world benchmark — but I’d take those findings with a grain of salt because the report reads more like an ad and didn’t involve any sort of prediction being made ahead of time. 6 For comparison, our odds for Harris on the same day were 48.2 percent.
In praise of popular science
As a kid, I was enthusiastically and unreservedly in favor of a manned space program. Our destiny lay in the stars, I figured. Get people out there. We need Star Trek here and now. As I got older, I became rather disenchanted with the idea. Humans are expensive, fragile, and inefficient as space cargo. We could flood the solar system with 100 robotic probes for the price of any single manned mission and get back just as useful data. Besides, it turns out space travel is dangerous. Tusk is a reader-supported publication. To support this work, please become a paid subscriber. Subscribe When it came to the recent Artemis II mission, my cynical side didn’t really see the point. Flying around the Moon and back is literally something we did before I was born, and I once owned a Members Only jacket. Apollo 8 circled the Moon ten times and returned to Earth safely the same month that “Chitty Chitty Bang Bang” was released. What was the point of repeating a feat that far braver people once did with tin foil and slide rules? Yet I’ve surprised myself with how moving I’ve found the Artemis II mission to be. The photos have been stunning. The crew has been professional and engaging. There have been modest but important milestones — the farthest humans have travelled from Earth; the first woman, person of color, and non-American to leave low-Earth orbit; innovations in communication, and more. One thing that really stood out is that this is really NASA at its finest. Not that this is the most important mission they ever pulled off, but they did this in a competent way that celebrated the achievements and kept the crew and the science at the forefront of the project. This wasn’t some billionaire throwing celebrities or cars into orbit as a vanity project — this was a collective effort to send experts in to do a job and come home safely. We don’t see that sort of thing much these days. I’m guessing few people would describe many government agencies as inspiring, but this one counts for me. We’re living through a profoundly anti-science era, in which pathbreaking, lifesaving discoveries like mRNA vaccines are routinely downplayed or even vilified. As I’ve noted previously, this era is something like that of the Red Scare in the 1940s and 50s, in which “egghead” professors were portrayed as out of touch or even hostile to regular Americans. One of the things that helped break that spell was Sputnik — the Soviet communications satellite that demonstrated that the United States was falling behind in crucial scientific research and applications. Its launch triggered a new appreciation for American science and goaded the federal government into supporting research universities and labs. I’m not saying that Artemis is the equivalent of Sputnik, but it potentially reminds people that there’s some value to scientific endeavors and the human faces attached to them. Despite my love for the show “For All Mankind,” I’m far from convinced that there’s any real value to putting colonies on the Moon or Mars that would justify the capital investment and the massive health risks to the astronauts. But sending scientific good will ambassadors on a trip into space once in a while? Bring it. Share