Tentative thesis: China is unlikely to accept a subordinate position in AI capabilities to the US, just as the US is unlikely to accept a subordinate position in AI capabilities to China or anyone else.
This thesis suggests that there are two likely paths for the future international AI (non)regulation:
1) a continuing race between the US and China, perhaps joined by some late entrants, possibly with some rules established (e.g. mutual ban on autonomous weapons or something like that), or
2) a “pause or stop deal” that will establish some sort of ceiling on capabilities, which will be equal for China and the US.
What I think is far less likely is a deal in which China would accept lower AI capabilities than the US.
I simply don’t see a good reason why Chinese leaders would accept subordinate position. They know that China has the ability to catch-up to the US in various technological domains, as evidenced by the fact that in, like, 1990, China was behind in more or less everything, and now they are pushing the technological frontier in many areas.
Some estimates floating around the web suggest that, if the US would stop AI development now, China could reach current US capabilities within months. That is maybe overly optimistic/pessimistic depending on where you stand, but I very much doubt that catch-up would take more than a decade. And Chinese government, being patriotic about the abilities of the Chinese nation, probably will not have an absurdly pessimistic estimate.
Moreover, the Chinese regime is in many ways oriented around this idea of catching up (this is a big difference between China and EU). It is a central plank of the official historical narrative of the People’s Republic of China (see for example preamble to their constitution: https://english.www.gov.cn/archive/lawsregulations/201911/20/content_WS5ed8856ec6d0b3f0e9499913.html) that the century between First Opium War and the Communist victory in the Chinese civil war in 1949 was the century of humiliation, during which various imperialist powers exploited China due to the fact that it was technologically backward country unable to stand up to them. I happen to think this narrative is largely correct, but even if it isn’t, what matters is that Chinese government believes it and is determined to never again allow China to be put in the position where a technologically superior power could walk over it.
This has some implications:
A) Since China is currently behind, any sort of “pause or stop deal” between the US and China would likely have to involve either the US stopping further development and allowing China to catch up, or the US somehow crippling its own capabilities to a current Chinese level. Of course, for either option, the US would have to place a lot of trust in the Chinese government.
B) As long as China is behind, export controls on hardware and other attempts to limit Chinese capabilities are at best of doubtful utility as a way to get to a "pause or stop" agreement. In fact, imho there are making a deal less likely. They antagonize Chinese leaders, make them even more paranoid and less inclined to a constructive agreement. But also, to the extent they are successful in limiting Chinese capabilities, they make it less likely to that the US government would agree to an equal deal, since that would mean sacrificing larger lead, which is harder to politically accept for the US. Perhaps this is a reason why nuclear arms treaties in the Cold war were agreed only after the USSR more or less caught up to US nuclear arsenal?
Now, attempts to limit Chinese capabilities can still be seen as useful if you think it is important for the US to “win” the competition, for example because there are reasons to believe US AI will be safer. Or because you think the world would be a better place if the US “wins”, even though both countries would develop similarly dangerous systems. Or of course if you think that the US victory would be preferable even if American AI would be more dangerous than Chinese AI.
Btw. for a well written snapshot of Chinese opinion/vibes, I recommend this: https://www.chinatalk.media/p/how-anthropic-became-chinas-goliath
I would like to express my support for Bruce Friedrich’s post. And he has strong effective altruist support: Peter Singer and Jeff Sebo have raised the same concern about how effective animal advocacy evaluates different approaches to change (Sebo & Singer, 2018).
1. Difficulty measuring the benefits of activism should not lead us to discount it.
Sebo and Singer explicitly identify the risk of measurability bias. Their concern is that the methods we use to evaluate activism may favor interventions whose benefits are easier to count. An approach may have important benefits even when those benefits are difficult to measure. As they explain:
"It is rather an argument that, if one aspires to reduce animal suffering as much as possible and to use evidence and reason in the course of doing so, then one should take care not to focus disproportionately on the direct, individual impacts of particular interventions since doing so will lead to an incomplete, and probably incorrect, analysis of what animal activists should be doing individually and collectively."
This matters for dietary advocacy because changing social norms may help people adopt dietary changes even if the immediate effects of individual interventions are small.
Expressed another way, there is a fair amount of peer-reviewed research that has found that social stigma is a reason people do not change their diets and, if they do change their diets, change them back (Markowski & Roxburgh, 2019; Hodson & Earle, 2018; MacInnis & Hodson, 2017; Minson & Monin, 2012; Cole & Morgan, 2011).
Many people simply want to be “normal,” and whatever normal is, they will do. This suggests that resistance to dietary change may reflect more than a personal commitment to eating meat. People may know that animals are suffering, and care, and still not go vegetarian or vegan because they are afraid of social isolation. Making vegetarianism and veganism seem normal and decreasing stigma could help more people make and keep the change.
We do not need to wait for a global tipping point to begin this work. Norms can change within particular communities. For example, we see many African Americans embracing plant-based eating as it becomes a norm within parts of the African American community, often framed around health and food justice rather than animals (Harper, 2010; Botchway et al., 2022). I experienced this firsthand in Atlanta, Georgia. What stood out to me was that many of the restaurants that were “plant-based” did not mention animals at all; however, they did mention how African Americans had been given the worst food and how plant-based eating was now a way to reclaim their health. I knew several people who took their dates to completely vegan restaurants, even though they were not vegan themselves, because they viewed these as high-status restaurants that would impress their dates.
An analogy might be smoking: One could imagine effective altruists in the 1950s trying a host of strategies to decrease smoking, including, say, flyers and protests, finding few immediate results, and concluding that these approaches were not worth pursuing. They might then focus only on making cigarettes less harmful, perhaps by decreasing the amount of tar. But substantial change in the rate of American smokers was possible. It required norm-setting behavior.
Or take civil rights, which I raise to compare how movements create change, not to equate the experiences involved. One could imagine effective altruists studying individual efforts to decrease anti-Blackness in the South, finding limited utility in, say, distributing flyers, and concluding that all that could be done to help African Americans was to convince white Americans to treat African Americans better. However, more substantial change was possible. Here, too, it required norm-setting behavior.
Sebo and Singer make this point explicitly:
"For example, historical studies of the civil rights movement, the feminist movement, the LGBTQ+ movement, and more allow one to see how conciliatory and confrontational approaches to activism can interact in ways that make the whole different from and greater than the sum of its parts."
Their argument is that we need to assess how approaches work together, as well as what each achieves individually. This is difficult research, but, as they put it, “It might be difficult to do this work well, but it is not impossible to do so.”
We can see an attempt at norm-setting, or shocking the conscience, in the open rescue movement. Perhaps it will work, perhaps it won’t, but we should not assume it is impossible. Its potential effects include changing how people understand animals and animal exploitation, alongside the direct effects of individual rescues.
2. We do have evidence that we can create long-lasting dietary change.
It is incorrect to say that there is no evidence we can produce long-lasting dietary change. We do have evidence. The problem is that we just don’t know how to scale it. The technique with the strongest support so far for producing dietary change is what we refer to as “moral shocks,” in essence, watching Earthlings or Dominion.
For example, Kramer and Landry (2026) randomly assigned 1,149 University of Toronto undergraduates to watch either a 16-minute clip from Dominion showing industrial pork production or a control video about mushrooms. Instead of asking only about intentions, they had participants choose real meals, and about half of those choices were actually served. What is so strong about this study is that the changes lasted. Right after the video, participants chose 27 percent fewer meals with meat. A week later, they were still choosing 13.5 percent fewer meals with meat, 30 percent less pork, and 122 percent more plant-based meals. The effect did shrink over the week, and the participants were college students, so we need more research on how long these changes last and who they work for. But the changes did not disappear. I learned about this study from Seth Ariel Green, who described it as “a great study.” There are other studies that have found similar results.
So the question is not whether any intervention can change diets. We have evidence that some can. The problem is the difficulty of scaling up, of getting large numbers of people to watch Earthlings. However, it might be possible to significantly increase the number of people who have an Earthlings-like experience. For example, a number of street outreach programs are thinking through ways to have people agree to watch segments of Earthlings.
Relatedly, lectures in college classrooms seem to work. Tracking more than 100,000 meal selections, Jalil, Tasoff, and Bustamante (2023) found that students randomly assigned to a single 50-minute talk on the health and climate effects of eating meat reduced their meat consumption by 5.6 percentage points, with no signs of reversal over three years. The authors calculate a high return on investment even under conservative assumptions, about US$14 per metric ton of CO2-equivalent.
This matches my own experience. Virtually every year, I get to teach a class in which students read every word of Animal Liberation and watch every second of Earthlings. I know they do because I test them on it. Perhaps more importantly, as a professor, I can set a norm in which veganism is clearly not stigmatized: the professor is doing it. And this happens at a key stage in students’ lives, when they are often first deciding who they will be as adults, separate from their families. Unsurprisingly, the changes I see are substantial, with 10–20 percent of students becoming vegetarian or vegan, and this works in a completely different way from, say, a flyer. I know many other vegan professors. We could support them in teaching more classes about animals.
In any case, this evidence suggests that some interventions can produce lasting dietary change. So, no, not flyers per se, but yes, it is a possible area where more research could be helpful.
3. Seth’s meta-analysis is narrower than it may appear.
The meta-analysis identifies limitations in the existing research and excludes many studies that do not meet its inclusion criteria. Only after applying those criteria does it find a small average effect across the remaining studies. That finding does not establish that dietary change is impossible or that other interventions cannot produce larger effects. The authors explicitly call for rigorous evaluation of promising interventions, which supports further research, including more and better-designed studies (Green et al., 2026).
I am aware of a few studies that have found that individual approaches may not be effective, for example, distributing pamphlets or providing information about health and the environment (Haile et al., 2021; Bianchi et al., 2018). Yet a classroom lecture on health and environmental topics did produce lasting change (Jalil et al., 2023). Finding that a particular approach did not produce dietary change does not prove that dietary change itself cannot work.
It is worth noting that one of Seth’s co-authors, Maya Mathur, had earlier published a well-done meta-analysis that found that appeals to animal welfare reduced meat consumption, at least in the short term (Mathur et al., 2021). When I asked her how to think about the two, she told me she now considers Seth’s meta-analysis the better one: the field’s methods improved much faster than she expected, and the newer analysis rests on a much stronger evidence base. Her update also means that much of the rigorous research on this question is recent, and the newest rigorous studies are not uniformly discouraging, as the Dominion and classroom studies above show. Rigorous measurement does not show that dietary change is impossible. It shows that most interventions tested so far have produced small effects, and that some stronger ones may work. And, as Sebo and Singer caution, even the best-designed studies of individual interventions cannot capture how advocacy shifts social norms over time.
4. There are a lot of people who have already changed their diets in the U.S.
The proportion of vegetarians and vegans in the U.S. is probably 9 percent. I have extensive analysis on this point in two different posts (Stănescu, 2019, 2025). I realize this may be a significantly higher number than you have heard before, so please take a look at the research I have compiled.
Ironically, the strongest numbers are from the meat industry itself. They have monthly polls on this exact question, so we have lots and lots of data, every month. Since these studies are funded in part by the beef and pork checkoff programs, any industry bias is presumably in the opposite direction, which is part of why I prefer their data to, say, a poll by Vegetarian Times. Consistently, what they find is a fairly high number of vegetarians, vegans, and semi-vegetarians. For example, the August 2026 Meat Demand Monitor, released September 1, reports 4 percent vegan, 4 percent vegetarian, and 10 percent flexitarian/semi-vegetarian (Tonsor, 2026, p. 3). These are much higher numbers than most people assume. According to this industry-funded survey, 18 percent of respondents identify with these diets, including 4 percent who identify as vegan.
5. I agree with you that we need to be more effective.
A representative from PETA asked me to provide recommendations for their advocacy and communication strategies. Here are five areas I suggested to them where I think more research and support could be helpful:
Perhaps some of these approaches will work, and perhaps some won’t. But that is a reason to study them. If you are interested, I can share more data and thoughts on each of these points.
Overall, I do not think it is fair to describe dietary advocacy as an approach that has repeatedly failed. We do have evidence that some forms of it can work. How much more funding it should receive is still a question worth studying. But, as Sebo and Singer argue, we also need to consider how activism changes social norms over time. If we discount those changes simply because they are harder to measure, we may underestimate the value of funding this work.
Therefore, we should continue to fund research on the effectiveness of dietary advocacy and, until that research is finished, advocacy in that direction.
Again, I would like to express my appreciation to Bruce for his post. He and I have disagreed in other areas, but on this issue, I strongly support his points. I also appreciate each of you working in veganism, vegetarianism, or animal rights. You are doing the most important work there is, and I am grateful to each of you for your work.
References
Bianchi, F., Dorsel, C., Garnett, E., Aveyard, P., & Jebb, S. A. (2018). Interventions targeting conscious determinants of human behaviour to reduce the demand for meat: A systematic review with qualitative comparative analysis. International Journal of Behavioral Nutrition and Physical Activity, 15, 102. Article.
Botchway, M., Turner-McGrievy, G. M., Crimarco, A., Wilson, M. J., Davey, M., Wilcox, S., & Frongillo, E. A. (2022). “They eat what they eat, I eat what I eat”: Examining the perspectives and experiences of African Americans who adopt plant-based diets. American Journal of Lifestyle Medicine, 16(3), 363–373. Article.
Cole, M., & Morgan, K. (2011). Vegaphobia: Derogatory discourses of veganism and the reproduction of speciesism in UK national newspapers. The British Journal of Sociology, 62(1), 134–153. Article.
Green, S. A., Smith, B., & Mathur, M. B. (2026). Meaningfully reducing consumption of meat and animal products is an unsolved problem: A meta-analysis. Appetite, 216, 108233. Article.
Haile, M., Jalil, A., Tasoff, J., & Vargas Bustamante, A. (2021). Changing hearts and plates: The effect of animal-advocacy pamphlets on meat consumption. Frontiers in Psychology, 12, 668674. Article.
Harper, A. B. (Ed.). (2010). Sistah vegan: Black female vegans speak on food, identity, health, and society. Lantern Books.
Hodson, G., & Earle, M. (2018). Conservatism predicts lapses from vegetarian/vegan diets to meat consumption (through lower social justice concerns and social support). Appetite, 120, 75–81. Article.
Jalil, A. J., Tasoff, J., & Bustamante, A. V. (2023). Low-cost climate-change informational intervention reduces meat consumption among students for 3 years. Nature Food, 4(3), 218–222. Article.
Kramer, L. A., & Landry, P. (2026). How the sausage is made: Testing the effectiveness of an informative video in promoting sustainable food consumption. Scientific Reports, 16, 18492. Article.
MacInnis, C. C., & Hodson, G. (2017). It ain’t easy eating greens: Evidence of bias toward vegetarians and vegans from both source and target. Group Processes & Intergroup Relations, 20(6), 721–744. Article.
Markowski, K. L., & Roxburgh, S. (2019). “If I became a vegan, my family and friends would hate me:” Anticipating vegan stigma as a barrier to plant-based diets. Appetite, 135, 1–9. Article.
Mathur, M. B., Peacock, J., Reichling, D. B., Nadler, J., Bain, P. A., Gardner, C. D., & Robinson, T. N. (2021). Interventions to reduce meat consumption by appealing to animal welfare: Meta-analysis and evidence-based recommendations. Appetite, 164, 105277. Article.
Minson, J. A., & Monin, B. (2012). Do-gooder derogation: Disparaging morally motivated minorities to defuse anticipated reproach. Social Psychological and Personality Science, 3(2), 200–207. Article.
Sebo, J., & Singer, P. (2018). Activism. In L. Gruen (Ed.), Critical terms for animal studies (pp. 33–46). University of Chicago Press.
Stănescu, V. (2019, July 1). Guest post: Response to the claim that only 2% (or less) of people in the United States are vegetarian. Critical Animal. Post.
Stănescu, V. (2025, July 20). You are not alone: The real numbers on vegetarians and vegans in America. Win for Animals. Post.
Tonsor, G. (2026, September 1). Meat Demand Monitor: August 2026 (Vol. 7, Issue 8, p. 3). Kansas State University, Department of Agricultural Economics. Report.
Insofar as people are making decisions based on peer group pressure and the fear of being ousted, sounds like some leadership / courage is needed!
YC never had any sort of monopoly on the best startups though, its own terms are much better than they used to be, and its still definitely the most prestigious. The question isn't "why didn't Anthropic join YC", it's "why, if AI truly makes most developers 10x more productive and transforms the unit economics of customers, are the valuations of a cohort of prestigious "AI-enabled" startups round about their non-AI enabled 2019 peers?". There really aren't enough people founding research labs for that to be the issue
Of course, any early stage valuations graph is as much a graph of investor sentiment as anything else, but if AI was really making these companies significantly more productive, investors would have to be very bearish on AI or YC selection effect to have gone down the toilet for that not to show in the data. I presume median data looks similar?
The best counterargument is that AI also makes it easier for competitors leading to less defensible business models even if AI actually enables them to grow faster, but if that was the case investor sentiment should be much more bearish on incumbents...
It seems like the most common objection to an international AI treaty with China is that it’s pointless as they will just defect. But there’s a strong case to be made that mutual verification of compliance is a tractable problem, so I suspect raising salience of this in AI policy discussions would make an international agreement more politically feasible. Presumably it would be a matter of only committing to what we can mutually verify now, then investing in stronger verification techniques as they’re developed. I find IFP’s policy recommendations for accelerating verification R&D to be a good starting point for thinking about this, especially in a near-term practical-to-policymakers way.[1] Ideally we would have some policymaker-legible brief memo/executive summary for verification proposals to point to that gets frequently brought up where it matters, i.e. Get members of Congress to write open letters about it, get journalists to ask about it, emphasize it in interviews that AI safety experts go onto, emphasize it at AI dialogues with the White House by those who can attend, etc. Granted, many of them are likely using motivated reasoning to dismiss this regardless of any evidence presented as a result of industry lobbying, but I reckon it would help to put more pressure on them to address it. From what I've seen thus far, the response from international slowdown advocates has mostly been "it's also in China's interest to slowdown, superintelligence kills everyone!" Sure, but I think it helps to make our proposals more skeptic-robust.
Why am I not bullish on civilizational refuges for reducing bio xrisk? I believe bio xrisk is concentrated in E2H scenarios or AI takeover, and neither of these things benefit much from refuges. For both of these threat models, there is value in refuges but only if they are a safe staging area to take back control of the environment or against an AI, which means you probably need a full industrial base worth of people/infrastructure to be protected. Society-wide biohardening and PPE is therefore the better option even on purely longtermist grounds (and has the additional property of also being much more valuable on non-longtermist grounds as well!)