Tyler Cowen showing up to defend EA here gives me mild nightmares because I remember him calling the top in 2022: xitter.cf/MargRev/status/1503404… But his bottom line that EA is “directionally correct, even if you should not follow its recommendations all the way” seems directionally correct to me, and in line with what I was getting at last week about EA’s pragmatism: xitter.cf/albrgr/status/21058269… He says more attention to AI safety would be good, and gives EA some credit there. I’d add that EAs were saying this years before most others, which is a pretty good example of Tyler’s closing line that “interesting and valuable ideas” can spring from weird places. (And I say this as someone who was pretty skeptical on the AI safety work for a long time.) And I agree that EAs can be offputtingly arrogant. That’s one reason I’m really glad we practice “worldview diversification” at CG (coefficientgiving.org/resear…) - though I still wish we had a more robust intellectual grounding for it and a better/more principled way to translate it into ultimate allocations. I’m also sympathetic to Tyler’s view that some EAs “perceive a high risk of doom without sufficient evidence,” though I also think that the best available estimates from people who are trying as hard as possible to be neutral (e.g., xitter.cf/Research_FRI/status/20…) suggest plenty of reason for concern. I think a lot of the problem here is just the fundamental difficulty of our epistemic position here, as I think @Research_FRI and @michael_nielsen have done a good job pointing out. With concrete harms like the Hugging Face attack starting to show up in practice, I do sometimes feel like Tyler is unfairly shifting the epistemic goalposts (though not particularly in this case). And as a small correction - CG doesn't have a “$40 billion fund,” or really an endowment of its own at all. I think that number might have come from a back-of-the-envelope guess I gave here (semafor.com/article/09/03/20…), building on @nanransohoff’s estimate of what all AI-linked donors might eventually give each year. Anyway, I agree w/ the upshot that more EA at the margin would be a good thing!