2025-10-11
drapery falls
Dwarkesh interviews Nick Lane on his mitochondria-centric view of eukaryotes, one of those frameworks which would be useful even if it were wrong, because you can derive a lot of very interesting hypotheses which answer many seemingly unrelated questions. For example, why certain cell types seem to exhibit signs of aging more than others (through mechanisms like sarcopenia or dementia), to whether we should expect to find life in the Galilean moons.
Richard Hanania interviews Stephen Pinker on common knowledge: in my opinion this is much better than the one he did recently with Tyler Cowen, because here they go directly into the political implications of what common knowledge being useful means, whereas it seemed to me that Tyler much preferred to keep things more abstract.
Cremieux argues that polls asking American’s about their attitudes towards political violence tend to overstate their actual levels of support.
Ian Leslie on ideological fragmentation in western politics, with a sort of meta-moderate position towards the political moderation debates, arguing that politicians representing any single particular ideology will have trouble winning since they will alienate every other group, and yet they still need to give the impression to ideologues in their tent that they are indeed being represented. I don’t know, because it’s my impression that the vast majority of people do not actually strongly subscribe to these arbitrary categories, but actually hold to the very popular ideology of “my own personal utility”. This just isn’t properly recognized by policymakers because to them, ideological categories are a means by which they can easily aggregate preferences for specific policies. Yet if policies could be designed such that they always provided the maximum public benefit with minimal cost, that seems like it should be quite popular.
On that note, Martin Susrik has a short piece on regulations in Europe around wild animals are insensitive to the people who are affected by them. I was pretty confused when I found out that in European politics, it’s common for parties representing rural voters to also be libertarian, which seemed inconsistent to me, as my view of rural voters was that they wanted farm subsidies. After watching Clarkson’s Farm, this position started making a lot more sense to me1. It seems to me there’s actually a strong case that farmer subsidies should actually be provided if you’re going to apply all these onerous regulations, but that they should be directly tied together, so farmers can be compensated for things like not being able to kill or drive away badgers on their property that are infecting their cows with tuberculosis, or can choose to forego compensation in return for being able to ignore it2.
Will McAskill suggests that a new EA cause area should be making sure that AGI goes well. It does seem unbalanced that within EA the current focus is exclusively on safety or pausing, which assumes ASI, ignores the nevertheless significant impacts that would occur if we “only” get AGI. Aside from the actual merits of the proposal (which I think are good), I think this is a great idea because lively internal debate is exactly what makes EA interesting and attractive to newcomers.
Epoch AI has a series of interviews3 with mathematicians on LLMs, one with Greta Panova on their current strengths and limitations, and another with Ken Ono on how they can be used in theoretical math research. They also ask each mathematician to rate from 1-10 how transformative AI will be and whether they think it should be paused or accelerated.
Edit: possibly relevant complaint of British proceduralism by Jacob Zucker.
To preempt the criticism that this would lead to negative externalities, this system needs not to be applied to every single regulation. When it comes to things like maintaining water quality, farmers themselves would want such regulation to exist because they would also be victimized by any tragedy of the commons. But if a regulation is unpopular among those that it applies to, and yet is supposedly to the public benefit, then presumably one ought to consider whether this benefit is one that the public is actually willing to pay for.
Not sure if they will have more, so I didn’t post this until now.