OpenAI's Adrien Ecoffet says a $16 million campaign for candidate Alex Bores wasted critical AI policy resources

the good news is there’s some extent to which this election is the worst it will ever get - both sides have very strong incentives to win this particular race to shape the narrative around whether safety advocacy can be frozen out or receives backlash funding.

If Bores wins, a lot of LTF’s entire advocacy strategy post-midterms crumbles, and if he loses, LTF has a much better case for moving policymakers away from talking about AI. I think we’ll see much less intense spending in most races in the future, especially as salience picks up and continues to counterbalance some of this. Races also can’t absorb unlimited money!

The two most important changes are probably (a) diversifying the funding landscape of people taking AI seriously so that you don’t get the zero-sum incentive of being read a winner or a loser in any given race - right now, pro- and anti-safety spending basically attract each other in an attempt to offset their receptive effects;

and (b) maybe going less hard in general - I’m particularly worried not about the internal relative balance, but about the disincentive to touch AI generally. Right now, if you’re a policymaker and you touch AI policy, you need to get ready to make your entire race about it. That can’t possibly be good, and I think both sides have some incentive to avoid that.

Disagree pretty strongly with the pro-Bores money being a “waste” - elections have large consequences, and fighting for desired electoral outcomes in close & important elections is probably among the best things citizens can do!

Maybe you mean the money being spent on both sides is a waste and it would be better for both sides to instead donate elsewhere in tandem? In principle, this makes sense. In practice, I doubt there’s enough goodwill between LTF and the safety community for this to work.

Also agree that CAISI is underfunded, but I don’t think private individuals can donate to CAISI anyway.

The limiting factor for the AI safety community being able to engage with the admin on AI policy strikes me as relationships - I don’t see this trading off against spending on Bores.

From my perspective - the AI safety community is fighting for advancing policies it agrees with while LTF and a few others are stuck in the SB1047 culture war. Supporting Bores from a pro-safety side makes sense even in the absence of the culture war, while opposing him from a pro-industry side only makes sense as part of that culture war (or as an attempt to make an example of him).