This is a special post for quick takes by RobertM. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
If the three frontier labs (OpenAI, Anthropic, and GDM) actively want to be forced to engage in a coordinated slow down (or implement other costly safety measures), but find that the US government is unwilling to play ball, maybe they should try using a "friendlier" regulatory apparatus, like the UK's. It doesn't have to be the kind of thing that they couldn't get out of by aggressively fighting it if they merely need a legal pretext to be allowed to do it in the US at all, plus an external source of compliance validation and enforcement (which doesn't care about geography, particularly).
(edit: the subtext here is people talking about antitrust concerns as bottlenecks for cooperation between the labs, particularly around slowing down the development of frontier AI.)
Lehane said OpenAI had been in talks with Anthropic and Google for several weeks and did not believe the three AI firms needed an antitrust waiver to coordinate on safety issues, according to the report.
This has been bugging me all day. I am sincerely worried about regulatory capture, enough I might oppose a whole pacing bill over it.
It feels like the "thing they definitely need regulation for" should be a simple crux, and it is not. At this point they're confessing future models are certain death, which does not take government coordination to not do. This may be about labs #1 and #2 getting #3-5 on board, while they have leverage (the leverage is threatening to kill everyone).
It doesn't really look like it kneecaps open models or smaller competitors.
So yeah the regulation is the "stop Grok" step, and it would pick up any mid-competitor who gets big enough. If it were full duopoly, Dario's framework would not really need to be government-administered.
6RobertM
Often all you need to coordinate is to establish common knowledge of the facts around which you want to coordinate. If all parties involved actually want to move to the better equilibrium, it then becomes possible.