This is a special post for quick takes by Linch. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Have people looked into the ways in which AI safety legislation or regulations might speed up AI capabilities progress?
Many people here have talked about how technical AI progress for current-generation and next-generation might be more gated by technical AI safety than AI capabilities right now, especially at OpenAI (as a crude analogy, if you want to make cars go faster, there'll be points where the most important thing to work on are stuff like steering wheels and brakes, rather than direct engine and accelerator optimizations, because otherwise the cars are too unsafe to realistically go fast for long).
It seems pretty plausible to me that something similar might be true for near-term AI policies (to go back to the car analogy, perhaps in some situations the way you make cars go faster via policy is deregulation, raising the speed limit, etc. But it's also possible at points that improving highway safety, installing traffic lights, etc, will end up increasing long-term average speeds of cars much more than campaigning to increase the speed limit).
To the extent that you think speeding up AI progress is mostly bad and the short-term safety benefits don't compensate for them[1], it's probably good to map out ways in which short-term safety legislations and regulations are medium-term or long-term accelerationist in practice.