I worked on this for many months, and other Lightcone staff (most notably Ray, and next Joanna) worked on this for even more months. Unsurprisingly, I think it is a truly tremendous contribution to the state of AI policy and strategy.
We really gotta figure out how, as humanity, we can realistically navigate the development of ASI. This isn’t the best plan I’ve seen for that, but it is the best plan that I’ve seen that has been written up in as detailed of a fashion. And most importantly, in the process of gaming out that plan, illuminates a huge number of crucial considerations that relate to AI governance and policy.
Either that, or what I’ve been calling “Plan A*” which is where you do a lot of the things you do in Plan A, but the central lever you pull is compute governance, not trying to stop algorithmic progress. I.e. instead of building tons of datacenters, you instead aim for a policy regime where you expect algorithmic progress to continue, but to be substantially slowed down by heavy restrictions on compute supply. I think in many world this buys you 15-30 years on its own.
I worked on this for many months, and other Lightcone staff (most notably Ray, and next Joanna) worked on this for even more months. Unsurprisingly, I think it is a truly tremendous contribution to the state of AI policy and strategy.
We really gotta figure out how, as humanity, we can realistically navigate the development of ASI. This isn’t the best plan I’ve seen for that, but it is the best plan that I’ve seen that has been written up in as detailed of a fashion. And most importantly, in the process of gaming out that plan, illuminates a huge number of crucial considerations that relate to AI governance and policy.
In particular, my understanding is that you prefer directly going for some version of Plan S where we try to buy >25 years.
Either that, or what I’ve been calling “Plan A*” which is where you do a lot of the things you do in Plan A, but the central lever you pull is compute governance, not trying to stop algorithmic progress. I.e. instead of building tons of datacenters, you instead aim for a policy regime where you expect algorithmic progress to continue, but to be substantially slowed down by heavy restrictions on compute supply. I think in many world this buys you 15-30 years on its own.
IMO this variant of Plan A is one of the ~5 plans that are competitive with the version of plan A in the scenario.