Scott Alexander at Astral Codex Ten Introducing Plan A, a proposal for AI safety built on agreements that recognise the principals do not trust each other. It is offered as an overall plan, but also as a stand-up comparison model
... it’s supposed to be a floor. When some politician proposes a data center ban, or says that we have to gut safety regulation to compete with China, or promises a job retraining program, think to yourself: does this person have a vision for where all of this ends up? If so, is it as good as Plan A? If not, consider demanding that they do better.
The full plan comes from AI Futures, which had an uncannily good prediction in 2021 of what would happen from then to now. It includes a link to the actual Plan A. The ACX essay linked above is Scott's sales pitch for it.
I don't know enough to have a credible opinion beyond the observation that these people are much smarter than I am, have good credentials that they are well-meaning, and have spent a lot of time on what could go wrong and how we prevent/fix that.
No comments:
Post a Comment