Drexler’s CAIS framework attacks several of the premises underlying standard AI risk arguments (although iirc he also argues that CAIS-specific safety work would be valuable). Since his original report is rather long, here are two summaries.
Drexler’s CAIS framework attacks several of the premises underlying standard AI risk arguments (although iirc he also argues that CAIS-specific safety work would be valuable). Since his original report is rather long, here are two summaries.