AI Alignment Game Tree

Marie Reay 3.02

12:00 pm - 12:30 pm


Come and help populate the “game tree” of AI alignment strategies! We will be working in small groups to propose “builder” strategies that might help align AI systems and corresponding “breaker” strategies that might inhibit them. Some knowledge of existing alignment proposals is very helpful, but not required.

Optional pre-reading: the Alignment Research Center’s “Eliciting Latent Knowledge”; Evan Hubinger’s “An Overview of 11 Proposals for Building Safe Advanced AI”; John Wentworth’s “The Plan”; Dan Hendrycks’ “Pragmatic AI Safety”; and Evan Hubinger’s “Chris Olah’s Views on AGI Safety”.