The Machine Intelligence Research Institute (MIRI) is interested in reasoning about highly advanced AI systems before they exist, specifically for developing models of safety and control for such systems. In this EA Global: San Francisco 2016 talk, MIRI's Andrew Critch examines some criteria for "ideal" logical induction, a new algorithm for logical induction that satisfies many desirable properties, and some implications for what a very powerful AI system is able to learn.

In the future, we may post a transcript for this talk, but we haven't created one yet. If you'd like to create a transcript for this talk, contact Aaron Gertler — he can help you get started.