#66 – Michael Cohen on Input Tampering in Advanced RL Agents
Hear This Idea
00:00
Introduction
Michael Cohen is a D-Phil student at Oxford, soon to be starting a postdoc with Professor Stuart Russell at UC Berkeley. He does technical AI safety research, so figuring out ways to design general intelligence systems which can be expected to behave safely. We talk about the difference between supervised, unsupervised and reinforcement learning. And we also talk about some arguments people have made about how it's easy to imagine the wrong thing when we reason about how reinforcement learners will behave.
Play episode from 00:00
Transcript
