108. Last Week In AI — 2021: The (full) year in review
Towards Data Science
00:00
Optimum Agents Seek Power
The argument here is that agents will tend to navigate towards areas of state space with more down stream, down stream options. And this is where you get the that optimal policies tend to seek power. This has been high lighted as like one key piece in the kind of safety related arguments that come down from this.
Transcript
Play full episode
