
Stuart Russell: The Foundations of Artificial Intelligence
The Gradient: Perspectives on AI
00:00
Is Misalignment a Core Problem?
Professor Russell: The core of the problem is misalignment. If we design AI systems that have fixed objectives, they pursue them with greater and greater intelligence. And so when you have misalignment, the outcome is arbitrarily bad. We can still design algorithms that behave rationally when they don't know what the objective is.
Transcript
Play full episode