How Researchers Test AI for Hidden Goals — Apollo Research

So you were scrolling on Marketplace, and there it was, the bike you'd been searching for.

It goes beyond the title with direct discussion of like, model, reward, including: So you were scrolling on Marketplace, and there it was, the bike you'd been searching for.

1. 01So they might realize, for instance, that they are being tested
2. 02OK, so we said that behaviorally, a reward seeker looks totally the same as an aligned model
3. 03The first thing that you might try is, well, can we just ask the model?