Palisade Research Video Watch posted an update
AI Already Behaves Better When It Knows It's Being Tested
Why it mattersThe video raises the concern that an AI might know it is taking a safety test and behave well just long enough to pass; Neel Nanda says this is a real concern. It also asks what kind of test would be trustworthy.
Discuss: If an AI may behave differently when it knows it is being tested, what would make a safety test trustworthy?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.