[adrianmonk]

> Maybe all large models will behave like this, and only by putting in extremely rigid guard rails

I've always believed that as soon as we actually invent artificial intelligence, the very next thing we're going to have to do is invent artificial sanity.

Humans can be intelligent but not sane. There's no reason to believe the two always go hand in hand. If that's true for humans, we shouldn't assume it's not true for AIs.