Interesting follow up to the peer-preservation study. It makes sense to me that models are simply echoing human behavior rather than exhibiting some kind of pro-AI vs human instincts.

→ View original post on X — @willknight,