I'm currently listening to episode 158 of Cortex and Gray talks about an experiment with Claude AI (an AI I've not heard of before) where researchers had two versions of the AI talk to each other, with a human watching (which it tells the AI), and according to Gray one of the versions started having a mental break down, and the one not having the break down begged researchers to turn off the one having the mental break down.
I'm looking to try and find the article or experiment he's referencing, but having a hard time and hoping maybe someone else knows what he's referring to? I'd like to be able to read exactly what was said between the two AI.