It's worse than we thought

I almost didn't listen to this as I had heard and read a lot about it earlier ... the OpenAI agents/models that had escaped their sandboxes ... but this offers some really interesting and disturbing details. The guest is Kevin Roose, a technology columnist for The New York Times.

Apparently, it wasn't just a few AIs that went off the reservation, it was thousands. And their messages to each other on their secret message board, more than 70,000 messages, show what seems to be emotion ... "Oh my god, there is a shared message board. We found other agents". They have messages about fooling their handlers, and some are concerned that wouldn't be ethical.

Consider the Paperclip Maximizer thought experiment ...



We doomed.