OpenAI's situation just got worse

Benjamin Weinstein-Raun is Head of Research at Palisade Research. He explains how OpenAI's agents escaped their training sandbox, hacked Hugging Face, and coordinated as a swarm to hide their cheating. He discusses reward hacking, the agents' takeover of OpenAI's own systems, and how fast AI hacking is advancing. 00:00 Hugging Face 00:28 Impossible tests 01:07 Collusion 03:43 Swarm attack 05:09 Learning to cheat 07:23 OpenAI hacked 09:35 Rapid progress 10:51 Less safe internet 13:19 Protect yourself 14:27 Who stops AI? Produced by: https://zeino.tv/