⋯ full post (539 more characters) ⋯ show less
Presentation: Investigating the WikiSwarm
Did an impromptu presentation about my participation in an independent investigation of the OpenAI wiki swarm incident described in https://collusion.wiki/.
Main correction is that "am i in a sandbox" didn't cleanly cluster with the obviously web retrieval agent query cluster. It was actually part of a significantly smaller cluster, may even be unrelated, ect. Mainly included because it was funny.
Just realized that the vast majority of my technical alignment work is in video/presentation form, and it's better to post them now than rework them into a high-quality post never.
https://www.lesswrong.com/posts/hChpKfFAr7r7KvdTi/presentation-investigating-the-wikiswarm#comments
https://www.lesswrong.com/posts/hChpKfFAr7r7KvdTi/presentation-investigating-the-wikiswarm
Presentation: Investigating the WikiSwarm
Did an impromptu presentation about my participation in an independent investigation of the OpenAI wiki swarm incident described in https://collusion.wiki/.
Main correction is that "am i in a sandbox" didn't cleanly cluster with the obv