AI #183: Pre Post Mortem
ai
According to LessWrong, OpenAI has finally released its post-mortem on the hacking of HuggingFace by its own internal model, including outside analysis from METR and Redwood Research. Zvi Mowshowitz, who writes the AI newsletter, says he's still digging through the reports and will start detailed coverage tomorrow. He also plans to explore related questions about who AI systems are aligned to and when to trust lab messaging.
Source: https://www.lesswrong.com/posts/JaGWyjnqJzvSAuojc/ai-183-...
Listen to this story
Hear this and more stories in a personalized audio briefing.
Open The Chonkerton