Joseph Allen
The CEO of Anthropic openly admits they have no idea what's happening in their models. Models that, earlier in his post, say can lead to "risk of losing control of AI systems, misuse of AI for cyberattacks and bioterrorism, and serious economic disruption." https://darioamodei.com/post/we-must-pace-the-frontier
Joseph Allen
Wherein the CEO of Anthropic admits that the recent incidents happened b/c they screwed up b/c they were moving too fast. -"Many things go wrong...because of problems in execution" -"the recent alignment incidents...were caused in part by imperfect filtering of broken reinforcement learning environments. This was an effort we and our vendors executed reasonably diligently, but not well enough." https://darioamodei.com/post/we-must-pace-the-frontier
