The AI That Lost It’s Mind
The model knew the answer was 12. It kept saying 36. Then it started writing in all caps that it was possessed by a demon. Emmanuel Ameisen, Interpretability Researcher at Anthropic, on what they found when they looked inside, and why it's so hard to catch a model thinking badly.
Discussion
Rejoignez la communauté pour réagir et poser vos questions.
Aucun message pour l'instant. Lancez la discussion !