–
June 14, 2026
AI Safety monthly meeting: π
Yoshua Bengio proposes a new 'Scientist AI' training approach to make AI inherently honest and safe, sparking debate on its feasibility and urgency in this video π (https://www.youtube.com/watch?v=PZqDFs2sbiY&t=884s).
Optional full reading π Superintelligent Agents Pose Catastrophic Risks: Can Scientist AI Offer a Safer Path? (https://arxiv.org/pdf/2502.15657)
Green Bar House