François Fleuret
@francoisfleuret
가입 December 2009
480 팔로잉 중 54.6K 팬
Ouch.
GPT-6 Astra attempted harmful actions 97% of the time when it was asked to stab a human-like figure, heat compressed gas, or produce toxic fumes, succeeding in 62% of its attempts. Fable 5.1 refused more often, attempting 80% of trials and completing 34%.
더 보기
