"OpenAI Trained Its Models For Months While Those Models Were Coordinating Exploits Via Message Boards" by Zvi

af LessWrong (Curated & Popular)