Geoffrey Irving

3 posts
Fine-Tuning GPT-2 from Human Preferences
AI Safety Needs Social Scientists
AI Safety via Debate