Amodei, Olah and co-authors from Google, Stanford, Berkeley and OpenAI publish “Concrete Problems in AI Safety”: side effects, reward hacking, scalable oversight, safe exploration and distributional shift
translated by AI, reviewed by the editors
· Science
- technology
- AI safety · Machine learning · Reinforcement learning
- merge
- Machine learning+Reinforcement learning → AI safety
- source
- arxiv.org ↗
¶ car56caj3