Speech 8: OpenAI-Ilya Sutskever: Meta-Learning and Self-Play

Contents:

Theory:

• Introduction & supervised learning

• Reinforcement learning

• Meta learning

Application:

• Hindsight experience replay

• Sim2Real with meta learning

• Learning a hierarchy of actions with meta-learning

• Limitation of high capacity meta-learning

• Self-play

• Can we train AGI using the self-play?

• Learning from human feedback