Скачать с ютуб Social Reinforcement Learning talk at RLDM в хорошем качестве

Social Reinforcement Learning talk at RLDM

Social learning helps humans and animals rapidly adapt to new circumstances, coordinate with others, and drives the emergence of complex learned behaviors. What if it could do the same for AI? This talk describes how Social Reinforcement Learning in multi-agent and human-AI interactions can address fundamental issues in AI such as learning and generalization, and improve human-AI interaction. I demonstrate the difference between social learning and imitation learning, and show that when agents can learn how to socially learn from experts, they can generalize to fundamentally different environments at test time than they have experienced during training. I then present a method for selectively learning who and what to imitate by computing when following other agents’ policies would pay off under the learner’s own preferences. This technique, called PsiPhi-Learning, is a step toward enabling more human-like social learning. In the last part of the talk I discuss early work on training language models with human feedback, focusing on implicit cues present in the text itself, rather than manually curated binary labels. Together, this work argues that Social RL is a valuable approach for developing more general, sophisticated, and cooperative AI.