UB 2023: Practical Reinforcement Learning: Lessons from 30 years of Research, Keynote by Peter Stone

The field of reinforcement learning (RL) has a long history of theoretical results that indicate when RL algorithms should work. Throughout this history, there has been a complementary thread of research that tests the theoretical assumptions by seeking to determine when (and how) RL algorithms do work in practice.

Drawing on 30 years of research results, mostly from the Learning Agents Research Group at UT Austin, this talk summarizes lessons learned about practical RL into four high-level topics: 1) Representation - choosing the algorithm for the problem's representation and adapating the representation to fit the algorithm; 2) Interaction - with other agents and with human trainers; 3) Synthesis - of different algorithms for the same problem and of different concepts in the same algorithm; and 4) Mortality - dealing with the constraint that in most practical settings, opportunities for learning experience are limited.

The talk concludes with a brief introduction to one of the largest ever commercial deployments of an RL agent, Gran Turismo Sophy, which in 2021 won a head-to-head competition against four of the world's best drivers in the Gran Turismo high speed racing game.

Across the research, business, political and societal spectrum, there are different ideas, excitement, and concerns about the future of AI. Academics and research community members are encouraged to voice their individual point-of-view at Upper Bound without restraint. Amii is committed to providing venues to discuss the future of AI in all its forms.