AI Seminar Series 2023: Khurram Javed

The AI Seminar is a weekly meeting at the University of Alberta where researchers interested in artificial intelligence (AI) can share their research. Presenters include both local speakers from the University of Alberta and visitors from other institutions. Topics can be related in any way to artificial intelligence, from foundational theoretical work to innovative applications of AI techniques to new fields and problems.

On March 10th, Khurram Javed— PhD student at the University of Alberta — presented “The Big World Hypothesis and its Ramifications on Reinforcement Learning" at the AI Seminar.

The big world hypothesis is that for many problems, the world is multiple orders of magnitude larger than the agent, and the agent cannot represent the optimal value function and policy even in the limit of infinite data. In this talk, Javed argues why the big world hypothesis is reasonable for many real-world problems. He shows — using existing and his own work — that algorithms that work well in the over-parameterized setting can fail when the big world hypothesis holds. Javed then shares an experimental protocol to benchmark algorithms under the big world conditions. Finally, he talks about promising solution methods for RL in big worlds. The main conclusion is that computationally cheap algorithms that learn continually are promising solution methods when the big world hypothesis is true. 

Watch the full presentation below:

Want to learn how you can kick-start your AI career? Find out more about Amii's Career Accelerator to find out more.