Explore the Paired Open-Ended Trailblazer (POET) algorithm in this comprehensive video explanation. Delve into the paper's innovative approach to generating increasingly complex learning environments and their solutions. Learn about evolutionary search algorithms, environment encoding, agent modeling, and the POET main loop. Discover how POET creates a powerful curriculum for agents to master challenging environments, and compare it to direct-path approaches. Gain insights into environment mutation algorithms, result analysis, and potential extensions of this groundbreaking research in open-ended learning.
Overview
Syllabus
Intro to POET
Evolutionary search explained
POET algorithm
Environment encoding
Agent and reward modeling
POET main loop
Environment mutation algorithm
Results
Direct-path curriculum comparison
POET "unstucks" agents
Extensions and outro
Taught by
Aleksa Gordić - The AI Epiphany
