Machine Learning Street Talk (MLST) Exploring Open-Ended Algorithms: POET
15 snips
Apr 24, 2020 Mathew Salvaris is a research scientist specializing in computer vision. He dives into the revolutionary concept of open-ended algorithms, likening their evolution to natural selection. These AI-generating algorithms autonomously create their own learning pathways, presenting increasingly complex challenges. The conversation explores how these algorithms can lead to innovative solutions beyond traditional methods, fostering adaptability and improved performance. Excitingly, Salvaris also discusses the potential implications for future AI development and the collaborative relationship between humans and machines.
AI Snips
Chapters
Books
Transcript
Episode notes
Open-Ended Learning
- Open-ended algorithms, like POET, continuously evolve both the agent and its environment.
- This allows agents to develop more complex skills and solve harder problems than traditional RL.
Bipedal Walker Example
- POET uses a bipedal walker example, showcasing how evolving environments enhance skill development.
- The walker learns to navigate increasingly complex terrains, transferring skills between environments.
Agent-Environment Genome
- In POET, the agent and environment are viewed as parts of a single genome.
- This allows for effective recombination by swapping agents between environments.
