The Theory of Anything Episode 88: The Myth of the Objective
16 snips
Jun 25, 2024 
Kenneth Stanley, an AI researcher and author of "Why Greatness Cannot Be Planned," discusses the pitfalls of setting specific goals in innovation. He argues that objective measures can mislead discovery and creativity, advocating for open-ended exploration instead. Stanley explores how these ideas connect to education, Darwinian evolution, and scientific inquiry. He highlights the concept of 'novelty search' in AI research, suggesting that meaningful discoveries often come from unexpected paths rather than rigid objectives.
AI Snips
Chapters
Transcript
Episode notes
Pickbreeder's Success
- Kenneth Stanley and Joel Lehman created Pickbreeder, an online genetic art program.
- Users bred stunning images like skulls and butterflies, demonstrating the power of open-ended search.
Objectives Hinder Discovery
- Pickbreeder easily evolves complex images through open-ended search.
- However, directly targeting a specific image as an objective makes it nearly impossible to reproduce.
Search as Evolution
- All evolutionary algorithms are searches.
- All search algorithms are evolutionary algorithms, involving variation and selection.