Open-endedness: A new grand challenge for AI
summary
Argues for creating open-ended algorithms that continually invent novel, increasingly complex tasks rather than optimizing a single objective; contrasts Deep RL on Atari and standard evolutionary algorithms with biological evolution, a never-ending algorithm that over ~3.5 billion years produced photosynthesis, flight, and evolutionary process produced human intelligence as one surprising outcome. Claims the goal is algorithms that keep producing unpredicted innovations, and points to evolutionary-style search as the primary tool.
AI-written summary · about
related talks
50mOptimal control, no! Than what? (...)
52mThe Journey of AI in Games: A Technical and Philosophical overview
17mBeyond Automation: AI and the Evolution of Robotic Intelligence
49mBeyond GenAI
04mFrom big data to human-level artificial intelligence
13mArtificial General Intelligence: Why aren't we there yet?
39mNeuro-Diverse AI: Enhancing Creativity through Unconventional Strategies
32mThe Road to General AI