Lex Fridman Podcast #221 – Douglas Lenat: Cyc and the Quest to Solve Common Sense Reasoning in AI
42 snips
Sep 15, 2021 Douglas Lenat, the founder of Cyc, shares insights from his 37-year journey in common-sense reasoning for AI. He discusses the complexities of knowledge representation, comparing trees to graphs for AI understanding. Lenat emphasizes the dual impact of the internet on knowledge and ignorance. He also reflects on foundational AI discussions from 1984, the evolution of the semantic web, and the interplay of human emotions in AI design. Their conversation underscores the need for deep, contextual intelligence to tackle significant global challenges.
AI Snips
Chapters
Transcript
Episode notes
Common Sense in AI
- AI systems need common sense knowledge to avoid brittleness and truly understand the world.
- This knowledge acts as a foundation, enabling them to handle novel situations and explain their reasoning.
Trash Truck Incident
- Douglas Lenat shares an anecdote about driving behind a trash truck whose back exploded.
- This illustrates the need for AI to handle unexpected real-world situations and utilize common sense knowledge.
Scale of Common Sense
- Cyc's knowledge base requires tens of millions of rules, not just facts, to capture common sense.
- These rules act as guidelines for making sense of facts and navigating everyday situations.