
Jensen Huang: NVIDIA - The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494
Jensen Huang discusses NVIDIA's extreme co-design approach and rack-scale engineering that powers the AI computing revolution
In this episode, Douglas Lenat discusses Cyc, his monumental 37-year endeavor to create a machine-readable encyclopedia of common-sense knowledge. Lenat explains that Cyc attempts to solve one of AI's fundamental challenges: enabling machines to understand and reason about the world with the kind of common-sense knowledge that humans take for granted. The project involves encoding millions of facts and relationships into a comprehensive knowledge base using formal logic and ontological engineering. Lenat describes how Cyc works by building a carefully structured knowledge representation system where concepts are related to each other in ways that preserve logical consistency. The discussion explores how to form and maintain a knowledge base of unprecedented scale, addressing challenges like global consistency versus local consistency in reasoning systems. Lenat emphasizes that knowledge representation is not merely about storing facts but about understanding the deep relationships and contexts that give those facts meaning. The conversation touches on the semantic web and how Cyc's approach could enable better data interpretation across the internet. Lenat discusses the development of OpenCyc, an open-source version of the knowledge base, which was created to engage the broader research community in expanding and improving the system. He addresses the inference problem, one of Cyc's central technical challenges, which involves efficiently deriving new knowledge from existing facts without computational explosion. Throughout the episode, Lenat reflects on philosophical questions about machine consciousness, reasoning, and what it means for machines to think. He shares insights about legendary AI pioneer Marvin Minsky and how his work influenced the field. The discussion includes thoughts on mortality, the nature of consciousness, and what Lenat would say to artificial intelligence if it could understand him. Lenat also offers advice to young people interested in AI research, emphasizing the importance of ambitious long-term projects. The episode captures both the technical depth of Cyc's architecture and the broader philosophical implications of trying to encode human understanding into machines. Lenat's perspective on AI differs from current deep learning trends, advocating instead for the continued importance of symbolic reasoning and explicit knowledge representation. His lifetime commitment to this project reflects a belief that certain fundamental AI challenges require sustained focus and that shortcuts through pure machine learning may not be sufficient for achieving true machine understanding.
“Common sense is not common, and that's the whole problem we're trying to solve with Cyc”
“The difference between knowledge and data is that knowledge is organized, structured, and interconnected in ways that enable reasoning”
“We need to understand that AI reasoning is not just about pattern matching but about understanding the deep relationships between concepts”
“Cyc is an attempt to codify the kind of knowledge that humans acquire through living and experiencing the world”
“The inference problem is the central challenge because we can encode millions of facts, but deriving useful conclusions from them efficiently is what matters”