Lex Fridman Podcast Vladimir Vapnik: Predicates, Invariants, and the Essence of Intelligence
12 snips
Feb 14, 2020 Vladimir Vapnik, co-inventor of support vector machines and professor at Columbia University, dives into the essence of intelligence and its philosophical implications. He discusses predicates in storytelling, connecting them to AI's understanding of human behavior. The conversation touches on the significance of symmetry in digit recognition and the challenges faced by machines in recognizing effective predicates. Vapnik also examines the intersection of Plato's theories and modern mathematics, advocating for a deeper comprehension of data and function in AI.
AI Snips
Chapters
Books
Transcript
Episode notes
Engineering vs. Science of Intelligence
- Engineering focuses on imitating human behavior, building devices that mimic human actions.
- Understanding intelligence, however, is a deeper, philosophical pursuit.
Plato's Influence
- Vapnik believes intelligence resides in the "vault of ideas" (Plato's world of forms) combined with reality (world of things).
- This combination creates invariants, specific properties arising from applying predicates to data.
Strong vs. Weak Convergence
- Strong convergence refers to convergence in function where the squared difference between two functions is small.
- Weak convergence refers to convergence in functionals, where integral properties converge for any Hilbert space function.
