AI summary by TwinMind
Richard Feynman on Computation (Stephen Wolfram) | AI Podcast Clips
Lex Fridman
12m 16s
Feynman-Wolfram Collaboration at Caltech and Thinking Machines
- Worked together at Caltech and as consultants at Thinking Machines Corporation
- Thinking Machines located down the street, ultimately ill-fated company
- Wolfram predicted company wouldn't work with their strategy
- Feynman's response: "what do we know about running companies, just let them run their company"
- Feynman viewed Wolfram's business interests as distraction from pure science
- Wolfram saw companies as "effective machine for figuring things out and getting things to happen"
Feynman's Approach to Mathematical Computation
- Feynman had elaborate personal methods for doing integrals
- His meta-idea: take intuitive methods and make computers follow them
- Actual implementation different - built "bizarre industrial machine"
- Main problem: turning results into human-understandable form, not doing the integral
- Feynman gave Wolfram pile of calculational methods for particle physics from 1950s
- Said "it's more use to you than to me"
- Wolfram still has the files, never returned them due to "finiteness of human lives"
- Feynman's attempt to systematize integral methods for particle physics
Feynman's Calculation vs Intuition Paradox
- Feynman remarkably good at calculating complex problems
- Thought calculation was easy because he excelled at it
- Would do complicated quantum field theory calculations
- Wouldn't tell anyone about calculations, thought they were unimpressive
- Instead presented simple intuitive explanations he invented afterward
- Having done calculations made intuition easier - "easier to have good intuition when you know the answer"
- Led to mystified colleagues who followed intuition but couldn't understand why it worked
- The real reason it worked: Feynman had done all the underlying calculations
Quantum Computing Collaboration (1980-81)
- Worked together on quantum computers before anyone had heard of them
- Typical dynamic: Feynman would calculate by hand on blackboard
- Wolfram would use computer, leading to arguments about methodology
- Feynman complained Wolfram was one-third his age
- Many quantum computing issues they identified (particularly measurement process) still relevant today
- Example of "remarkable repetition of history" in science and technology
- Things eventually get nailed down but often take decades
Rule 30 Cellular Automaton Discovery
- Wolfram working on Rule 30 cellular automaton at Thinking Machines
- Used Connection Machine parallel computer to generate giant printout
- Same printer type used for microprocessor layouts - high resolution, large format
- Physical analysis: crawling on floor with rulers measuring pattern features
- Feynman's key question: "How did you know Rule 30 would produce complicated behavior?"
- Wolfram's answer: "I didn't know, I just enumerated all possible rules and observed"
- Feynman's relief: "I feel a lot better, I thought you had some intuition I didn't have"
- Demonstrated principle: "you really can't have intuition about irreducible [computation]"
Computational Irreducibility and Human Limitations
- Hard for brilliant physicists like Feynman to accept lack of compressed intuition
- Can't have "compressed clean intuition about how the whole thing works"
- Feynman on edge of understanding computational principles
- Found computation interesting, was "poking at" the intuition problem
- Discovery challenge: even systematic enumeration sounds easy but isn't
- Wolfram initially missed Rule 30 due to preconceptions - "had intuition it shouldn't be there"
Wolfram's Transition from Physics to Computational Thinking
- Started with reductionist physics: "thrown the universe, go figure out what's inside"
- Shifted to building computer tools and languages
- Computer language design similar to physics: drill down to find primitives
- Key difference: building up from primitives to create "artificial universe"
- Transition from studying natural universe to creating any universe desired
- Experience building computer languages gave different attitude toward possibilities
- Enabled exploration of artificial universes vs being constrained by natural universe
Programming as Cognitive Extension
- Programming creates "another brain" beyond mind and pen
- Computer as tool like telescope but fundamentally different
- More general capability for understanding and seeing patterns
- Building computational languages requires creating abstraction levels
- Must be open to different structural possibilities
- Wolfram aware of tendency to "miss the obvious"
- Attempts to factor this awareness into approach, though may not always succeed
