Keynote: A Mathematician Muses about Macros, @Models, and ML
summary
A mathematician explains why adoption depends on specifics not hype and demonstrates it: a Newton fractal coded as a SwiftUI prototype (200×200 grid) that took minutes and looked pixelated, then reimplemented with Metal reimplementation for real time zoom to get interactive performance. Argues ML’s failure modes are concrete — dependence on large clean datasets and sensitive initialization causing divergent outputs — and points to Bonjour as an example of a small idea that became pervasive. Advocates watching what’s new over isolated sessions.
AI-written summary · about
chapters (20)
- 0:00From Harvey to Bernoulli
- 2:37Escher and hyperbolic space
- 4:38Foundations for WWDC
- 8:02The importance of maths
- 10:07Geometric proofs
- 13:40Changing the rules
- 17:35Colleagues and mentors
- 20:08Trusting your intuition
- 22:11Finding your place in ML
- 24:35The story of Bonjour
- 29:47Storytelling and adoption
- 32:44Choosing technologies
- 35:28Patience in development
- 37:15Community support
- 39:44Calculus and derivatives
- 43:33Fundamental theorem
- 46:11Newton's method
- 52:19Complex dynamics
- 55:49Continuous learning
- 59:17Looking ahead
related talks
54mKeynote: Can you trust your (large language) model?
44mKeynote Speaker - Simon Willison
02mKeynote: Scaling Intelligence: Infrastructure as the Foundation
01h 07mKeynote + Conversation
54mKeynote: Can you trust your (large language) model?
05mKeynote: Teaching LLMs New Knowledge
34mAttention is not all you need - Keynote
44mThe LLM Revolution: From Inception to the Future of Thinking Machines