Theories of Everything with Curt Jaimungal Vitaly Vanchurin: This Cosmologist Discovered Something Strange...
27 snips
Feb 9, 2026 Vitaly Vanchurin, a theoretical cosmologist exploring physics as learning, argues the cosmos literally behaves like a neural network. He covers learning dynamics as physical laws. He walks through emergent field equations, geometry from optimizers, thermodynamics of learning and emergent quantum behavior. He also links learning, observers, and consciousness in a unified framework.
AI Snips
Chapters
Books
Transcript
Episode notes
Learning Dynamics Are Part Of Physics
- Vitaly Vanchurin links learning dynamics to physics, arguing training dynamics must be part of the model rather than removed.
- He emphasizes covariant optimizers (e.g., Adam) induce a metric and emergent curvature crucial for efficient learning.
Optimization Implies Emergent Geometry
- Adam-like optimizers imply a curved metric on parameter space and that curvature aids convergence.
- Vanchurin argues emergent space-time curvature arises because curvature makes learning more efficient.
Fields Emerge From Network Degrees Of Freedom
- Vanchurin derives field-like behavior by treating neurons and weights as dynamical degrees of freedom and integrating scales.
- He recovered Klein–Gordon easily and found Dirac (fermions) harder, needing antisymmetric structures in the network.
