Reproduce “The world as a neural network” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Vitaly Vanchurin
Vanchurin proposes that learning in a neural network could yield familiar physical descriptions in different limits. He treats weights and biases as trainable variables and neuron states as hidden variables, deriving quantum-like, classical, and gravitational equations under selected assumptions. These conditional arguments do not test whether the universe is a neural network or follows these mechanisms.
Key insights
- Fundamental physical laws, spacetime, and even observers can emerge from simpler, underlying neural network dynamics rather than being built-in.
- The universe can be conceptualized as an AI system, fundamentally linking physics with machine learning and information theory.
- The phase of the quantum wave-function can be identified with the free energy of an ensemble of the network's hidden variables.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.