What happens when you teach neural networks the deepest symmetries of nature?

This book is a hands-on introduction to one of the most exciting frontiers in science: the convergence of artificial intelligence and quantum field theory. Written for physicists curious about machine learning and for AI practitioners drawn to fundamental physics, it bridges both worlds with clarity, rigor, and working Python code.

The journey begins with a surprising discovery. The renormalization group, one of the most powerful tools in theoretical physics, maps directly onto the information flow through neural network layers. Gauge symmetry, the principle that governs every fundamental force, provides architectural blueprints for AI systems. Readers build a neural network from scratch that identifies phase transitions without being taught any physics, demonstrating how AI can rediscover fundamental principles from raw data alone.

The book then examines how AI tackles each type of quantum field. Neural networks reveal exotic scalar field phases that traditional methods miss. DeepMind's FermiNet achieves chemical accuracy for molecules with up to 30 electrons. MIT's gauge-equivariant normalizing flows reduce lattice QCD autocorrelation times by a factor of 100, conquering the critical slowing down that has stalled simulations for decades. Transformers compress million-term scattering amplitudes into single equations.

The final chapters look ahead to AI systems that do not merely calculate but create. Systems like MELVIN design quantum experiments no human has imagined. Language models solve bootstrap equations. Neural networks propose pathways toward grand unification. The book closes with the emerging partnership between quantum computers and classical AI, a combination that may finally unlock QFT's deepest unsolved problems.

Includes 17 chapters, a glossary, working code examples, and a companion GitHub repository.