What's Really Going On in Machine Learning? Some Minimal Models

The Hidden Workings of Machine Learning

Why does machine learning work at all? Neural networks power everything from image recognition to language models, yet the science behind their success remains strangely elusive. In this essay, Stephen Wolfram strips machine learning down to its barest forms—minimal models built from simple rules—and shows that even at this level, systems can learn. What emerges is a surprising picture: machine learning doesn't rely on carefully engineered structures but on the natural complexity of the computational universe.

Seen this way, machine learning is less about hidden design and more about sampling complexity. Wolfram's exploration offers not only clarity about why AI works but also perspective on its limits—why some successes resist explanation, and why the field may never yield a simple unifying theory.

Information and Media Inquiries

October 28, 2025 Publication
Publicity and Interviews: sw-media@wolfram.com
Non-Fiction

Contents

• The Mystery of Machine Learning
• Traditional Neural Nets
• Simplifying the Topology: Mesh Neural Nets
• Making Everything Discrete: A Biological Evolution Analog
• Machine Learning in Discrete Rule Arrays
• Multiway Mutation Graphs
• Optimizing the Learning Process
• What Can Be Learned?
• Other Kinds of Models and Setups
• So in the End, What's Really Going On in Machine Learning?
• Historical & Personal Notes
• Thanks

About the Author

Stephen Wolfram is an award-winning scientist and bestselling author, and the creator of some of the world's most respected software systems—Mathematica, Wolfram|Alpha and Wolfram Language. For more than 35 years, he has been CEO of the global technology company Wolfram Research, as well as responsible for a series of groundbreaking advances in basic science, including the recent Wolfram Physics Project.

Other Books by Stephen Wolfram