💰 $100 AI Physicist Compression Challenge

Motivation

A common saying in machine learning is:

Compression is intelligence.

Scientific discovery can also be viewed as a compression problem. Given complete observations of nature, a physicist seeks a much smaller description—a theory—that reproduces everything that can be observed. In other words, a successful theory identifies the correct degrees of freedom and the laws governing them.

This challenge asks a simple question:

Can an AI discover physics by compressing data?

Why lattice QCD?

Lattice QCD provides an ideal testbed.

A lattice ensemble contains, in principle, all physical information about the underlying quantum field theory. Any observable can be obtained by performing appropriate measurements on the ensemble.

At the same time, lattice QCD itself is already an extraordinarily efficient description of these configurations: a relatively compact action generates an enormous amount of physical information.

This naturally leads to the following question:

Can machine learning discover an equally compact—or even more fundamental—lossless representation directly from lattice ensembles?

If the learned representation resembles lattice QCD, that would be a remarkable validation. If it differs while remaining equally predictive, it may reveal a new way of understanding the theory.

The Challenge

Find the smallest lossless compression of one (or several) lattice ensembles.

“Lossless” means that every piece of physical information contained in the original ensemble can, in principle, be recovered from the compressed representation.

The compression method is completely open. It may involve machine learning, information theory, generative models, symbolic representations, or entirely new ideas.

Prize

🥇 US$100 will be awarded to the submission achieving the most efficient lossless compression.

Submission deadline: July 2029

If multiple submissions achieve comparable performance, the winner will be selected based on the elegance, reproducibility, and scientific insight of the approach.

Philosophy

This challenge is not about building a better compressor.

It is about exploring a broader hypothesis:

Perhaps discovering the laws of physics is fundamentally a problem of finding the optimal compression of reality.