The Specialty News
AI

GPT-6 Astra Helps Propose a Fix for a 25-Year Standard Model Gap

UC Santa Barbara physicist Nathaniel Craig says the AI took the lead on the proof behind a candidate lattice definition of the full Standard Model.

By The Specialty News DeskEdited by 4 min read
GPT-6 Astra Helps Propose a Fix for a 25-Year Standard Model Gap
Photo: UC Santa Barbara

The Standard Model predicts the magnetism of an electron to within a few parts in a trillion, yet in more than fifty years nobody has written down a complete, rigorous definition of it. On November 15, UC Santa Barbara physicist Nathaniel Craig described a candidate one on OpenAI's science blog, built on a grid of spacetime, and said GPT-6 Astra took the lead in developing the detailed proof. If it holds up, the theory behind every known particle gets a firmer mathematical floor, and an AI model gets real credit on one of physics' long-standing open problems.

Right Across 18 Orders of Magnitude, Still Undefined

The Standard Model describes the known elementary particles and three of the four fundamental forces. Its predictions match experiments across eighteen orders of magnitude, from the faint 21-centimeter radio emission of hydrogen to collisions at the Large Hadron Collider.

Most of those predictions come from perturbation theory. Physicists switch the interactions off, solve the easy problem exactly, then add the interactions back as a series of small corrections. It works beautifully for the electron's magnetism. It is little help for the mass of the proton, which comes mostly from the motion of its quarks and the gluons of the strong force. Switch the strong force off and the proton falls apart.

The workaround is the lattice: chop spacetime into a grid of points, define the theory exactly on that grid and let a computer calculate. Lattice methods have given physicists a first-principles account of the proton's mass. They have never covered the whole Standard Model, and the reason comes down to handedness.

The Left-Handed Glove Problem

Craig explains it with gloves in a mirror. Hold up a left-handed glove and the mirror shows a right-handed one. The strong force and electromagnetism treat left- and right-handed particles the same way. The weak force does not: it treats left- and right-handed quarks and electrons differently, which makes the Standard Model a chiral theory. Put a left-handed particle on a lattice with the most natural rules and a right-handed partner you never asked for shows up, a problem physicists call doubling.

Decades of work chipped away at it. Paul Ginsparg and Nobel laureate Ken Wilson wrote down the rules a lattice needs; Rajamani Narayanan and Herbert Neuberger found particles that obey them, called overlap fermions; Martin Lüscher showed how to tell left from right among them; Yoshio Kikukawa and colleagues explored how to use this for the weak force. One piece remained: a phase, a bookkeeping angle that has to come out consistent when the force fields on the grid are moved around a loop and brought back to where they started.

Lüscher worked out that phase rule for simpler theories, such as a chiral version of electromagnetism. Craig had a strategy for the full Standard Model. What he could not do was prove it worked in every case it needed to cover.

Where Astra Took the Lead

Craig's idea was to copy the force fields onto a succession of ever finer grids, compare the allowed particle states on each pair of neighbouring grids, and add those comparisons up to get the phase rule on the original one. Turning that into mathematics took several rounds of work with GPT-6 Astra. The model suggested the right way to carry the fields onto finer grids and showed that the comparisons converge to a definite answer.

“Astra helped turn that strategy into a construction and took the lead in developing the detailed proof.”— Nathaniel Craig

People checked the work too. Lattice theorists Maarten Golterman, Andreas Kronfeld and Yigal Shamir went over Astra's proof and asked the probing questions that built Craig's confidence. The October preprint runs to more than 120 pages with its supplement, and comes with a mathematical companion and a formalization in Lean, a language computers use to check proofs line by line.

Craig calls the result "much more a beginning than an end". It is a definition in principle, valid under a bound on field strengths at the scale of the grid, and much work remains before it is useful for realistic simulations. He is now checking a property called reflection positivity, which ties the lattice model to a quantum theory in which probabilities add up. The step from the grid to continuous spacetime is still to come, and Craig points to OpenAI's recent proof that Yang-Mills theory exists, one half of a Millennium Prize problem, as the piece that could connect them. The Standard Model has passed every test for half a century; now it has a candidate definition for experts to put to the test.

How the Phase Rule Gets Built

A visual summary of this story

The Brief

Stay curious

AI and technology: what changes and why it matters.
Your daily selection, in English or Spanish.

Free forever. Unsubscribe anytime.

Conversation

Start the conversation

No account needed. Comments are checked automatically — keep it civil.

More stories

Keep reading