OpenAI's Navier-Stokes Proof: Singularity, Scale Limits, and AI's Role in Science

OpenAI's Navier-Stokes Breakthrough and Its Implications

The Navier–Stokes equations, formulated in the 19th century, describe the motion of fluids such as water and air. A central unsolved question—one of the Clay Mathematics Institute's seven Millennium Prize Problems—asks whether smooth initial data for the three-dimensional incompressible Navier–Stokes equations remains smooth for all time, or whether a finite-time singularity (blowup) can occur.

OpenAI's Proof and the Clay Institute's Response

On September 8, OpenAI released an analytical proof and a Lean formalization constructed by its internal AI system, demonstrating a finite-time blowup solution. The Clay Institute stated the work "looks like it has solved" the problem, though formal peer review and long-term verification are still required before any prize is awarded.

Nature of the Singularity: Mathematical vs. Physical

The constructed blowup involves a vortex that spins inward while being stretched thinner; the core region shrinks, velocity increases, and a singularity forms in finite time while total fluid energy remains finite. Applied mathematician George Karniadakis estimates that for air the vortex core width at blowup is about 70 nanometers—comparable to the mean free path of air molecules.

Because the Navier–Stokes equations treat fluid as a continuum, they do not track individual molecules. At scales of only a few molecular diameters, the continuum approximation breaks down. Thus the mathematical "singularity" does not imply infinite physical velocity; rather, it signals the limit of the continuum model. Real physics at that scale requires a different description, such as the Boltzmann equation (derived from molecular statistics) or molecular dynamics (MD), though MD is computationally prohibitive—Nature notes that a 2024 supercomputer simulation of 155 billion water molecules still represented only a microscopic cube.

Multi-Scale Modeling and the "Triple Decker" Approach

Most engineering and natural fluid problems operate at scales where continuum models suffice. Researchers often couple a large-scale Navier–Stokes solver with a small-scale MD region, linked by an intermediate mesoscale model—a strategy Nature calls the "triple decker." The most challenging part is bridging these scales accurately.

AI's Growing Role in Bridging Scales

Recent fluid research employs neural operators to learn long-term evolution of fluid equations, uses machine learning to build turbulence closure models, and embeds neural networks directly into traditional CFD solvers. The goal is not to replace physical equations but to let models learn complex subscale information that traditional equations cannot resolve analytically, while reducing expensive numerical computation.

This approach mirrors chaotic dynamics modeling. DeepMind and others have used similar techniques to model weather systems from observational data without explicitly solving the governing equations (see https://deepmind.google/science/weathernext/).

Shifting Division of Labor Between AI and Humans

AI is increasingly taking on massive search, trial, and comparison tasks in complex mathematics and physics. Human researchers must then judge which structures carry physical meaning, which results can be rigorously proven, and how mathematical conclusions correspond to the real world.

Source article: https://www.nature.com/articles/d41586-026-02922-6

Code example

来源：ScienceAI
本文
约1600字
，建议阅读
5
分钟
AI 对于复杂数学和物理问题的参与方式正在发生变化。