OpenAI's Navier-Stokes Claim Draws Fire from Mathematicians and Physicists
Two weeks after OpenAI claimed to have solved a Clay Mathematics Institute problem related to Navier-Stokes, the controversy keeps growing, with academics raising challenges along three dimensions: mathematics, physics, and training data. One widely accepted point is that by Clay's judging criteria the result may count as "solved," but criticism continues over OpenAI's exaggerated official tweets, the proof's actual value, and its data sources.
Confirmed
- Multiple mathematicians (as compiled and relayed by @ProfNoahGian) believe the Navier-Stokes problem has multiple layers and OpenAI solved only part of it, not all, and that its official tweets overstate the result.
- A mechanical engineering professor in thermo-fluids (as relayed by @GaryMarcus) said bluntly that the solution is physically irreproducible—"we learn nothing about the nature of turbulence" from it—it's merely "a mathematical curiosity."
- A Scientific American report (as relayed by @joshgans) says the proof was generated by an internal OpenAI LLM and may have gamed a loophole in the problem statement, effectively solving the wrong problem—something that could have won the Clay Mathematics Institute's $1 million prize.
- Mathematician Sandra Stojanovic noted that the mathematical strategy used is explicitly permitted under the Clay Millennium Prize judging criteria.
Not yet confirmed
- Whether training relied on unpublished work remains unclear even after Gary Marcus consulted multiple mathematicians.
Why it matters
- Gary Marcus breaks the controversy into three layers: whether it counts as solved under Clay's criteria, whether training relied on unpublished work, and whether the claims are overstated. After debating Sabine Hossenfelder, he further stressed that solving Clay's problem does not mean solving the problems physicists care about.
- The episode touches on the credibility of AI research claims: even if the math meets the bar, missing physical meaning and exaggerated official messaging can mislead public understanding of AI capabilities.
2026-09-22 ~ 2026-09-22 · 8 related posts
- Episode 1: Debate Erupts Over Agent Swarm's Role in Navier-Stokes Proof(2026-09-18, 12 posts)
- Episode 2: Noam Clarifies Multi-Agent Systems Contributed Under 10% to Millennium Problem(2026-09-19, 3 posts)
- Episode 3: AI Solves Navier-Stokes with 130B Tokens, Sparking Debate(2026-09-20, 2 posts)
- Episode 4: Report: OpenAI Used ~10,000 Agents to Tackle Navier–Stokes(2026-09-20, 2 posts)
- Episode 5: OpenAI's Navier-Stokes Claim Draws Widespread Backlash from Mathematicians(2026-09-20, 16 posts)
- Episode 6: OpenAI's Navier-Stokes Claim Draws Fire from Mathematicians and Physicists(2026-09-22, 8 posts)
Primary sources
- "Solving Navier-Stokes took intelligence": debate over defining AI intelligence — austinc3301 · 2026-09-22
- [source] Did OpenAI Solve the Wrong Navier-Stokes Problem? Experts Cry Loophole — joshgans · 2026-09-22
- Gary Marcus says OpenAI's Navier-Stokes claim is overstated as mathematicians debate the proof — GaryMarcus · 2026-09-22
- [source] Mathematicians weigh in on OpenAI's Navier-Stokes claim: only parts solved — ProfNoahGian · 2026-09-22
- Thermofluids professor: OpenAI's Clay problem solution is not physically reproducible — GaryMarcus · 2026-09-22
- Gary Marcus: OpenAI solved the physics problem as posed, but not what physicists actually need — GaryMarcus · 2026-09-22
- [source] OpenAI's 10,000 agents proved a Navier-Stokes variant in 88 hours — sparking plagiarism claims — DeepLearningAI · 2026-09-22
1 near-duplicate retellings: GaryMarcus