Gary Marcus: Riemann breakthrough vindicates neurosymbolic AI, not pure LLMs
GaryMarcus · x · 2026-11-21
Gary Marcus weighs in on the AI-assisted Riemann hypothesis progress (Alpöge et al.):
- Impressive if it holds up, but he stresses it has scarcely been diligenced and is not yet confirmed.
- He argues it's a vindication, not a challenge: pure LLMs still can't do this kind of mathematics — what worked were neurosymbolic systems descended from DeepMind's AlphaProof/AlphaGeometry line, exactly as he argued in his July 2024 essay on why the future of AI is neurosymbolic.
- Even if confirmed, he says, it wouldn't much change the problematic economics of the pure-LLM approach.
Bottom line: anyone serious about AI and math should reread his essay; the Riemann result confirms what he's argued for 25 years.
Related event: Gary Marcus: Riemann Progress Would Vindicate Neurosymbolic AI(2 posts)→
More from AGI Musings
- Benjamin Bratton: Solving good writing is harder than frontier math, yet humanities call AI bullshit — bratton · 2026-11-21
- Dahlia Ohara: even very smart engineering minds struggle to grasp the fundamental nature of AI — repligate · 2026-11-21
- AI Posts Math Results on Navier–Stokes and Riemann Hypothesis — While We Still Call LLMs Parrots — haider1 · 2026-11-21
- When Intelligence Becomes Abundant, Power Becomes Scarce — peizNLP · 2026-11-21
- AI-powered review harness finds fundamental flaws in claimed GRH proof — soumitrashukla9 · 2026-11-21
- Yudkowsky's two-year-old post on AI proving the Riemann Hypothesis resurfaces — nwilliams030 · 2026-11-21