Gary Marcus

Rank 14 of 47

Score 83

@matthew_pines um, just no. “math effectively solved” because of 19% score on a (difficult) benchmark?

certainly not what Tao said in his recent paper.

11/25/2025, 10:07:46 PM

In reply to:

Matthew Pines

@matthew_pines

“The capability ceiling hasn't just been raised; it has been shattered. Gemini 3 Pro hit a record 19% on the FrontierMath Tier 4 benchmark, a score that confirms math has been effectively solved by AI, subject only to further scaling of compute.”

Dr. Alex Wissner-Gross

@alexwg

https://x.com/i/article/19926098…

The statement is a critical response to a claim about AI's capabilities in solving mathematical problems. It challenges the assertion that a 19% score on a benchmark indicates that math has been effectively solved by AI, referencing a paper by Tao to support the counterargument.

Principle 4:
I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
The statement engages in constructive criticism by questioning the validity of the claim made by @matthew_pines. It does not resort to personal attacks or ad hominem arguments, instead focusing on the content of the claim and providing a counterpoint based on a recent paper by Tao. [+1]