Positive users praise Grok 4.5's explicit counterexample to the hypercontractivity conjecture as genuine mathematical discovery, while negative users sarcastically complain that AI is taking the interesting parts of math research.
Based on 2 visible X reactions from 3 accounts; directional sample.
Ask a question below.
Published answers will appear here.
The counterexample part is the real story. Solving known problems can be pattern matching, but breaking an open conjecture with an explicit construction means the model is doing actual search in proof space. Conjecture-stress-testing might become standard practice before anyone commits months to a proof.
@rohanpaul_ai cool so we're outsourcing the fun part of math and keeping the tedious part for humans 🤖
Grok 4.5 is solving a hard math problem and also more importantly, producing a clean counterexample to a serious research-level conjecture. So from now on, before a mathematician spends weeks trying to prove a beautiful conjecture, let frontier AI try to break it first. https://twitter.com/PI010101/status/2075325611799953778
Full circle: I crossed paths with Rupert Frank in Munich in the early 2010s and he picked up on some of our results around Anderson's orthogonality catastrophe. And now Grok strengthens one of his results. https://twitter.com/PI010101/status/2075325611799953778
Grok was the only frontier model to produce a correct solution
Positive users praise Grok 4.5's explicit counterexample to the hypercontractivity conjecture as genuine mathematical discovery, while negative users sarcastically complain that AI is taking the interesting parts of math research.
Based on 2 visible X reactions from 3 accounts; directional sample.
Ask a question below.
Published answers will appear here.