"The model didn't figure it out": mathematician pushes back on LLM math autonomy claims
just_steve_h · x · 2026-10-09
In a debate over OpenAI's math release, some argued the model "figured out" proofs on its own. juststeveh counters that nothing was self-discovered: billions of text fragments are fed through algorithms to produce large probability matrices, whose outputs—constrained by inputs and other algorithms—are just text strings.