Debate erupts: we understand LLMs far better than the human brain, researcher argues

AlexTensor · x · 2026-09-27

A debate flared on X over whether LLMs truly reason. Researcher Aran Nayebi pushed back on GradyBooch's dismissal of LLMs as "statistical engines walking a static manifold," noting that human brains are also non-deterministic statistical engines navigating the environment—yet we don't deny humans reason.

@andrewprock countered that claiming we understand the brain anywhere near as well as LLMs is absurd: "we can build an LLM, but we can't build a brain," calling the professor-level sloppiness unacceptable.

The core tension: statistical non-determinism isn't proof of no reasoning, and buildability shouldn't be the sole yardstick of understanding.

Original post →

More from AGI Musings

AGI Musings channel →