Gary Marcus vs. Bo Wang: Can LLMs Actually Solve Biology?

BoWang87 · x · 2026-11-23

Gary Marcus argues an inductive claim that current systems can solve biology is flawed: math is quickly verifiable, biology needs lab work, and pure LLMs still can't do math on their own—only neurosymbolic systems combining LLMs with symbolic tools can, so capability claims must distinguish pure LLMs from LLM-based pipelines.

MIT professor Bo Wang quote-posts a sarcastic rebuttal: "Ok they can solve the Riemann Hypothesis but they cannot do biology…", poking at Marcus's habit of moving the goalposts as models advance. A representative spat over AI capability boundaries and verification paradigms.

Original post →

More from AGI Musings

AGI Musings channel →