'Theory Is All You Need': The Paper Arguing Prediction Underrates the Machines

maier_ak · x · 2026-09-21

The author shares a commentary on the paper 'Theory Is All You Need', calling it 'a strong case against prediction that underrates the machines.' The paper argues that next-token LLMs cannot run the theory-experiment loop and that theory is a uniquely human advantage (see the author's follow-up rebuttal: active learning, RL, and causal models already design experiments and create new data).

Original post →

More from AGI Musings

AGI Musings channel →