Dietterich pushes back on LeCun's claim that LLM-written papers won't stand the test of time
tdietterich · x · 2026-09-23
Yale professor Yann LeCun speculated that most papers written with LLMs will not stand the test of time, partly because people will develop a bias against that writing style.
Prominent ML researcher Thomas Dietterich pushed back: is the concern only about writing? What about theorems proved by LLMs, or algorithm experiments conducted by LLMs? If the substantive research output is sound, dismissing entire papers on stylistic grounds doesn't hold — a representative debate on LLMs' role in science.
Related event: Scholars debate LLM-written papers as ML papers grow homogeneous(4 posts)→
More from AGI Musings
- Schmidhuber rebuts Musk's AI timeline: no robot can match a plumber, let alone beat science by 2028 — AjdDavison · 2026-09-23
- AI2 CEO Etzioni backs Gary Marcus as the reality check on 'superintelligence' hype — GaryMarcus · 2026-09-23
- Richard Socher's 'The Eureka Machine' on full-stack scientific superintelligence is out — RichardSocher · 2026-09-23
- Karpathy: prompting is fading away, graphs are the endgame architecture — pranavmarla · 2026-09-23
- Framing automated science as a hypothesis-experiment-refute loop, with agent swarms competing — burny_tech · 2026-09-23
- Elon Musk declares intelligence is improving exponentially — TinfoilTricorn · 2026-09-23