LLMs Can Make Scientific Breakthroughs, Researcher Challenges Zahavy Position
A blog post disputes the claim that large language models lack the intuitive leaps needed for major scientific discoveries like Einstein's relativity.
Conversation activity · last 5 days peak 5/hr
Summary, timeline and people extracted by Claude from 10 items across 1 source · 8h ago. Quotes are verbatim.
An anonymous researcher (posting as coffeeaddict1) published a rebuttal to Tom Zahavy's ICML 2026 position paper "LLMs Can't Jump," which argues that LLMs cannot make abductive leaps required for paradigm-shifting scientific breakthroughs. The rebuttal contends that while Zahavy correctly identifies abduction as Einstein's path to General Relativity, alternative deductive paths—such as Richard Feynman's quantum field theory approach—could lead to the same discovery, suggesting LLMs might achieve breakthroughs through different reasoning routes than Zahavy assumes.
- Zahavy argues LLMs cannot make paradigm-shifting scientific discoveries because they lack sensory grounding and physical priors needed for abductive reasoning, using Einstein's General Relativity as proof.
- coffeeaddict1 counters that multiple deductive paths exist to the same scientific conclusions (e.g., Feynman's quantum field theory route to General Relativity), suggesting LLMs could achieve breakthroughs through reasoning routes different from Zahavy's assumed path.
- Community criticism emphasizes the paper is unsupported opinion and calls for rigorous quantitative methods to evaluate LLM scientific discovery capabilities.
- Zahavy clarifies his position paper is personal and does not claim LLMs can never make scientific discoveries, only that they may face specific limitations in abductive reasoning without sensory experience.
How it unfolded
-
Reaction Rebuttal posted to Hacker News
The coffeeaddict1 blog post appears on Hacker News with minimal initial engagement (1 score, 0 comments at posting time).
-
Event coffeeaddict1 publishes rebuttal blog
A blog post challenges Zahavy's thesis by arguing that General Relativity could be derived via Feynman's deductive route using quantum field theory principles, not just Einstein's abductive approach.
-
Report Zahavy clarification on Twitter/X
Tom Zahavy posts follow-up clarifications on social media, stating his paper is a personal position paper, not a company view, and does not claim LLMs can never make scientific discoveries.
-
Reaction Calls for rigorous quantitative evidence
Commenter killerstorm criticizes the position paper as unsupported opinion and proposes a methodology to rigorously test whether LLMs can derive recent scientific breakthroughs.
-
Reaction Historical accuracy questioned
Commenter quantum_mcts challenges the paper's historical framing of Einstein and Special Relativity, noting the groundwork in electrodynamics and Maxwell equations is oversimplified.
-
Reaction Early pushback on environmental factors
Commenter bob1029 argues the limitation may be environmental rather than inherent to LLMs, noting complex LLM interactions in game engines challenge the "no jumping" thesis.
-
Event Zahavy position paper gains traction
Tom Zahavy's ICML 2026 position paper titled "LLMs Can't Jump" is posted to OpenReview and gains significant discussion (297 HN score, 213 comments).
What people are saying verbatim
“I want to lodge my takes on the how LLMs can advance science through major breakthroughs, and what the assumptions behind my takes imply for safety risks and continual learning.”
coffeeaddict1 · Blog post on yongzx.github.io · Aug 7
“LLMs are capable of deduction or induction. Abduction––specifically relying on sensory experience––was the way of how Einstein came up with The Theory of General Relativity.”
coffeeaddict1 · Blog post on yongzx.github.io · Aug 7
“My position paper recently got some traction here, so I wanted to share a few thoughts and clarify a few things. First things first: some people are framing this as "DeepMind is throwing cold water on AI for science" or claiming the paper argues LLMs can never make real scientific discoveries. This is NOT the case.”
Tom Zahavy · Twitter/X (quoted in HN comment by defgeneric) · Aug 4
“on the basis of the general principles of quantum field theory and of experimental results it is possible to conclude that gravity, as any other force, has to be mediated by exchanges of a virtual particle, which in this case is a massless neutral spin-2 quantum, the graviton.”
Richard Feynman · Blog post on yongzx.github.io (quoted from Feynman article) · Aug 7
“I think this is more of a function of the harness and the environment than the LLM. I've seen some LLM interactions over complex environments like Godot and Unity that challenges the notion that there is no "jumping" going on at all.”
bob1029 · Hacker News · Aug 4
“This is literally an opinion of one dude which is not backed by any kind of quantitative evidence.”
killerstorm · Hacker News · Aug 4
Voices from the web unedited
-
Worth reposting a follow-up tweet from the author Tom Zahavy [1] after this made the rounds on X/Twitter recently:> A few reflections on my "LLMs Can’t Jump" paper:> My position paper recently got some traction here, so I wanted to share a few thoughts and clarify a few things.> First things first: some people are framing this as "DeepMind is…
-
Probably too late for this, but I have argued before that language is a fundamentally lossy encoding of the human experience. We do our best to describe what we're seeing and experiencing using language, which is fantastically expressive, but it has its limits. I think we see glimpses of this when we find ourselves saying things such as, "it's…
-
The theory is that creative leaps in theoretical physics require a grounding in sensory experience, but the obvious counter-argument is that humans can make creative leaps in abstract fields without such sensory grounding. They do address this at the end, saying"In abstract domains such as Mathematics or Computer Science, the Sense Experience (E)…
-
This is literally an opinion of one dude which is not backed by any kind of quantitative evidence.It's actually possible to answer this question rigorously:1. Define a scientific result which qualifies as a "jump". They should be frequent enough that they happen every year - otherwise one might say humans can't jump either.2. Identify all such…
-
The popular retelling of how Einstein created Special Relativity to "Resolve the contradictions of Michelson-Morly experiments" is very reductive to the history of the question. The epitome is the quote from the paper:> From the two postulates, Einstein derived the Lorentz trans- formation ...If Einstein derived them, who is "Lorentz"?The…
-
I think this is more of a function of the harness and the environment than the LLM. I've seen some LLM interactions over complex environments like Godot and Unity that challenges the notion that there is no "jumping" going on at all.An LLM in isolation from its environment might as well be a brain in a vat in some dark cave. You need an external…
-
Came for: "A computer once beat me at chess, but it was no match for me at kick boxing."TFA was actually about leaps of intuition, sadly.One of the experiments I've heard proposed around here is to somehow create an LLM from all text up to 1980 or 1990 and see if it can get back to making itself.
-
Why everybody is obsessed with replacing humans with LLMs when it seems like the most profitable use cases (like coding agents) rely on enhancing human capabilities?Until LLMs have some 0% error humans will have to be in the loop (even if they only serve to take responsibility of the process).