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 groundwork for Special Relativity was the study of electrodynamics and symmetries of Maxwell equations. The Einsteins paper was literally called "On the Electrodynamics of Moving Bodies" and never cites Michelson and Morley.
Einstein seems to be so conveniently dismissive that he knew about the seminal Michelson-Morley experiments but he probably knew about it too well [1].
Einstein is not the first great scientist who are in denial of other important prior contributions, and he also not the last one. Newton also probably knew too well about Al-Haytham (Alhazen), arguably the father of modern science, and his breakthrough experiments but never directly cited Alhazen's works in his seminal books on Optics.
[1] Millikan, Einstein, and the Birth of Relativity (4 letters):
"Abraham Pais, who knew Einstein well and wrote his scientific biography, was certain that Einstein did know about Michelson's experiment before 1905. He points out that Einstein was over seventy and in poor health when he spoke to Shankland; at the first interview he probably did not remember that Michelson's experiment is discussed in Lorentz's 1895 monograph, the famous "Versuch", which Einstein had definitely read before 1905."
Was the ethical requirement to discuss and acknowledge predecessors work as developed in Newton's time as in our own? I would naturally expect not because it would seem to me to be the kind of thing that develops over time, but I could be wrong as I am not a historian of science.
It's a not a requirement, it's just a natural to do as a research scientist to acknowledge prior contributions if you knew it. The same is true when you are at the time of Aristotle.
This isn't offered as a defense of Einstein since I couldn't speculate on what ideas he must have been consciously aware of, but several times in my life I've thought I had a novel idea only to find a very similar idea was in something I read and had just forgotten about. It's a very human phenomenon.
Science rightfully recognizes the mind that doesn't just first describe the idea but provides a robust framework to test it and communicate it.
I get really tired of Einstein being portrayed as if he’s the greatest scientist who ever lived, when he doesn’t even belong in that conversation. If Einstein never existed, the fields he was in were already moving strongly in the directions of his conclusions. If Maxwell or Newton had never existed, the world today would be quite different.
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 throwing cold water on AI for science" or claiming the paper argues LLMs can never make real scientific discoveries. This is NOT the case.
> This is a personal position paper, not the company's view on AI for science. This is also not my position. As a core contributor to AlphaProof (the first AI system to win an IMO medal), I know firsthand that my colleagues at DeepMind, other frontier labs, and academia have made amazing discoveries with LLMs and will continue to do so. This paper is NOT an "LLMs are a dead end" kind of thing.
> Rather, the paper is the result of a deep dive I took to study the invention of General Relativity. I wanted to explore what it would take for a modern AI system to make that exact kind of jump. Specifically, I focused on the equivalence principle—a key axiom that Einstein formulated through thought experiments grounded in his physical intuition. I was trying to figure out what it would take to give modern AI systems that sort of thinking.
> Giving AI this specific capability isn't necessarily the most urgent thing to do next. It is very likely that improving our current recipes will lead to many exciting discoveries in the near future. In fact, that is what I am personally working on these days (sorry to disappoint you!). It is also quite possible that I am wrong, and that simply scaling our current systems will lead to new inventions in physics and elsewhere.
> Nevertheless, this was my position last winter when I wrote the paper, and I'm sticking to it. I think that there are a few interesting ideas to explore in this space which could influence the next generation of AI systems. I was very lucky to receive a lot of interesting feedback about this position—thank you for all the messages!
>> Specifically, I focused on the equivalence principle—a key axiom that Einstein formulated through thought experiments grounded in his physical intuition.
It's weird because the equivalence principle is very unintuitive. Aristotle's Mechanics does not have it. It took almost two thousand years to discover inertia that is the most simple version of the equivalence principle. Einstein understood the idea of the the equivalence principle because he had a physics degree, not because he feel that in real life.
Moreover, if you ever have to study or teach Quantum Mechanics, physical intuition gets in the way. A lot of properties contradict the physical intuition but after a while you get use to them. If we continue with Einstein, the photoelectric effect does not aperar in real life.
That’s funny, when I first learned about the equivalence principle, my first thought was “of course!” I have always found it to be very intuitive. The great leap is being able to frame it that way.
I mean Einstein had help, he was networked with the best scientific minds of the planet and his discoveries were grounded in experimental results that contradicted existing theories (at least for specialized relativity), and without Riemann’s work he wouldn’t have been able to formulate his theory either. So not sure if AI couldn’t do that if you kept feeding it with new research results and let it correspond with top human scientists. Einstein was a genius but I don’t think his thought process is beyond what an LLM could simulate. And again this is probably the most impressive scientific achievement in theoretical physics in the 20. century so maybe it’s hanging the bar a bit high for LLMs.
If your glib comment is referring to me as a techbro and doing the annoying worst thing, then maybe you should explain how I am comfortably abusing the goalpost fallacy given the explosion of agentic AI capability.
I simply point out here, that fully accepting the paper’s premise, the paper’s conclusion isn’t limiting on frontier AI reasoning agents. The paper posits the necessity of multimodal world models and the limitations of LLMs. Frontier agents aren’t simply LLMs and do increasingly integrate increasingly capable multimodal models.
> Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
> Comments should get more thoughtful and substantive, not less, as a topic gets more divisive.
> When disagreeing, please reply to the argument instead of calling names. "That is idiotic; 1 + 1 is 2, not 3" can be shortened to "1 + 1 is 2, not 3."
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.
> 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.
Could be, but preventing leakage from more modern stuff can be challenging.
I can't find the citation right now, but I think people found it was leaking anachronisms? So this probably wasn't as well filtered as the creator had hoped?
Progress followed improvements in hardware, would you have to give access to modern hardware in the experiment for it to use? How much could it infer from it?
Lots of things changed, GPT-2 is small (1.5e9) and is also a base model, so it is only doing next-token/autocomplete rather than prompt-response like even the first ChatGPT-3.5 was doing.
Just for the sake of clarity: all LLMs up to today are still only doing next-token/autocomplete. The training process got additional stages to shape the model weights, but standalone LLMs are still deployed essentially identically.
If you gave GPT-2 a question and ended with a "?", it might answer, but also it might write several more questions in a similar category.
IMO, the mechanism isn't the important thing, the behaviour is. If you look at the step-by-step, we are also looking for the next word or motor action (and for whoever is about to suggest that we humans plan ahead, Transformer-based LLMs have been shown to also do this); as this is not a useful description of what it means to be a living brain, I'd say it's also not a useful description of what makes everything post-InstructGPT different from what came before.
I agree completely - behaviorally the models have changed drastically due to RLHF, RLVR and now maybe even more so due to agentic harnesses. But the mechanism of prediction hasn’t changed, that was all I was clarifying.
Why couldn't an LLM, if it was smart enough, generate and consume its own data?
I know the answer: because it leads to model collapse. But why is that? Wouldn't a smart model not collapse? It's seeming like they keep getting smarter because we keep pouring more of our own knowledge into them, not because they are actually getting smarter. And yes, sometimes a dumb but persistent bruteforcer can make new discoveries.
and current gen LLMs get a lot of training data beyond "all books + all of the internet"
the objection is then that producing this additional data would already confound it with pre "virtual cutoff date" knowledge (since the training data probably implies mathematical and SWE concepts that were developed post "virtual cutoff date")
The curious case here is how much of a description do we give it of itself? That would almost certainly dominate success rates.
My feeling is that a prompt would have to provide a vague description of a program that meaningfully passes something like a Turing test, an API to conform to, an expectation of novel construction (no 'ifs all the way down'), and then a requirement to search broadly and pursue promising ideas and not get hung up on the philosophy. Anything more precise feels like it would corrupt the test, but as it is that description feels doomed to loop before even trying the interesting parts.
I wonder if we could just tell it to invent itself without any description and see if it can I introspect enough through its own interface to figure out what it is.
Chessboxing was the invention of comics book artist Enki Bilal (and he's credited with this in Wikipedia). I first saw it in his Nikopol trilogy. Because life is weird, it then became a real thing.
It's unrelated to computers playing chess. It predates Kasparov's first defeat by Deep Blue. I don't remember any mention of computers being good at chess in the trilogy, either. Or any computers, for that matter.
The article seems like an interesting Gedankenexperiment. However, I think it overrotates on the GR analogy.
For example "..ARC captures the logical leap, it misses the manipulative component—the physical sensation and embodied simulation..." makes lots of assumptions on how such a discovery must occur, e.g. through "physical sensation and embodied simulation". Results matter, not the path there.
For example, quantization of energy, at the core of QM, wasn't discovered through "physical sensation and embodied simulation" at all. Planck simply found that if energy is quantized, then one obtained the observed black-body radiation spectrum. There was no "physical sensation and embodied simulation".
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).
> 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?
The only way to justify trillion dollar valuations is to sell cruelty-free robot slaves which replace white collar labor for pennies on the dollar.
Even though it's a complete and total sham, there's no American agency willing to or capable of prosecuting these firms for fraud, so they really have very strong incentives to keep up the lie.
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 environment to sample from and act upon to make forward progress.
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) may be grounded in high-dimensional topology or have other goals such as generality or minimality."
But if such sense experience is possible in abstract domains via some high-dimensional topology, why could a sufficiently advanced LLM not develop an equivalent high-dimensional topology for domains like physics and use it to make creative leaps?
> ...but the obvious counter-argument is that humans can make creative leaps in abstract fields without such sensory grounding...
But we have no idea at how good humans are at that. Given the appalling failures of humans to handle even basic statistical situations like identifying that the same thing happens over and over, it might be that they are hilariously bad at creative leaps in abstract fields, it is just we have had nothing better available to measure against. We've spent about as long as decision theory existed trying to convince people to use it instead of flailing. Limited success, usually in exceptional cases.
And the paper seems a bit dodgy, we have models created with sensory data available. No reason a LLM can't be trained on more sensory data than a human can accumulate in one lifetime. There is a lot of visual data on YouTube.
Most humans are probably bad at it. Some humans are very good at it. I'd even argue most science does not demand these kinds of leaps and is mostly concerned with incremental improvements, or proving or disproving other people's abductions
The case study they chose is literally Albert Einstein coming up with General Relativity, something most scientists of his time were not able to do
That's an interesting analogy. My gut sense is that theoretical mathematics requires a high level of intelligence versus more grounded domains. That may imply that deficiency in grounding can be made up for with intelligence and basically reverse engineering the gaps in grounding from first principles/limited grounding. The ultimate question would then be what is the tradeoffs between grounding and raw intelligence for the same outcome.
If there is enough cross over between real world knowledge engrams and abstract knowledge engram, would this allow for the jump?
One interesting (albeit sad) area which might be related are humans who are never raised with a first language. They seem to never developer abstract reasoning and even seem to lose the ability to develop it later in life. This might indicate there is some 'real world senses' -> 'direct language' -> 'indirect language' -> 'abstract abduction' hierarchy that develops, perhaps related to more real world abductions as a necessary side chain to developing abstract ones.
One of the obvious problems with this is just how difficult we find it to study intelligence purely in humans. We are measure a LLMs by a yardstick that is already known broken, but maybe this is still the right path.
I have been writing a 'paper' [1] on an adjacent topic for months now. At some point, I decided to make it an empirical paper vs position paper. I am still chasing the experiments (when I get some free time waiting for agentic loops)
For this paper specifically, after reading the abstract [2], I felt almost certain that the author would have used Judea Pearl's ladder of causation (https://web.cs.ucla.edu/~kaoru/3-layer-causal-hierarchy.pdf) but they did not. Would have probably been a better argument to make.
[1] paper in quotes because it may never get published (it is over 20 pages atm). the core argument is that lack of native adjacency resolution makes problems harder and sample inefficient, not impossible
[2] "Using Einstein’s formulation of General Relativity as a case study, we demonstrate that LLMs are structurally incapable of creating new foundational axioms, particularly when observational data is scarce. "
Also, the claim that 'LLMs are structurally incapable of creating new foundational axioms' is provably false depending on where you place 'fundamental'.
The early physics background is messy and incorrect. I didnt read the full position paper, but from its start: The Lorentz transformations were by Lorentz, well before Einstein’s paper on special relativity; the principle of relativity also existed before the Einstein paper. The math was all there, with steps taken by Maxwell, Voigt, Larmor, Lorentz, and Poincare. Einstein supplied a clean physical interpretation, making all inertial frames equivalent, making simultaneity frame dependent, and explaining length and time deformations without the need of the concept of ether. Skimming the end of the paper with the arguments about lack of abduction or inability to make the analogy without sensory experience, I see that this paper is unfounded speculation rather than solid/hard philosophical logic. As a position paper it is OK to appear, but i think it misses the point of how LLMs or other autoregressive learners of future states can build analogies and intuition that can help them formulate new theories of the world. Soon it will be more obvious to everyone, so I am not very worried about these writings.
Yes, I made a similar comment on the other mention today. The author seems to misunderstand what GR is / what it added to physics too (creating self-referential field equations to handle mass/energy equivalence - linear field equations without instantaneous 'action-at-a-distance' existed much earlier). The notion that it was a 'small signal' is totally false. Once you've hypothesized that the apparent mass of objects depends on the observer, you need to show that your theory gives consistent results for trajectories of objects in gravitational fields.
I found this paper really thought provoking, but I think the conclusion of “world models are the solution” leaves something to be desired. People are already equipping agentic systems with physical simulation tools and exploring action-conditioned world models. This is cool because you can change the rules of the simulation and observe what happens, but it doesn’t address the core question of what to change the rules to, or even what the goal should be in the first place.
It's a very shaky position, and the empirical track record of "LLMs can't..." is in itself a reason to call it into doubt.
Every "can't" of this nature was followed by a discovery of "they can, just poorly", and then by that "poorly" improving steadily generation to generation.
The paper doesn't provide a way to measure or quantify this elusive "jumping" capability, not even as an approximation. It just throws "can't jump" out there, as if "abduction" is an established class of problem with known computational properties and requirements that the LLM architecture fails to satisfy. It's none of those things - and the paper makes the claim without backing it by anything but rhetoric attempts at persuasion.
The proposed solution is also dubious. The empirical track record of dedicated "world models" for reasoning and problem-solving is, frankly, downright abysmal. Even integrating multimodal data into LLMs has failed to yield general reasoning capability gains.
LeCun's misadventures in the field aside, the main frontier lab that pushes in favor of "improving reasoning via multimodal fusion" is GDM - and Gemini isn't exactly a paragon of frontier reasoning capabilities. It has strong multimodal capabilities, but lags behind both OpenAI and Anthropic in performance outside that - while Anthropic is the lab that always treated multimodal grounding as an afterthought, and still trades blows with OpenAI at the very edge of the performance frontier. Multimodal grounding seems to work great as a way to improve an AI's ability to deal with those specific modalities, but it falters outside that.
Now, it's not impossible that everyone who tried multimodal world models for reasoning is just doing it wrong, and there is an undiscovered recipe for multimodal grounding that results in a step change in AI capabilities. But the results we have so far suggest it to be unlikely.
> the empirical track record of "LLMs can't..." is in itself a reason to call it into doubt ... Every "can't" of this nature was followed by a discovery of "they can, just poorly", and then etc
This tracks for me as someone making keeps of intuition in little-explored areas.
I just don't see any of the LLM users around at all. Clearly some force is guiding them all away from thinking any of the "leap of faith" thoughts that I am thinking.
I have a weird thought experiment: If you give a GPT-2/3 level LLM tools to search the internet - any document, can it build bigger, better LLMs?
You may think this is not a good test because an older (or say a smaller) LLM can study from the knowledge on the Internet and build. But we are like that - we can access the Universe through our senses.
Can we ever produce anything that is beyond this Universe? I think an LLM that is lacking in knowledge can build more complex systems as long as it can access more data.
I feel like you could just add some noise or randomness to the LLM and start approximating the leaps that the human mind uses to solve and understand unrelated things. Maybe that’s naive, it’s just coming from my organic computer in my skull.
Technically true, but the counter argument would be that the probability of this working would be ~ 2^(-(entropy_of_leap)) for an LLM (presumably intractable) and a human would succeed at a higher probability.
Hahaha you just derived temperature from first principles.
Turns out temperature is pretty bad too, you can find ways to sample from deeper in the distribution without distorting it. Great example is XTC (exclude top choices), In a few weeks/months it'll also have a proper scholarly paper with peer review.
LLMs can’t but humans supplied with data and reasoning from an LLM can make novel jumps without absolute prior knowledge, or at least with reduced need for years of knowledge.
And then such jump can be verified by a machine so human kind of plugs the intelligence gap.
That’s pretty exciting.
I always liked to provocatively call LLM „the new calculator”. Calculator for language.
We are so focused on creating a standalone intelligence that we didn’t notice how we massively augmented our own. That could be considered transhumanism holy grail if only interface brain-LLM was faster.
People need to understand that these things are tools. And every tool needs an operator to function. Tool doesn’t have its own goals, needs, wants or motives. It won’t do anything out of its own, it always exists in context of someone telling it what to do.
In light of that most of the panic and fear mongering is rather ridiculous. Calculator won’t replace you. It wont take over the world. It is just a tool.
You write a book with book generator? Cool, it can be used for this. We will judge output, not the methods. Sometimes we will judge people who have no taste in literature.
Calculator for language, is a nice insight. I visualize generally as LLM's as big mathematical expression that produce next word.
But then , one difference I found on LLM's is on scaling laws, where at some point, it have interesting emerging properties, that nobody thought would be possible now being possible.
Tools:
Though It is a tool, but powerfull tool that is automating existing manually done jobs, large scale. Adapting to new roles where we definite goals, needs and motives judging of AI output, at large level is an issue. And humans are used to day to day repeating job, doing same thing repeatedly. Now the AI is taking over that. I see that is a challenge.
Also I see we are moving to creative world, where we will spend more time on creating something really new, leaving mechanized parts to AI.
Clearly LLMs cant do leaps of intuition since their "intuition" is locked after training ends.
The only way a LLM can come up with new ideas if the "idea" appeared as a generalisation durring training or if it was achieved using reason in chain of thought.
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 groundwork for Special Relativity was the study of electrodynamics and symmetries of Maxwell equations. The Einsteins paper was literally called "On the Electrodynamics of Moving Bodies" and never cites Michelson and Morley.
Einstein seems to be so conveniently dismissive that he knew about the seminal Michelson-Morley experiments but he probably knew about it too well [1].
Einstein is not the first great scientist who are in denial of other important prior contributions, and he also not the last one. Newton also probably knew too well about Al-Haytham (Alhazen), arguably the father of modern science, and his breakthrough experiments but never directly cited Alhazen's works in his seminal books on Optics.
[1] Millikan, Einstein, and the Birth of Relativity (4 letters):
https://www.aps.org/archives/publications/apsnews/200403/let...
"Abraham Pais, who knew Einstein well and wrote his scientific biography, was certain that Einstein did know about Michelson's experiment before 1905. He points out that Einstein was over seventy and in poor health when he spoke to Shankland; at the first interview he probably did not remember that Michelson's experiment is discussed in Lorentz's 1895 monograph, the famous "Versuch", which Einstein had definitely read before 1905."
Was the ethical requirement to discuss and acknowledge predecessors work as developed in Newton's time as in our own? I would naturally expect not because it would seem to me to be the kind of thing that develops over time, but I could be wrong as I am not a historian of science.
It's a not a requirement, it's just a natural to do as a research scientist to acknowledge prior contributions if you knew it. The same is true when you are at the time of Aristotle.
This isn't offered as a defense of Einstein since I couldn't speculate on what ideas he must have been consciously aware of, but several times in my life I've thought I had a novel idea only to find a very similar idea was in something I read and had just forgotten about. It's a very human phenomenon.
Science rightfully recognizes the mind that doesn't just first describe the idea but provides a robust framework to test it and communicate it.
I agree with you in general. However:
> If Einstein derived them, who is "Lorentz"?
You can (re-) derive a lot of existing stuff.
> You can (re-) derive a lot of existing stuff.
Einstein was aware of Lorentz and the transform. He was aware of Poincaré as well. He knew the state of the art for his time.
Re-deriving the start of the art cleanly with fewer postulates is good!
I get really tired of Einstein being portrayed as if he’s the greatest scientist who ever lived, when he doesn’t even belong in that conversation. If Einstein never existed, the fields he was in were already moving strongly in the directions of his conclusions. If Maxwell or Newton had never existed, the world today would be quite different.
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 throwing cold water on AI for science" or claiming the paper argues LLMs can never make real scientific discoveries. This is NOT the case.
> This is a personal position paper, not the company's view on AI for science. This is also not my position. As a core contributor to AlphaProof (the first AI system to win an IMO medal), I know firsthand that my colleagues at DeepMind, other frontier labs, and academia have made amazing discoveries with LLMs and will continue to do so. This paper is NOT an "LLMs are a dead end" kind of thing.
> Rather, the paper is the result of a deep dive I took to study the invention of General Relativity. I wanted to explore what it would take for a modern AI system to make that exact kind of jump. Specifically, I focused on the equivalence principle—a key axiom that Einstein formulated through thought experiments grounded in his physical intuition. I was trying to figure out what it would take to give modern AI systems that sort of thinking.
> Giving AI this specific capability isn't necessarily the most urgent thing to do next. It is very likely that improving our current recipes will lead to many exciting discoveries in the near future. In fact, that is what I am personally working on these days (sorry to disappoint you!). It is also quite possible that I am wrong, and that simply scaling our current systems will lead to new inventions in physics and elsewhere.
> Nevertheless, this was my position last winter when I wrote the paper, and I'm sticking to it. I think that there are a few interesting ideas to explore in this space which could influence the next generation of AI systems. I was very lucky to receive a lot of interesting feedback about this position—thank you for all the messages!
[1] https://x.com/TZahavy/status/2082401499628376180
>> Specifically, I focused on the equivalence principle—a key axiom that Einstein formulated through thought experiments grounded in his physical intuition.
It's weird because the equivalence principle is very unintuitive. Aristotle's Mechanics does not have it. It took almost two thousand years to discover inertia that is the most simple version of the equivalence principle. Einstein understood the idea of the the equivalence principle because he had a physics degree, not because he feel that in real life.
Moreover, if you ever have to study or teach Quantum Mechanics, physical intuition gets in the way. A lot of properties contradict the physical intuition but after a while you get use to them. If we continue with Einstein, the photoelectric effect does not aperar in real life.
That’s funny, when I first learned about the equivalence principle, my first thought was “of course!” I have always found it to be very intuitive. The great leap is being able to frame it that way.
I mean Einstein had help, he was networked with the best scientific minds of the planet and his discoveries were grounded in experimental results that contradicted existing theories (at least for specialized relativity), and without Riemann’s work he wouldn’t have been able to formulate his theory either. So not sure if AI couldn’t do that if you kept feeding it with new research results and let it correspond with top human scientists. Einstein was a genius but I don’t think his thought process is beyond what an LLM could simulate. And again this is probably the most impressive scientific achievement in theoretical physics in the 20. century so maybe it’s hanging the bar a bit high for LLMs.
Moreover anyone glomming onto this paper for goal-post-shifting “AI can never” should:
1. Read the last sentence of the abstract, and
2. Reflect that frontier reasoning agents already increasingly integrate multimodal models.
The worst thing about the AI boom is how tech bros feel comfortable abusing the goalpost fallacy. So annoying.
If your glib comment is referring to me as a techbro and doing the annoying worst thing, then maybe you should explain how I am comfortably abusing the goalpost fallacy given the explosion of agentic AI capability.
I simply point out here, that fully accepting the paper’s premise, the paper’s conclusion isn’t limiting on frontier AI reasoning agents. The paper posits the necessity of multimodal world models and the limitations of LLMs. Frontier agents aren’t simply LLMs and do increasingly integrate increasingly capable multimodal models.
> Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
> Comments should get more thoughtful and substantive, not less, as a topic gets more divisive.
> When disagreeing, please reply to the argument instead of calling names. "That is idiotic; 1 + 1 is 2, not 3" can be shortened to "1 + 1 is 2, not 3."
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.
> 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.
Could be, but preventing leakage from more modern stuff can be challenging.
This was attempted with Victorian public domain content: https://www.estragon.news/mr-chatterbox-or-the-modern-promet...
I can't find the citation right now, but I think people found it was leaking anachronisms? So this probably wasn't as well filtered as the creator had hoped?
Progress followed improvements in hardware, would you have to give access to modern hardware in the experiment for it to use? How much could it infer from it?
> would you have to give access to modern hardware in the experiment for it to use?
At a minimum, yes. IIRC, the sum total of all compute manufactured over history only reached the minimum needed to train an OK LMM in the mid 00s.
> How much could it infer from it?
Only way to find out is to try.
I think this can't work because an LLM needs too much data, and before the internet there probably just wasn't enough to get close to what we have now
Even simpler: Can GPT-2 anticipate and build Gwen/Deepseek? I think the answer is almost trivially "no", so I wonder what changed?
Lots of things changed, GPT-2 is small (1.5e9) and is also a base model, so it is only doing next-token/autocomplete rather than prompt-response like even the first ChatGPT-3.5 was doing.
Just for the sake of clarity: all LLMs up to today are still only doing next-token/autocomplete. The training process got additional stages to shape the model weights, but standalone LLMs are still deployed essentially identically.
If you gave GPT-2 a question and ended with a "?", it might answer, but also it might write several more questions in a similar category.
IMO, the mechanism isn't the important thing, the behaviour is. If you look at the step-by-step, we are also looking for the next word or motor action (and for whoever is about to suggest that we humans plan ahead, Transformer-based LLMs have been shown to also do this); as this is not a useful description of what it means to be a living brain, I'd say it's also not a useful description of what makes everything post-InstructGPT different from what came before.
I agree completely - behaviorally the models have changed drastically due to RLHF, RLVR and now maybe even more so due to agentic harnesses. But the mechanism of prediction hasn’t changed, that was all I was clarifying.
Why couldn't an LLM, if it was smart enough, generate and consume its own data?
I know the answer: because it leads to model collapse. But why is that? Wouldn't a smart model not collapse? It's seeming like they keep getting smarter because we keep pouring more of our own knowledge into them, not because they are actually getting smarter. And yes, sometimes a dumb but persistent bruteforcer can make new discoveries.
> if it was smart enough
and i think this is exactly the crux;
the really big models need really big datasets
and current gen LLMs get a lot of training data beyond "all books + all of the internet"
the objection is then that producing this additional data would already confound it with pre "virtual cutoff date" knowledge (since the training data probably implies mathematical and SWE concepts that were developed post "virtual cutoff date")
If it is smart enough to generate data it can consume to train itself better, it is already smart enough to not need to do that.
Maybe we can synthesize large amounts of limited information. I thought that new training data is mostly synthetic anyway.
The curious case here is how much of a description do we give it of itself? That would almost certainly dominate success rates.
My feeling is that a prompt would have to provide a vague description of a program that meaningfully passes something like a Turing test, an API to conform to, an expectation of novel construction (no 'ifs all the way down'), and then a requirement to search broadly and pursue promising ideas and not get hung up on the philosophy. Anything more precise feels like it would corrupt the test, but as it is that description feels doomed to loop before even trying the interesting parts.
I wonder if we could just tell it to invent itself without any description and see if it can I introspect enough through its own interface to figure out what it is.
Is that how chessboxing was invented? Genuinely asking.
Nope.
Chessboxing was the invention of comics book artist Enki Bilal (and he's credited with this in Wikipedia). I first saw it in his Nikopol trilogy. Because life is weird, it then became a real thing.
It's unrelated to computers playing chess. It predates Kasparov's first defeat by Deep Blue. I don't remember any mention of computers being good at chess in the trilogy, either. Or any computers, for that matter.
Article was plenty interesting to me.
The article seems like an interesting Gedankenexperiment. However, I think it overrotates on the GR analogy.
For example "..ARC captures the logical leap, it misses the manipulative component—the physical sensation and embodied simulation..." makes lots of assumptions on how such a discovery must occur, e.g. through "physical sensation and embodied simulation". Results matter, not the path there.
For example, quantization of energy, at the core of QM, wasn't discovered through "physical sensation and embodied simulation" at all. Planck simply found that if energy is quantized, then one obtained the observed black-body radiation spectrum. There was no "physical sensation and embodied simulation".
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).
> 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?
The only way to justify trillion dollar valuations is to sell cruelty-free robot slaves which replace white collar labor for pennies on the dollar.
Even though it's a complete and total sham, there's no American agency willing to or capable of prosecuting these firms for fraud, so they really have very strong incentives to keep up the lie.
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 environment to sample from and act upon to make forward progress.
Why does the cave need to be dark if its just a brain in a vat?
'Cause the brain is in a glass vat. and brains have eyes too. https://www.youtube.com/watch?v=pc3uWxs9adw
Is this a reference to the movie "White man can't jump"? If so they are in a surprise because the movie says otherwise.
Best comment on this from 6 months ago: https://news.ycombinator.com/item?id=46870575
"Best" because of the irony that it was written by Claude? https://news.ycombinator.com/item?id=46901199
Even myself, I really can't remember a time where I had this "jump". Is very subjective to feel this jump
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) may be grounded in high-dimensional topology or have other goals such as generality or minimality."
But if such sense experience is possible in abstract domains via some high-dimensional topology, why could a sufficiently advanced LLM not develop an equivalent high-dimensional topology for domains like physics and use it to make creative leaps?
> ...but the obvious counter-argument is that humans can make creative leaps in abstract fields without such sensory grounding...
But we have no idea at how good humans are at that. Given the appalling failures of humans to handle even basic statistical situations like identifying that the same thing happens over and over, it might be that they are hilariously bad at creative leaps in abstract fields, it is just we have had nothing better available to measure against. We've spent about as long as decision theory existed trying to convince people to use it instead of flailing. Limited success, usually in exceptional cases.
And the paper seems a bit dodgy, we have models created with sensory data available. No reason a LLM can't be trained on more sensory data than a human can accumulate in one lifetime. There is a lot of visual data on YouTube.
Most humans are probably bad at it. Some humans are very good at it. I'd even argue most science does not demand these kinds of leaps and is mostly concerned with incremental improvements, or proving or disproving other people's abductions
The case study they chose is literally Albert Einstein coming up with General Relativity, something most scientists of his time were not able to do
That's an interesting analogy. My gut sense is that theoretical mathematics requires a high level of intelligence versus more grounded domains. That may imply that deficiency in grounding can be made up for with intelligence and basically reverse engineering the gaps in grounding from first principles/limited grounding. The ultimate question would then be what is the tradeoffs between grounding and raw intelligence for the same outcome.
Isnt the sensory grounding even in abstract cases some (limited) intuition that simulates in a mental world model?
If there is enough cross over between real world knowledge engrams and abstract knowledge engram, would this allow for the jump?
One interesting (albeit sad) area which might be related are humans who are never raised with a first language. They seem to never developer abstract reasoning and even seem to lose the ability to develop it later in life. This might indicate there is some 'real world senses' -> 'direct language' -> 'indirect language' -> 'abstract abduction' hierarchy that develops, perhaps related to more real world abductions as a necessary side chain to developing abstract ones.
One of the obvious problems with this is just how difficult we find it to study intelligence purely in humans. We are measure a LLMs by a yardstick that is already known broken, but maybe this is still the right path.
Previous: https://news.ycombinator.com/item?id=49162791
Also:
https://news.ycombinator.com/item?id=49136070 https://news.ycombinator.com/item?id=49096837 https://news.ycombinator.com/item?id=46890333 https://news.ycombinator.com/item?id=46870562
All of these are titled “LLMs Can’t Jump”
Also https://news.ycombinator.com/item?id=49136070
You posted this twice now. Why not just edit your original comment?
I have been writing a 'paper' [1] on an adjacent topic for months now. At some point, I decided to make it an empirical paper vs position paper. I am still chasing the experiments (when I get some free time waiting for agentic loops)
For this paper specifically, after reading the abstract [2], I felt almost certain that the author would have used Judea Pearl's ladder of causation (https://web.cs.ucla.edu/~kaoru/3-layer-causal-hierarchy.pdf) but they did not. Would have probably been a better argument to make.
[1] paper in quotes because it may never get published (it is over 20 pages atm). the core argument is that lack of native adjacency resolution makes problems harder and sample inefficient, not impossible
[2] "Using Einstein’s formulation of General Relativity as a case study, we demonstrate that LLMs are structurally incapable of creating new foundational axioms, particularly when observational data is scarce. "
Also, the claim that 'LLMs are structurally incapable of creating new foundational axioms' is provably false depending on where you place 'fundamental'.
The early physics background is messy and incorrect. I didnt read the full position paper, but from its start: The Lorentz transformations were by Lorentz, well before Einstein’s paper on special relativity; the principle of relativity also existed before the Einstein paper. The math was all there, with steps taken by Maxwell, Voigt, Larmor, Lorentz, and Poincare. Einstein supplied a clean physical interpretation, making all inertial frames equivalent, making simultaneity frame dependent, and explaining length and time deformations without the need of the concept of ether. Skimming the end of the paper with the arguments about lack of abduction or inability to make the analogy without sensory experience, I see that this paper is unfounded speculation rather than solid/hard philosophical logic. As a position paper it is OK to appear, but i think it misses the point of how LLMs or other autoregressive learners of future states can build analogies and intuition that can help them formulate new theories of the world. Soon it will be more obvious to everyone, so I am not very worried about these writings.
Yes, I made a similar comment on the other mention today. The author seems to misunderstand what GR is / what it added to physics too (creating self-referential field equations to handle mass/energy equivalence - linear field equations without instantaneous 'action-at-a-distance' existed much earlier). The notion that it was a 'small signal' is totally false. Once you've hypothesized that the apparent mass of objects depends on the observer, you need to show that your theory gives consistent results for trajectories of objects in gravitational fields.
https://news.ycombinator.com/item?id=49177965
LLMs don't have legs yet. If you're smart but can't touch things you only keep being smart
I found this paper really thought provoking, but I think the conclusion of “world models are the solution” leaves something to be desired. People are already equipping agentic systems with physical simulation tools and exploring action-conditioned world models. This is cool because you can change the rules of the simulation and observe what happens, but it doesn’t address the core question of what to change the rules to, or even what the goal should be in the first place.
We just put different things together and then we evaluate it.
In math its simple: does the verification say its okay.
If its mechanical: is any property better than what we have already.
etc.
Neither can I, if I'm being honest.
It's a very shaky position, and the empirical track record of "LLMs can't..." is in itself a reason to call it into doubt.
Every "can't" of this nature was followed by a discovery of "they can, just poorly", and then by that "poorly" improving steadily generation to generation.
The paper doesn't provide a way to measure or quantify this elusive "jumping" capability, not even as an approximation. It just throws "can't jump" out there, as if "abduction" is an established class of problem with known computational properties and requirements that the LLM architecture fails to satisfy. It's none of those things - and the paper makes the claim without backing it by anything but rhetoric attempts at persuasion.
The proposed solution is also dubious. The empirical track record of dedicated "world models" for reasoning and problem-solving is, frankly, downright abysmal. Even integrating multimodal data into LLMs has failed to yield general reasoning capability gains.
LeCun's misadventures in the field aside, the main frontier lab that pushes in favor of "improving reasoning via multimodal fusion" is GDM - and Gemini isn't exactly a paragon of frontier reasoning capabilities. It has strong multimodal capabilities, but lags behind both OpenAI and Anthropic in performance outside that - while Anthropic is the lab that always treated multimodal grounding as an afterthought, and still trades blows with OpenAI at the very edge of the performance frontier. Multimodal grounding seems to work great as a way to improve an AI's ability to deal with those specific modalities, but it falters outside that.
Now, it's not impossible that everyone who tried multimodal world models for reasoning is just doing it wrong, and there is an undiscovered recipe for multimodal grounding that results in a step change in AI capabilities. But the results we have so far suggest it to be unlikely.
> the empirical track record of "LLMs can't..." is in itself a reason to call it into doubt ... Every "can't" of this nature was followed by a discovery of "they can, just poorly", and then etc
What LLMs are "fundamentally incapable" of doing has striking parallels to https://en.wikipedia.org/wiki/God_of_the_gaps
For those who may not be aware this is a clever title derived from a film title https://en.wikipedia.org/wiki/White_Men_Can%27t_Jump
This tracks for me as someone making keeps of intuition in little-explored areas.
I just don't see any of the LLM users around at all. Clearly some force is guiding them all away from thinking any of the "leap of faith" thoughts that I am thinking.
I have a weird thought experiment: If you give a GPT-2/3 level LLM tools to search the internet - any document, can it build bigger, better LLMs?
You may think this is not a good test because an older (or say a smaller) LLM can study from the knowledge on the Internet and build. But we are like that - we can access the Universe through our senses.
Can we ever produce anything that is beyond this Universe? I think an LLM that is lacking in knowledge can build more complex systems as long as it can access more data.
Interestingly, halucinations might be the way to achieve that.
I feel like you could just add some noise or randomness to the LLM and start approximating the leaps that the human mind uses to solve and understand unrelated things. Maybe that’s naive, it’s just coming from my organic computer in my skull.
Technically true, but the counter argument would be that the probability of this working would be ~ 2^(-(entropy_of_leap)) for an LLM (presumably intractable) and a human would succeed at a higher probability.
Hahaha you just derived temperature from first principles.
Turns out temperature is pretty bad too, you can find ways to sample from deeper in the distribution without distorting it. Great example is XTC (exclude top choices), In a few weeks/months it'll also have a proper scholarly paper with peer review.
LLMs can’t but humans supplied with data and reasoning from an LLM can make novel jumps without absolute prior knowledge, or at least with reduced need for years of knowledge.
And then such jump can be verified by a machine so human kind of plugs the intelligence gap.
That’s pretty exciting.
I always liked to provocatively call LLM „the new calculator”. Calculator for language.
We are so focused on creating a standalone intelligence that we didn’t notice how we massively augmented our own. That could be considered transhumanism holy grail if only interface brain-LLM was faster.
People need to understand that these things are tools. And every tool needs an operator to function. Tool doesn’t have its own goals, needs, wants or motives. It won’t do anything out of its own, it always exists in context of someone telling it what to do.
In light of that most of the panic and fear mongering is rather ridiculous. Calculator won’t replace you. It wont take over the world. It is just a tool.
You write a book with book generator? Cool, it can be used for this. We will judge output, not the methods. Sometimes we will judge people who have no taste in literature.
Calculator for language, is a nice insight. I visualize generally as LLM's as big mathematical expression that produce next word.
But then , one difference I found on LLM's is on scaling laws, where at some point, it have interesting emerging properties, that nobody thought would be possible now being possible.
Tools: Though It is a tool, but powerfull tool that is automating existing manually done jobs, large scale. Adapting to new roles where we definite goals, needs and motives judging of AI output, at large level is an issue. And humans are used to day to day repeating job, doing same thing repeatedly. Now the AI is taking over that. I see that is a challenge.
Also I see we are moving to creative world, where we will spend more time on creating something really new, leaving mechanized parts to AI.
Clearly LLMs cant do leaps of intuition since their "intuition" is locked after training ends.
The only way a LLM can come up with new ideas if the "idea" appeared as a generalisation durring training or if it was achieved using reason in chain of thought.
Or some randomness is aomehowntroduced in the output, which happens after every word, unless you set the temperature to zero.