Chomsky and the two cultures of statistical learning (2011)
norvig.com
norvig.com
Norvig confuses engineering for science. If an engineer wants something that works (and guarantees that it will work) then you use the most effective tool available to achieve that goal. No problem there. But there's absolutely no reason to think that tells you anything scientific about the phenomenon you're modeling. For example, aeronautical engineering says very little about how birds fly. And - currently relevant - LLMs learn to use and understand language, but they use millions of times more training data, and millions of times more computing power, than a child does. So an LLM may be a great technology, but it's probably not a good model for how humans learn and use language.
Chomsky OTOH doesn't see any use for statistics in science, at least not in linguistics. Fortunately for the field of linguistics, Chomsky's influence in linguistics is waning and linguists are beginning to use statistical tools to do science.
I don’t think Norvig is confused. He explicitly addresses this point right at the top of the article.
His argument is that successfully engineering a system using a statistical method provides some bits of evidence for the hypothesis that the system itself is statistical.
Which is obviously true, so long as we take care to weigh evidence properly.
> there's absolutely no reason to think that tells you anything scientific about the phenomenon you're modeling
No, this is not how evidence works. If you can accurately model a system with method X and you cannot come up with a model that works with method Y, you should think it’s more likely (but not certain) that X is the actual physical model.
And indeed modern cognitive science (see Predictive Processing) also has a bunch of evidence for the brain as a hierarchical predictor, rather than a rule engine.
To your points - planes are not model organisms of birds, but they do shed insight on the bounds that birds must obey, eg thrust-to-weight etc. And the sample efficiency point really needs to die. The human brain is born with an extremely complex prior baked in. An LLM begins pretraining with almost nothing (just a very simple architecture and a few simple algorithms like attention), and has to learn the basic architecture of cognition from scratch. A better baseline would be how many samples were processed by every animal in our genetic ancestry that had a brain, since brains existed.
Could you expand on this point?
I was also going to suggest coin flips, but it’s not quite the same… I think any chaotic system is generally best modeled that way, and if you cared about the effect the system has on other things, it’s unlikely you can do better since measuring the initial state will not be feasible.
Math is not physics and physics is not math. Writing a Superman story does not mean the language and pictures represent physics
That's the difference being highlighted here; stats are not evidence of gravity and electromagnetism; physical phenomena when materials influenced by those forces are relative to local gravity, electromagnetic fields are our evidence
Stats of 1960s demographics did not capture immutable physical phenomena because entropy-> generational churn -> those people are now dead
The languages do not matter; they're just to communicate some ground truth.
Shackling society to the past through careful statistical methods relative to finance and politics to avoid societal change is not a scientific conclusion; there's no reason to use 30-40 years ago as the center of our social universe
The concern with stats consumption is we walk away from a result as if its an immutable forever conclusion as we ignore that gravity and electromagnetism, etc have not stopped changing ground truth; past truths that no longer matter to our next choices
You can write many imagined stories, but one of them will match observed reality. It won't be about an airborne male model who wears his underpants on the outside. But you can write a story that describes reality. And you can describe observable phenomena with statistics. Of course, you can describe plenty of pseudoscientific drivel with statistics too.
in college, i was somewhat involved in cogsci wrt semantic mapping of linguistic concepts based on human surveys (this was well before LLM/AI-word2vec), and it surprises me today how the mathematical properties of relatedness/distance seem now to drive machine intelligence, something which we did not consider at the time at all.
maybe i'm getting in my metaphysical sentimental stage, a turn a lot of adults make after middle age, but i wouldn't fully dismiss the fact that there may be an equivalence.
There's pretty interesting (I think) research in how effectively it's possible to scale down LLMs. If you're interested, take a look at the BabyLM conferences:
(I have no connection to them.) The work there demonstrates that gap between what we can do now with human-relevant amounts of training text, and what remains to be discovered. Quite a bit, it looks like.
I mean shifting influence of gravity relative to an object with enough mass type ratios are everywhere
The language of math started as visual art to depict such ratios; in modern times this ground truth associations is frequently overlooked or never understood to exist in the first place without engagement
Differential diagnosis in medicine is exactly the same process; such processes and their successes are everywhere.
The issue we have in software is the outcome isn't as obvious as jabbing a blood draw needle in a patients eye; we spent a lot of time wasting time on illusory needs
Can you use a human language well enough and it not sound intelligent? And no, I don't mean just syntactically correct because semantics is a part of human languages too. I think the astounding thing is that they managed to model a language this well. It's like trying to learn English without a functional understanding of the word "Please".
My take is that a lot of the intelligence we natively ascribe to humans exist in human languages and just like the nutrients in your food, by harnessing a language that possesses latent intelligence, you too can behave like an intelligent person. And how did the nutrients of intellect got in there in the first place? Just like languages evolve nowadays: via contributions from speakers over time. Not unlike soil fertilisation.
For example, if I compose your post as :
--
"Good prose embodies intellect gradually evolved by myriad humans over time, so it wows me to see LLMs create it, though I soon realised that the intelligence is not being created by the LLM but rather is already present in the human-evolved input datasets - the prose itself! - which they statistically arrange."
--
..the particulars of the language are entirely different (and omit some of your chosen nuance) but most readers would receive the same "core truth" from it, the same measure of beneficial information.
Which highlights the great fluidity of language; that is another of the key underlying mechanisms allowing LLMs to flourish.
Regardless, the fluidity of language is self-evident to an extent : it is partly embodied in the entire concept we call figures-of-speech (a linguistic phrasing can be literal, or sarcastic, or described by metaphor, ...). Partly in the common reality that our dictionaries list multiple possible definitions for many words. Partly in the fact that (many) synonyms exist. Partly in sheer human individuality.
It's funny, I used to think it was complete and utter horseshit when literature professors would say that the meaning of a text is created by the reader, not the author.
Since LLMs have become commonplace, I have completely reversed my opinion on that.
That you see recognizable intelligence in the output of stochastic generation is because you are imposing meaning on it. It is why "claudish" is so hard to read--you are trying to impose meaning on a sequence of words that was stochastically generated, and it takes mental effort to do that.
If Claude can write formal mathematical proofs (that a machine can check), or it can write software that works, where's the head of the reader that imposes the meaning?
> you see recognizable intelligence in the output of stochastic generation is because you are imposing meaning on it
> a sequence of words that was stochastically generated
As was your comment (and mine, and the others on this page and elsewhere).
With Claude I know that there is no source meaning or belief, but it helps to impose Dennett's "intentional stance" and treat it as if there were, up to a point. That there is such a "point" allows be to manipulate it in my attempts to extract factual information implicit in its training data, expressed probabilistically via its stochastic processes.
It most certainly does not. It's obvious to anyone who critically examines LLMs, without letting the hype override sense, that there's zero intelligence there.
That is, you can apply for grants and write papers in the transformational grammar paradigm endlessly. You can pose and answer questions that make sense within in the system.
Chomskyism was disappointing in two respects: (1) was that it was an inadequate basis for machine understanding of text, (2) was that structuralists were hoping everything in culture could be understood in terms of "... having the structure of language" but syntactic structures weren't all adequate for that task.
(1) and (2) both pointed to Chomskyism being hermetic and not something really useful or inspiring to other fields and maybe just a little less "true".
Also, I'll mention that there is a lot of empirical knowledge about natural language syntax in generative grammar, even if the details aren't quite right about their theoretical underpinnings. And a lot of them were "low-hanging fruit" waiting for someone to point out how to find them and characterize them. That Chomsky did that is the best argument for his intellectual value - permitting all that "normal science" to happen.
Science doesn't guarantee technological applications, whether or not that science is correct.
Finally, just saying: use of a word like "Chomskyism" makes you sound like one of those creationists who call evolution "Darwinism".
This is wholly fallacious ... you may disagree with what you interpret as the implications of attaching "ism" to Chomsky, but in no way does that equate it to attaching "ism" to Darwin -- each of these attachments and their validity are independent.
BTW, even what you see as pejorative in such an attachment is a cultural interpretation ... e.g., it is far more common among British evolutionary biologists than American to view "Darwinism" neutrally -- Dawkins for one uses the term and describes himself as a Darwinist. See, e.g., https://www.edge.org/conversation/richard_dawkins-chapter-3-...
Yes this is one of my critiques of the academic system (I have finished a PhD and will hopefully never work at a university ever again). Many papers in the introduction start with explaining how the study builds on plenty of previous research, but that alone doesn’t prove that something is a good direction to investigate. If academics were completely honest, many papers would write “We researched topic X because we got funding for it. But since we actually don’t care about X, we just wrote a proposal about our favorite niche topic that only one of the co-authors cares about, namely topic Y, and wrote it such that it sounds like X. Next we show that Y is actually very cool in the hope to convince others that it is cool plus to convince others of the conclusion that our most senior professor already found during the PhD which was 30 years ago. Apart from the professor, we don’t believe in the main argument of the paper but since he got the grant we’ll just go with it in the hope to one day be able to push our own ideas forward”.
this distinction is superficial. Scientists use less precise models all the time. Like, you don't solve the schrodinger equation even when you're a scientist doing organic chemistry. You reach for lewis dot models. Science as a discipline is about building better models of the world, not uncovering fundamental truth.
but from all models, it's the closest, "all models are wrong, but some are useful"
Care to back this up? What evidence do we have to suggest that back propagation, self attention, and byte pair encoding are mechanistically similar to human cognition? Similarity of the outputs says nothing about similarity of the machinery underneath. Turboprop engines tell us nothing about how birds fly - doesn't matter that they successfully transport millions of people around the world.
I would say that when it comes to higher level cognition, logic engines are more human like even if they only have a tiny fraction of the practical utility of LLMs.
That is not true.
Turboprop engines and flying birds are different things, but there is a set of common properties.
A blade of a screw propeller functions exactly in the same way as a flapping wing of a bird, except that the wing cannot rotate without limits around the body of the bird, so it must stop, change angles and revert the motion, while the blade can rotate continuously around the axis, functioning on one side like a falling wing and on the other side like a raising wing. The similarities between propeller blades and insect wings are even greater, because insect wings cannot change their shapes so much as those of birds.
For now, our understanding of the networks composed of the neurons of animals is much less advanced than our understanding of wings and propellers, which is why there is a much greater uncertainty about whether there are any similar mechanisms in biological neural networks and in LLMs.
"Nothing" may be a little too strong -- physics is physics, after all -- but (small) birds and airplanes have Reynolds numbers about 1000x different and different underlying material science.
Airplanes are engineered (with a very small number of exceptions) to be stable during flight, whereas bird flight is inherently unstable, requiring a lot of specialized cognition so birds can fly.
I would think a lot of cognitive ornithologists (who I really hope exist!) would be interested in finding out about this.
At this point, I'm at the complete end of my knowledge of bird flight!
The "except" in your argument is doing an incredible amount of work. The rotation of a turboprops generates forward force, wings (whether a bird's or a plane's) generate lift. And plane's wings are called "wings" because they, and not the turboprop, mimic the structure and function of a bird's wings.
Regardless, this whole line of argument is missing the forest for the trees.
And I'd be hard pressed to identify many cognitive characteristics of LLMs during execution that resemble human thinking, aside from those that are readily parrotable.
As a model of human cognition, LLMs deep suck.
We considered the hypothesis that humans are a general intelligence that figures language out by exposure. The evidence has consistently come back that that is not how humans actually work. A model where humans are hardwired with a lot of knowledge about language, then fill in the gaps based on exposure still appears to be far more accurate.
The only thing LLMs show is that it is possible to learn language de-novo.
I agree, but would add, "But not under conditions that humans learn language."
Is there some other competing model which can give results as good? If not, the statement seems true in the trivial sense that it's the _only_ model which we've observed to give good results.
Regarding "hard-wired"; it's not if you consider that "hard-wiring" of an individual is the result of millions of years of evolution.
You can't count by text, because text is not the only thing children learn on. So unlike the bytes comparison the text comparison is totally invalid.
> We happen to feed only distilled data to LLMs though. That's why they appear so efficient.
That is wild handwaving that really explains nothing.
But then, most of the comments on this page are wild handwaving, pontificating like a blind man describing an elephant. Very little engages with or matches the quality of Norvig's arguments.
Have you ever had kids? Either you never had, you forgot, or you are intentionally deceiving us (and perhaps yourself) to support your point. If you never had kids, I can tell you you have very poor imagination about how they learn language.
Any specific entries?
"This year, we have merged the Multimodal and Interaction tracks into the Strict and Strict-small tracks. This is for two primary reasons: Firstly, there were insufficient submissions to warrant standalone tracks. Secondly, none of the submissions in the Interaction track outperformed models submitted to the Strict track, and *none of the submissions from Multimodal outperformed our baselines*. We will still allow participants to use multimodal data or interactive training procedures if they choose, but such data or learning procedures must conform to the rules of the Strict or Strict-small tracks.
Emphasis mine.
I think that's a pretty weak point to your argument. They are saying that none of the contestants outperformed their baseline multimodal models and that there was very little multimodal submissions in the first place.
The multimodal baselines are _lower_ than the text only baselines. See Table 3. In fact, 10M of text outperforms 100M of text + images.
> I understand how Chomsky might arrive at the conclusion that probabilistic models are unnecessary, from his study of the generation of language.
I think the point turns here for me. We have probabilistic models now that generate unbelievably convincing language. Norvig's following argument on the value of interpreting language statistically becomes secondary.
In order to stay moored in reality, our models must follow from evidence. I think it would take a peculiar obstinance to deny that the latest models show evidence of generating language.
An LLM must construct all of its grammatical structure, world models, etc. from scratch during pretraining.
LLM pre-training is more like recapitulating evolution of the mammalian brain than childhood development.
LLMs process millions of times more text data and use millions of times more compute power to learn a language, mostly because we know how to do that technologically.
Children use much less text data, but also use a lot of other kinds of data, including interacting with other humans, to learn language. We have no idea how that works. And, on top of that, people can interact with pets of various species in the same way as children, but those pets never learn language.
So, something strange is going on that children can do, but that LLMs and puppies can't. Probably, part of it is that human children are much better at learning by copying what adults are doing, and part of it is that humans evolved to learn and use language. The details of what's going on there is what linguistics is about.
The question isn't "are neural networks just like human brains?" It is "can statistical models generate and understand language?"
The answer to the actual question under discussion appears to be a resounding yes. It does not appear that learning language is innate in the way that Chomsky has argued.
It may be that learning language as efficiently as human children will involve some innate property that we have not yet discovered. But that is a perfect example of moving the goal posts. We go from "there is something special about the brain that allows us to learn language, something that cannot be reproduced statistically" to "there is something special about the brain that allows us to learn language efficiently, more than statistical models".
But the situation has moved on, moving even further beyond the assertions Peter makes about the utility of the black-box generative approach.
In particular, LLMs are telling us a lot about those low-probability, never before seen sentences — it’s called CALU, the creative aspects of language use.
LLMs exhibit CALU; as we know they will happily create such sentences. And inspecting the LLMs does provide insight:
“It does change the epistemic situation substantially. A working system can be inspected, perturbed, retrained, lesioned, and compared with other systems. Building is not understanding, but it gives understanding a strong foothold.
Explaining creative language use in LLMs would not, by itself, explain creative language use in humans. But the artificial case matters even if the mechanisms turn out to be substantially different. At a minimum, LLMs are an existence proof of one pathway by which open-ended, context-sensitive language can arise.”
For much more, see this substack by Chandra Sripada, a U of Michigan linguist:
https://chandrasripada.substack.com/p/ai-and-the-retreat-of-...
He is interviewed by Sean Carroll on his excellent Mindscape podcast:
https://preposterousuniverse.com/podcast/2026/08/10/363-chan...
Chomsky and the Two Cultures of Statistical Learning (2011) - https://news.ycombinator.com/item?id=46286533 - Dec 2025 (119 comments)
On Chomsky and the Two Cultures of Statistical Learning - https://news.ycombinator.com/item?id=35081102 - March 2023 (135 comments)
On Chomsky and the Two Cultures of Statistical Learning (2011) - https://news.ycombinator.com/item?id=16489828 - March 2018 (12 comments)
On Chomsky and the Two Cultures of Statistical Learning (2011) - https://news.ycombinator.com/item?id=11951444 - June 2016 (102 comments)
On Chomsky and the Two Cultures of Statistical Learning - https://news.ycombinator.com/item?id=2591154 - May 2011 (107 comments)
(Reposts are fine after a year or so; links to past threads are just to satisfy extra-curious readers)
There's enormous irony in Chomsky's mathematicization of linguistics sweeping away the linguists who came before him compared to his own stubborn refusal to acknowledge the fruits of the statistical approach. As far as I know, he's more or less been swept aside.
Appeal to Authority is over-emphasized in academia in general though.
Why the "simplest" model? Is a probabilistic model not the only model able to deal with uncertainty? What other kind of models are possible to represent uncertainty and inherent ambiguity that are not probabilistic?
A probabilistic model is the most complicated model. It doesn't even attempt to model the problem, it just looks at the results of an enormous number of tries with different inputs, goes through them all, finds the most similar cases to the current input, and blends them together in a way that carefully avoids any self-contradictory or absurd result.
We see this plainly with LLMs. You train an LLM on a thousand times the (high-level, pre-digested) input that the average human has gotten, and the average human thinks it's a hallucinating moron. It doesn't start to converge on something that passes the smell test to you average human idiot until it has been trained on a million times the input the average human has gotten.
It's not doing what a human does, it's guessing what a human (or any system) might do when confronted with X input. They're not even similar activities, even if they have identical output in the limit. It's like confusing gravity for equations about gravity. Being able to make perfect predictions about the positions of moving, nearby frictionless spherical things in a vacuum at time t doesn't tell us a thing about what gravity is. It tells us how it looks. It should, we distilled it by watching things for a very long time.
> Is a probabilistic model not the only model able to deal with uncertainty?
No, all models deal with uncertainty. The entire point of a model is that it is what you know about a system when you don't know its inputs. The variables in a model are uncertainty. The model is dealing with it.
edit: I think you're interpreting "uncertainty" as a numerical probability. A numerical probability is an artifact of the set of data you're calculating it from, not a thing with independent existence. "Uncertainty" is just a thing that you don't know yet.
If you assume that "uncertainty" is a knowable numeric probability that contributes to some calculation, you're begging the question when you say that the only way you can model it is through probabilistic calculations.
For the domain I am working in, a model is considered deterministic if for a set of input it always provide the same output and the input are assumed to be perfectly known. A model become probabilistic, once the input or the model itself are considered as uncertain. As an example: previously weather models always provided singe prediction for tomorrow, for example rain or no rain. But now, such models are probabilistic in the sense that an ensemble of model forecast are made (with slightly different input representing our uncertainties) to provide an estimate of the probabilities. Indeed such probabilistic model is not the simplest, but to most complete and so far the only practical way to handle with uncertainty for system like the weather as far as I am aware of.
Ternary logic is a super simple model; where true AND unknown equals unknown, and true OR unknown equals true. This deals with uncertainty and ambiguity, just not the probability; notionally you could say everything is 50-50.
Same thing if you have e.g. text-to-speech which returns "the b??? dog" rather than "the (0.8:black, 0.1:blue, 0.1:boy) dog" - no probabilities. Of course, maybe later in the processing you add probabilities back in.
Then there's fuzzy logic, which uses scores that are typically numbers, but those don't express actual probabilities, but more like vibes. They don't necessarily add up to 1, for instance, and you don't apply Bayes' formula to them, or even multiply them, you can use addition too - yolo.
Edit: and also of course backtracking, beam search, that sort of stuff - needn't use probabilities, just heuristics.
Anyway, the probabilistic model doesn't exist. Simple probabilistic models work quite badly on speech recognition (word classification, etc.). A lot of structure needs to be added before they start performing.
(One could make a meta statement here about how faulty some human's modeling of their linguistic inputs is.)
1. Norvig is correct that observation and fact/data collection are necessary for science. However, Chomsky is correct that the ultimate goal of science is to determine the principles governing something. Ergo: you need the first to do the second. The intellect abstracts from source data, but nothing is in the intellect that is not first in the senses.
2. Chomsky here is uninterested in things like speech - he only cares about the underlying syntax. But syntax - expressed and instantiated in matter - is known through speech, and probabilities presuppose an underlying determinate syntax. Ergo: a complete science of speech is end-to-end, from the "reading" or interpretation of noisy data through syntactic processing all the way through to expression.
3. Chomsky focuses heavily on internal generation. Norvig points out that interpretation is a separate, highly practical challenge. Figuring out what someone means in a noisy, ambiguous world is an exercise in prudence which involves weighing contingencies and probabilities to arrive at the most likely truth. Of course, this can only be done against a presupposed acceptance of the function of speech which is to communicate meaning and what is true (hence why lies work and why they are an abuse of speech). Probabilistic reasoning is still reasoning. It's reasoning in the face of incomplete information (against a backdrop of presuppositions) which is a component of prudential reasoning.
Note that the base epistemology of science is inference to the best explanation, not "truth".
Have to say that I confronted myself with the fact that I can't read longer posts anymore due to overuse of AI. Guess a detox is required...
"..if we think of truth as something of granite-like solidity and of personality as something of rainbow-like intangibility and reflect that the aim of [human writing] is to weld these two into one seamless whole, we shall admit that the problem is a stiff one and that we need not wonder if [human writers], for the most part failed to solve it." - Virginia Woolf
[human writing, human writers] = [biography, biographers] in the original quote. My edit is because I infer that her statements key point is that of the biological subjectivity inherent to all human prose.
Because : consensus of what? of sheer numbers of human adherents? of other statistic? historical evidence? of widest agreement by those with higher education in that field? of those with the most money to afford advertising? other? a consensus of multiple consensus(es)?
The dynamic of myriad possible vectors through language to express a given 'truth' remains, whether we define truth as empirical or consensus.
Your founding premise, that observed data (that is one definition of 'empirical fact') and "consensus" are somehow opposing ideologies seems flawed however.
If you would prefer to read the term consensus in my original post, rather than empirical fact, then please do - it does not change my point.
Linguistics, obviously.
Philosophers love him, because he always gives them an argument, and that's all they really care about.
There probably wouldn't even be cognitive science without Chomsky. And psychology is still responding to Chomsky's nuking of behaviorism in the 1950's.
So, there's a lot there, but admittedly most of it is for politics.
Just look at his Wikipedia entry. He finished his dissertation in 1955, starting teaching at MIT the same year, and continued to publish seminal linguistic works over the next 12 years. That's all before he gained any notoriety whatsoever as a critic of the U.S.'s role in the Vietnam War.
Manufacturing Consent was published in 1988!
Additionally, I strongly doubt Chomsky had any "name recognition" all the way through to the late 90s. Outside of a single episode of Firing Line, where would the general public have heard his views?
That said, as a computer science student I learned of his hierarchy of formal languages before I learned of his politics.
A formal theory of finite automatons, both deterministic and non-deterministic, is due to Michael Rabin & Dana Scott, for which they received a Turing prize. 1959.
No offense but your question is terribly ignorant, naive, and biased. Just spend a few seconds reading his Wikipedia article.
It's sort of like criticising Galileo's putative experiment about dropping two balls from the leaning tower of Pisa by pointing out that a feather indeed falls slower than a iron ball.
do you want a cracker -> you want a cracker?
I think he is dumb -> * I think is dumb
It's a lazy dismissal from someone who doesn't know the relevant facts.
Of older English words, there are extremely few genuine exceptions. Seize is one.