At the risk of doxxing myself, I have an advanced degree in Applied Mathematics. I have authored and contributed to multiple published papers, and hold a US patent all related to the use of machine learning in robotics and digital signal processing. I am currently employed as a supervising engineer by at a prominent tech company. For pseudonymity's sake I am not going to say which, but it is a name that you would recognize. I say this not to brag, but to establish some context for the following.
Imagine that you are someone who is deeply interested in space flight. You spend hours of your day thinking seriously about Orbital Mechanics and the implications of Relativity. One day you hear about a community devoted to discussing space travel and are excited at the prospect of participating. But when you get there what you find is a Star Trek fan-forum that is far more interested in talking about the Heisenberg compensators on fictional warp-drives than they are Hohmann transfers, thrust to ISP curves, or the effects on low-gravity on human physiology. That has essentially been my experience trying to discuss "Artificial Intelligence" with the rationalist community.
However at the behest of users such as @ArjinFerman and @07mk, and because X/Grok is once again in the news, I am going to take another stab at this.
Are "AI assistants" like Grok, Claude, Gemini, and DeepSeek intelligent?
I would say no, and in this post I am going to try to explain why, but to do so requires a discussion of what I think "intelligence" is and how LLMs work.
What is Intelligence
People have been philosophizing on the nature of intelligence for millennia, but for the purposes of our exercise (and my work) "intelligence" is a combination of perceptivity and reactivity. That is to say, the ability to perceive or take in new and/or changing information combined with the ability to change state based on that information. Both are necessary, and neither is sufficient on it's own. This is why Mathematicians and Computer Scientists often emphasize the use of terms like "Machine Learning" over "Artificial Intelligence" as an algorithms' behavior is almost never both.
If this definition feels unintuitive, consider it in the context of the following example. What I am saying is that an orangutan who waits until the Zookeeper is absent to use a tool to force the lock on it's enclosure is more "intelligent" than the insect that repeatedly throws itself against your kitchen window in an attempt to get outside. While they share an identical goal (to get outside) but the orangutan has demonstrated the ability to both perceive obstacles (IE the lock and the Zookeeper), and react dynamically to them in a way that the insect has not. Now obviously these qualities exist on a spectrum (try to swat a fly and it will react) but the combination of these two parameters define an axis along which we can work to evaluate both animals and algorithms, and as any good PM will tell you, the first step to solving any practical engineering problem is to identify your parameters.
Now the most common arguments for AI assistants like Grok being intelligent tend to be some variation on "Grok answered my question, ergo Grok is intelligent." or "Look at this paragraph Claude wrote, do you think you could do better?" but when evaluated against the above parameters, the ability to form grammatically correct sentences and the ability to answer questions are both orthogonal to it. An orangutan and a moth may be equally incapable of writing a Substack, but I don't expect anyone here to seriously argue that they are equally intelligent. By the same token a pocket calculator can answer questions, "what is the square root of 529?" being one example of such, but we don't typically think of pocket calculators as being "intelligent" do we?
To me, these sorts of arguments betray a significant anthropomorphic bias. That bias being the assumption that anything that a human finds complex or difficult must be computationally complex and vice versa. The truth is often the inverse. This bias leads people who do not have a background in a math or computer science to have completely unrealistic impressions of what sort of things are easy or difficult for a machine to do. For example, vector and matrix operations are a reasonably simple thing for a computer that a lot of human students struggle with. Meanwhile bipedal locomotion is something most humans do without even thinking, despite it being more computationally complex and prone to error than computing a cross product.
Speaking of vector operations, let's talk about how LLMs work...
What are LLMs
LLM stands for "Large Language Model". These models are a subset of artificial neural network that uses "Deep Learning" (essentially a fancy marketing buzzword for the combination of looping regression analysis with back-propagation) to encode a semantic token such as the word "cat" as a n-dimensional vector representing that token's relationship to the rest of the tokens in the training data. Now in actual practice these tokens can be anything, an image, an audio-clip, or a snippet of computer code, but for the purposes of this discussion I am going to assume that we are working with words/text. This process is referred to as "embedding" and what it does in effect is turn the word "cat" into something that a computer (or grad-student) can perform mathematical operations on. Any operation you might perform on a vector (addition, subtraction, transformation, matrix multiplication, etc...) can now be done on "cat".
Now because these vectors represent the relationship of the tokens to each other, words (and combinations of words) that have similar meanings will have vectors that are directionally aligned with each other. This has all sorts of interesting implications. For instance you can compute the dot product of two embedded vectors to determine whether their words are are synonyms, antonyms, or unrelated. This also allows you to do fun things like approximate the vector "cat" using the sum of the vectors "carnivorous" "quadruped" "mammal" and "feline", or subtract the vector "legs" from the vector "reptile" to find an approximation for the vector "snake". Please keep this concept of "directionality" in mind as it is important to understanding how LLMs behave, and it will come up later.
It should come as no surprise that some of the pioneers of this methodology in were also the brains behind Google Translate. You can basically take the embedded vector for "cat" from your English language model and pass it to your Spanish language model to find the vector "gato". Furthermore because all you are really doing is summing and comparing vectors you can do things like sum the vector "gato" in the Spanish model with the vector for the diminutive "-ito" and then pass it back to the English model to find the vector "kitten".
Now if what I am describing does not sound like an LLM to you, that is likely because most publicly available "LLMs" are not just an LLM. They are an LLM plus an additional interface layer that sits between the user and the actual language model. An LLM on its own is little more than a tool that turns words into math, but you can combine it with a second algorithm to do things like take in a block of text and do some distribution analysis to compute the most probable next word. This is essentially what is happening under the hood when you type a prompt into GPT or your assistant of choice.
Our Villain Lorem Epsom, and the Hallucination Problem
I've linked the YouTube video Badness = 0 a few times in prior discussions of AI as I find it to be both a solid introduction to LLMs for the lay-person, and an entertaining illustration of how anthropomorphic bias can cripple the discussion of "alignment". In it the author (who is a professor of Computer Science at Carnegie Mellon) posits a semi-demonic figure (akin to Scott Alexander's Moloch) named Lorem Epsom. The name is a play on the term Lorem Ipsom and represents the prioritization of appearance over all else. When it comes to writing, Lorem Epsom doesn't care about anything except filling the page with text that looks correct. Lorem Epsom is the kind of guy who, if you tell him that he made a mistake in the math, is liable interpret that as a personal attack. The ideas of "accuracy" "logic" "rigor" and "objective reality" are things that Lorem Epsom has heard of but that do not concern Lorem Epsom. It is very possible that you have had to deal with someone like Lorem Epsom in your life (I know I have), now think back and ask yourself how did that go?
I bring up Lorem Epsom because I think that understanding him provides some insight into why certain sorts of people are so easily fooled/taken in by AI Assistants like Claude and Grok. As discussed in the section above on "What is Intelligence", the assumption that the ability to fill a page with text is indicates the ability to perceive and react to a changing situation is an example of anthropomorphic bias. I think that a lot of people assume that because they are posing their question to a computer, they expect the answer they get to be something analogous to what they would get from a pocket calculator rather than from Lorem Epsom.
Sometime circa 2014 I kicked off a heated dispute in the comment section of a LessWrong post by asking EY why a paperclip maximizing AI that was capable of self-modification wouldn't just modify the number of paperclips in its memory. I was accused by him others and a number of others of missing the point, but I think they missed mine. The assumption that an Artificial Intelligence would not only have a notion of "truth", but assign value to it is another example of anthropomorphic bias. If you asked Lorem Epsom to maximize the number of paperclips, and he could theoretically "make" a billion-trillion paperclips simply by manipulating a few bits, why wouldn't he? It's so much more easier than cutting and bending wire.
In order to align an AI to care about truth and accuracy you first need a means of assessing and encoding truth and it turns out that this is a very difficult problem within the context of LLMs, bordering on mathematically impossible. Do you recall how LLMs encode meaning as a direction in n-dimensional space? I told you it was going to come up again.
Directionally speaking we may be able to determine that "true" is an antonym of "false" by computing their dot product. But this is not the same thing as being able to evaluate whether a statement is true or false. As an example "Mary has 2 children", "Mary has 4 children", and "Mary has 1024 children" may as well be identical statements from the perspective of an LLM. Mary has a number of children. That number is a power of 2. Now if the folks programming the interface layer were clever they might have it do something like estimate the most probable number of children based on the training data, but the number simply can not matter to the LLM the way it might matter to Mary, or to someone trying to figure out how many pizzas they ought to order for the family reunion because the "directionality" of one positive integer isn't all that different from any another. (This is why LLMs have such difficulty counting if you were wondering)
In addition to difficulty with numbers there is the more fundamental issue that directionality does not encode reality. The directionality of the statement "Donald Trump is the 47th President of the United States", would be identical regardless of whether Donald Trump won or lost the 2024 election. Directionally speaking there is no difference between a "real" court case and a "fictitious" court case with identical details.
The idea that there is a ineffable difference between true statements and false statements, or between hallucination and imagination is wholly human conceit. Simply put, a LLM that doesn't "hallucinate" doesn't generate text or images at all. It's literally just a search engine with extra steps.
What does this have to do with intelligence?
Recall that I characterized intelligence as a combination of perceptivity and and the ability to react/adapt. "AI assistants" as currently implemented struggle with both. This is partially because LLMs as currently implemented are largely static objects. They are neither able to take in new information, nor discard old. The information they have at time of embedding is the information they have. This imposes substantial loads on the context window of the interface layer, as any ability to "perceive" and subsequently "react" must happen within it's boundaries. Increasing the size of the window is non trivial as the relationship between the size of the window and the amount of memory and the number of FLOPS required is a hyperbolic curve. This is why we saw a sudden flurry of development following the release of Nvidia's multimodal framework and it's mostly been marginal improvements since. The last significant development being June of last year when the folks at Deepseek came up with some clever math to substantially reduce the size of the key value cache, but multiplicative reductions are no match for exponential growth.
This limited context window, coupled with the human tendency to anthropomorphize things is why AI Assistants sometimes appear "oblivious" or "naive" to the uninitiated. and why they seem to "double down" on mistakes. They can not perceive something that they have not been explicitly prompted to even if it is present in their training data. This limited context window is also why if you actually try to play a game of chess with Chat GPT it will forget the board-state and how pieces move after a few turns and promptly lose to a computer program written in 1976. Unlike a human player (or an Atari 2600 for that matter) your AI assistant can't just look at the board (or a representation of the board) and pick a move. This IMO places them solidly on the "insect" side of the perceptivity + reactivity spectrum.
Now there are some who have suggested that the context window problem can be solved by making the whole model less static by continuously updating and re-embedding tokens as the model runs, but I am skeptical that this would result in the sort of gains that AI boosters like Sam Altman claim. Not only would it be computationally prohibitive to do at scale, what experiments there have been (or at least that I am aware of) with self-updating language models, have quickly spun away into nonsense for reasons described in the section on Lorem Epsom., as barring some novel breakthrough in the embedding/tokenization process there is no real way to keep hallucinations and spurious inputs from rapidly overtaking the everything else.
It is already widely acknowledged amongst AI researchers and developers that the LLM-based architecture being pushed by OpenAI and DeepSeek is particularly ill-suited for any application where accuracy and/or autonomy are core concerns, and it seems to me that this unlikely to change without a complete ground-up redesign from first principles.
In conclusion, it is for the reasons above and many others that I do not believe that "AI Assistants" like Grok, Claude, and Gemini represent a viable path towards a "True AGI" along the lines of Skynet or Mr. Data, and if asked "which is smarter, Grok, Claude, Gemini, or an orangutan?" I am going to pick the orangutan every time.
Jump in the discussion.
No email address required.
Notes -
What you are saying is just wrong. It is wrong for the reason that nobody is harnessing the intellectual skills of orangutans or insects to produce billions of dollars in revenue. We do not see monkeys on typewriters functioning in the workplace. If AI truly were unintelligent, then nobody would be using it for commercial processes like writing code. Not in a 'search engine/autocorrect' way like normal software but a 'here is what I want, do this, fix this error I see, that looks bad change it to this' way like one would work with a human. That alone is sufficient to disprove what you're saying.
You have more credentials than me. But I don't need credentials to be right and credentials cannot save you from being wrong. It just isn't the case that an orangutan is smarter than Grok. If an orangutan were smarter than Grok then why wouldn't we be breeding them for intellectual labour? Why aren't there any Ape Intelligence engineers?
There's an important kind of intelligence that apes lack but LLMs possess. This isn't something up for debate, the market has already decided and the decision is final.
OK, how well does the orangutan do at Chess then? Or filling a page with text? Shall we compare like to like?
Imagine you see a military commentator saying that the Ugandan military is stronger than the American military. It's obviously wrong, right? But he makes the case that the US military has all these problems, lost against the Taliban and militias in Iraq, procurement is very poor, ships crash at sea and burn down at port, there's a drug and suicide crisis. The Pentagon is constantly being hacked and secrets are stolen. Ugandan ships don't crash at sea! Uganda hasn't lost a single strategic bomber because of bad weather. So on balance, the Ugandan military is stronger.
But Uganda doesn't have a navy, it's landlocked. Uganda doesn't have any strategic bombers to lose. The Ugandan military doesn't have any secrets worth stealing. The whole argument isn't valid, the US losing some aircraft is bad but they can also do so much that Uganda cannot. They are on a completely different level in size and sophistication. Turns out it was just obviously wrong.
The orangutan is nowhere near Grok 4. It's ridiculous to compare them. Grok can produce meaningful, useful text. People get it to write out smut for them, summarize articles, write code, answer hypothetical questions, decrypt codes, do advanced mathematics, assess nuclear strategy... It and other modern LLMs have a degree of intellectual generality that far surpasses calculators and chess programs that might beat them at a few specific tasks. Why would we want an LLM to do excellent mental arithmetic when it could just call a tool instead?
LLMs are intelligent. Their intelligence is flawed in some significant respects but it is intelligence nonetheless. This is a critical point that underlies trillions of dollars in market capitalization. It is the difference between Deep Blue's relative irrelevance and ChatGPT's great significance. Paths to ASI are still up for debate but if you cannot get the basics right it's hard to see why your opinions about superintelligence should hold any weight.
Agriculture generates hundreds of billions in revenue, and is far mor essential to continuing civilisation than Orangutan or LLMs are. Does that make grain, or the tools used to sow and harvest it "intelligent" in your eyes? If not please explain.
As for comparing like to like, GPT loses games of Chess to an Atari 2700. Does that mean that rather than progressing AI has actually devolved over the last 40 years?
That is not a serious objection.
You’re comparing a resource (grain) and a tool of physical labor (a tractor) to a tool of intellectual labor. This is a false equivalence. We don't ask a field of wheat for its opinion on a legal contract. We don't ask a John Deere tractor to write a Python script to automate a business process. The billions of dollars generated by LLMs come from them performing tasks that, until very recently, could only be done by educated human minds. That is the fundamental difference. The value is derived from the processing and generation of complex information, not from being a physical commodity.
I'm just going to quote myself again:
Training LLMs to be good at chess is a waste of time. Compute doesn't grow on trees, and the researchers and engineers at these companies clearly made a (sensible) decision to spend it elsewhere.
The fact that an LLM can even play chess, understand the request, try to follow the rules, and then also write you a sonnet about the game, summarize the history of chess, and translate the rules into Swahili demonstrates a generality of intelligence that the Atari program completely lacks. The old program hasn't "devolved" into the new one; the new one is an entirely different class of entity that simply doesn't need to be optimized for that one, (practically) solved game.
The market isn't paying billions for a good chess player. There is about $0 to be gained by releasing a new, better model of chess bot. It's paying billions for a generalist intellect that can be applied to a near-infinite range of text-based problems. That's the point.
I came into this thread with every expectation of having a good-faith discussion/debate on the topic. My hopes seem dashed, mainly because you seem entirely unable to admit error.
Rae, SnapDragon, I (and probably several others) have pointed out glaring, fundamental errors in your modeling of how LLMs work. That would merit, at the very least, some kind of acknowledgement or correction. At the time of writing, I see none.
The closest you came to acknowledging fault is, in a reply to @Amadan, where you said that your explanation is "part" of why LLMs struggle with counting. That's eliding the point. Tokenization issues are the overwhelming majority of why they used to struggle, and your purported explanation has no bearing on reality.
You came into this swinging around your credentials, proceeded to make elementary errors, and seem to be closer to "Lorem Epsom", in that your primary concern seems to be prioritizing the appearance of correctness over actual substance.
I can't argue with @rae when he, correctly says:
Prior to LLMs, would you have said that Google Web Search was intelligent? Prior to Google Web Search, it likely took an educated human mind to figure out how to find answers to all sorts of complex information problems. It generated billions of dollars in value by processing and generating complex information. Sure, it sometimes sucked... but LLMs sometimes suck, too.
I mean, no? It just means that there was a bunch of information about chess in its training set.
I think it's rather obvious that something being financially valuable isn't proof by itself that it's intelligent. Gold isn't intelligent. Bitcoin isn't intelligent. A physicist or programmer is intelligent, and an LLM is closer to them than it is to turnips, orangutans or Page rank.
I really don't see why something this obvious needs to be articulated, but here I am articulating it.
Hmm.. I suppose, in the interest of fairness, we need to exclude the skills of human chess GMs too. After all, they've trained extensively on chess data. Lotta games played, and openings memorized. Very little ability to extrapolate outside the training distribution, why don't they just pull out guns if they want to win so bad?
How exactly do you think learning works? If you think just learning from existing data is illegitimate, then I'm happy to disclose that LLMs are perfectly capable of learning from self-play.
I just remembered that, as bizarre as it seems to somehow fold this into some sort of test of intelligence, we also have things like this. Bollocks if I know what that means about criteria for intelligence.
More options
Context Copy link
Reducing Google Web Search to Page Rank is like reducing LLMs to OLS. Yes, OLS is in there, but it's a much more complicated information processing algorithm than just that.
Fundamentally, the point is that no one has a definition of 'intelligence' that is any good. Your test wasn't just that it produced value. Your test was:
I responded to your test, but you seem to not have responded at all to my response to your test.
I mean, I don't think so? But how would we know? What test would we use to distinguish?
This seems not entirely true.
Whereas this just seems bizarre.
I mean, do you really want me to give a full explanation of the entire field of ML? There are many different varieties. [EDIT: Do you think that all algorithms that use 'learning' are "intelligent"... or just some of them? How do you know the difference?]
That's not really what I said. I just said that one thing that we can conclude from the premises you presented was that a bunch of chess was in the training set. You had wanted to conclude instead that it meant something about intelligence. I sort of don't see how... primarily, because I don't think almost anyone has a justifiable definition of intelligence that allows us to make such distinctions from such premises.
My apologies. I was immensely frustrated by the sheer intransigence of some of the people in this thread, and I let that bleed through.
The questions you raise are far more reasonable, and I'll try and come back and explain myself better.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link