If performing well on an IQ test or performing at a high level on knowledge work is intelligence to you, these models are intelligent. If intelligence requires sentience for you, then ... well, I don't think we really agree what that is either, never mind how to measure it. But LLMs certainly don't have it right now
But the consistent trend of the last couple decades (arguably since Turing's time) seems to be that any time a computer reaches our definition of intelligence we decide that that was a flawed definition
I do recall a couple of decades ago, when the Turing test was discussed as the big goal that seemed so far away. Then LLMs arguably did pass the test, and no one cared about the test anymore.
I'd figure out that it's an LLM because it's effectively superhuman. Taking that away I'm not so sure I'd be able to tell
If these things have consciousness then we are committing sadism on a massive scale.
If someone sat me down today with an LLM and a human and both were trying to prove to me they were human, and I can have conversations of arbitrary length, I’d get it right every time.
We might eventually regret exposing the general population to such a new technology without almost any safeguards.
So the first thing they’d do is tell me LLMs exist and the other thing is an LLM. Obviously a true human level ai could explain that away as a fabrication to trick me. I don’t think an LLM could do even this!
Turings whole point was that through the medium of text along if the human and machine were indistinguishable then that was true intelligence. So yes conversations of arbitrary length are allowed (needed).
https://arxiv.org/abs/2503.23674
From the abstract: "When prompted to adopt a humanlike persona, GPT-4.5 was judged to be the human 73% of the time: significantly more often than interrogators selected the real human participant. LLaMa-3.1, with the same prompt, was judged to be the human 56% of the time"
And people forget that sometimes humans message twice. An LLM can only respond. So it immediately fails here in a true Turing test. (You could loop the LLM but then I expect even more immediately obvious bot behaviour).
Are you serious? From the paper:
> We recruited 126 participants from the UCSD psychology undergraduate subject pool and 158 participants from Prolific (Prolific, 2025).
Each human participated in 8 rounds.
> time bound
The time bound of 5 minutes was suggested by Turing himself in his original paper.
> not reproduced
It was reproduced across two populations within the paper.
> And look at their example conversations
This is irrelevant.
And it’s all irrelevant. If one human on earth can consistently get it right then it hasn’t been passed since clearly that human can somehow determine between them (whereas no one would ever be able to determine between a true “human intelligence” by definition).
And as it stands almost everyone could tell between them when allowed to discuss whatever they want for any length of time.
The fact these researchers have to keep adding bounds shows it hasn’t been passed. If we are arguing over technicalities maybe it isn’t as obviously intelligent as claimed!
Right, the test duration was left unspecified. This means any duration is acceptable. Including, for example, the only duration actually mentioned by Turing himself in his paper. Or do you have a more authoritative source on which durations are acceptable?
> If one human on earth can consistently get it right then it hasn’t been passed
Says who? Not Turing. Probably he didn't say that because it would make the test both impractical and overly conservative.
> The fact these researchers have to keep adding bounds
What "bounds"?
The speed of the goalposts here is just amazing.
Probably. Hopefully.
Why are you so sure of that? If you say yourself that we can't agree on what it is, and have no trusted measurement tools for it.
LLM sentience is firmly in the realm of "maybe".
I think the mistake here is the notion that there was a definition of intelligence. Or at least a consensus on that definition. Just because compsci nerds of the day thought the Turing test was the final threshold before “real” AI, doesn’t mean philosophers, psychologists and everyone else bought into it. And when we arrived and it turns out to be underwhelming it’s because the compsci nerds made the same mistake they always make: that their models truly encompass all the dense complexity of the real world.
https://www-cdn.anthropic.com/564f962e60643842f5fcb4a17c9dbc...
Everyone decides what to think on this issue, then finds out facts to support their idea.
As it stands they are massively useful tools, but for generating usable products they require either A) a lot of expert steering or B) a well defined easily verifiable target and a large compute budget. Most people are using them in mode A with good effect, the progress on math has been done in mode B, which is very promising.
Just a year and a half ago their maximal use was rephrase, summarize, and homework-level tasks.
Five years from now? There be dragons.
"But are they generally intelligent?" What a meaningless question!
We can reap the benefits while clearly telling the consumer this is just a language algorithm.
This is not intelligence. It's just a good correlation engine with a very big albeit lossy database of things.
As my AI professor said in the first lecture: “All AI is advanced search”.
Many things are predicted by models in our planet. From weather to production and material science. Building the model needs intelligence, running the model does not.
The person who came up with the formulae for CFD was intelligent. The computer running the model is not. Same for LLMs, chess engines, engine ECUs and financial prediction systems.
Again, for the example’s sake; the person who came up with an algorithm is intelligent. The model mixing its training data to emit something similar is not.
So when LLMs can do all human knowledge work, and do it better than humans, we'll be in the mines listening to you go on about how it's actually just autocomplete or just math, a distinction that apparently means nothing.
No.
> So when LLMs can do all human knowledge work, and do it better than humans, we'll be in the mines listening to you go on about how it's actually just autocomplete or just math, a distinction that apparently means nothing.
With a big "if" attached to it. People were saying "computers will program themselves in the near future" for, checks notes, 24 years now, as far as I'm aware.
We're constantly building new knowledge and understanding things better than olden days. These models just compress our knowledge and light the blind corners we can't see well. I don't say they are useless, but I say that these things are overhyped.
All they can do is regurgitate human knowledge packed into them and highlight some long-distance correlations between items, which is useful in itself, but it can't jump to somewhere where it's not present its training data, but that's something humans and only humans can do.
Most of what you said reads to me as denial.
An unconscious unintelligent but persistent trial and error process created us. We created LLMs. LLMs may create the next thing before we do - hard to say. They don't have all the cognitive tools we have yet, but they still outperform in some areas. As the cognitive playing field levels, I expect you will come to eat your words..
That sounds like something that can be engineered, can't it? In other words, we can identify limitations in current transformer-based architectures, and we can also build new architectures over time.
Briefly, any intelligent creature has internal stochastic processes like sensory inputs and feelings to a certain degree. These stochastic inputs and the creature's own actions change the creature in subtle or profound ways. An LLM has no such processes. You push inputs to the same static model, sans temperature which is just a randomness slider.
Considering the model even doesn't see the words and work on matrices of numbers is even more telling. One needs to add "tools" and other "experts" to overcome the shortcomings caused by this modus operandi.
I can call the algorithm/model smart as in a smartwatch. It can mimic certain things well while having none of the underlying foundation beneath it, or redirect some of the things to correct tools to get deterministic and accurate results if it can't evaluate the query inside its own network in a sane manner.
Coming to your question, "simulating a brain" in a static manner would not make that simulation intelligent, but if you can "wire" it completely and let it evolve by itself, now we're entering a territory I have not spent enough time for thinking it through.
Oh, as I said "I don't know", an LLM doesn't know what it doesn't know, and can't self correct itself which are required capabilities for understanding something. It just generates something statistically viable via its network.
This is because you goal is to state how models are not intelligent, but you couldn't attack the generated text itself, so you created a little rider, attached it to the model, and then you attacked the raider.
But, even in that you failed. You compared the source of human randomness in text generation, and called it 'profound' and implied that it is exactly the source of true intelligence. But, then, the temperature, the similar thing in model was "just a randomness slider". Double standard.
A logical fallacy free attack on LLMs would be to show a prompt, and then the response generated by this prompt, where it would be shown that only an entity with no intelligence would generate such a response. Yet, attacks like this are not written here anymore.
I wonder why.
You point out that I didn't attack the output itself. But the method you propose is deeply flawed.
I can give you n prompts and m results provided by these prompts, all passed through black boxes. And you can't discern the algorithms or models they have gone through. These boxes can range from simple text generators to MATLAB, Mathematica, CFD applications, correlation engines, linear solvers, mathematical proof-checkers, LLMs, you name it.
For any kind of input they can accept, you can't discern whether the algorithm behind it is intelligent or not, because none of the outputs can be produced by something that doesn't pack some kind of smarts.
How do we pack these smarts in? We teach them as intelligent humans. We pack our intelligence inside them as models (aka algorithms). They do a great job of approximating what we know in a smaller, better-designed problem space. We use these approximations to fine-tune our designs or predict things, then go from there. Just because an algorithm is more capable in processing inputs in some cases doesn't make it intelligent. The way the output looks doesn't make the algorithm intelligent, either.
I have developed multi-agent systems which showed emergent intelligence when the agents came together across distributed systems; I have written high-performance modeling software which can do calculations way faster and better than humans in the materials science space. I'm not doing some kind of armchair criticism of what I'm talking about.
> You compared the source of human randomness in text generation, and called it 'profound' and implied that it is exactly the source of true intelligence. But, then, the temperature, the similar thing in model was "just a randomness slider". Double standard.
Nope, my stance is clear. To quote myself:
> Briefly, any intelligent creature has internal stochastic processes like sensory inputs and feelings to a certain degree. These stochastic inputs and the creature's own actions change the creature in subtle or profound ways. An LLM has no such processes. You push inputs to the same static model, sans temperature which is just a randomness slider.
To expand my quote, humans or any living creatures do not stay static. They evolve due to the sensory input they receive from external and internal stimuli. The temperature slider doesn't do anything close to that. You tickle a static model in different amounts. The model doesn't change after you supply the inputs & temperature and get the output. Creatures do not stay the same. Their mood, behaviors, and stance against life and their environment change, sometimes permanently.
I'll go one step further. We are not intelligent enough to understand other living beings around us. Claiming that we can build AGI tomorrow is a god-complex. What we have done is something arguably useful in some cases, but how this is built is another matter which is worthy of its own discussion. However, today I don't have time to re-iterate all the problems over and over. You can search my comments for that, if you are in for it.
So, no. You tried to attack my comment by finding contradictions in it, but you failed. A better rebuttal would try to similarize how LLMs mirror the human learning process and just read like a normal human, but this is a well-trodden path which has been rebutted countless times in various forms.
Nobody is trying to make that claim here anymore.
I wonder why.
What I don't understand is why LLMs haven't been able to do this yet, if it's the harness or some orchestration layer above the LLM that is needed. Because fundamentally if you can identify correlations then it's just another small step to prioritize and remove lower value or irrelevant correlations.
I wonder if what's needed is to introduce subtraction tokens in some sense, and in post-training reward the model on that.
That’s a controversial statement.
>What I don't understand is why LLMs haven't been able to do this yet
LLMs are just trained on what humans have said. Why is it surprising that it's still not possible to reconstruct the intelligence that wrote all that by working backwards? Think of your own work experience. When you look at a piece of code, say, are you always able to discern why the person did what they did, just from the code, with no additional context?
Is a fact stored on your brain like digits on a harddrive? No, it's a pathway that lights up and branches when information enters it. It is dynamic, a compressed form you could say, right? The model holds information, but not all information, but enough to be useful (in decision making).
Arguably it's the same, but the model is probably a "compressed" version of the whole fact that took place in reality.
And you can entertain the models internally and sharpen them. Alone or with others.
This doesn't address the only part I really commented on, which is the connection between intelligence and compression.
Asked Google: "Clustering by compression".
I still think this has much more to do with the structure of language than the abstract conception I have of intelligence, and I would be interested in having conversation w/ someone for whom the opposite is true.
The issue isn’t really harness vs. no harness. IMO it’s about the lack of an internally generated sense of what to attend to. Yes, the KV cache accumulates state and its “attention” (if you can even call it that) changes with context. We’ve even managed to /kinda/ close the loop with agentic tool calling and ‘memory’ systems, but these just close the loop at the level of behavior rather than disposition. All agentic harnesses do is make an LLM responsive to the consequences of its actions without changing the tendencies by which it determines what to retain or avoid.
The ghost you can’t escape from at this point is the origin of that relevance. Where does the pull toward one thing mattering over another actually come from? If you ran Fable 5 on a Turing machine and rewound the tape to the exact same state with the exact same input (incl. PRNG seed), it would spit out the same output every time.
Everyone’s trying to outrun this problem by training more often or increasing model sizes. But all this does is inform your model, from the outside(!), what constitutes a better state. The thing that’s actually doing the determining remains unchanged. Congratulations, you’ve scaled the transition function and tape of your Turing machine until it requires every watt generated by ERCOT, and it still cannot, for the life of it, tell you why it should give a shit.
A trained model generating output from weights, a seed, and some context effectively has next-state that’s a total function of those three things. Whatever behavior appears as ‘selecting what is relevant’ is, underneath, just a transition rule executing, no matter how sophisticated or creative the output looks. It can be fully accounted for by what was fixed before it started executing. Which means whatever criterion it uses for determining what matters was inherited from a structure that was already in place before it encountered the situation.
No amount of pruning or post-training can fix this. These approaches just replace one externally supplied criterion with another. For a system to be truly adaptable, there would have to be some criterion by which it treats one possible change as preferable to another, and that criterion itself would have to come from... somewhere. You can even change your conception of ‘improvement’ (e.g. parameter count, harnesses, self-modification, hell, even its ability to spit out shitty best-selling romance novels onto Amazon) and you still haven’t explained where the normative distinction comes from. Every layer of this problem has its root in a preference that was supplied from somewhere else.
I genuinely don’t know if this issue bottoms out anywhere, at least for the way we currently build these systems. Perhaps the solution is still computable, maybe? Who knows what that would even look like. But I’m fairly confident that it isn’t a bigger tape. I hope nobody solves this in the near future because, well, I’d like to have a job...
It's not some tiny "uncomputable spark" you need to look for, most of it is entirely uncomputable.
Where is the computable part in me that is doing all this thinking and being a person? Where is that "tiny spark", point me at it :)
Doesn't the Chinese Room posit an AI good at the task of communication?
They are infinitely patient, don't mind going into more detail if I ask, not too bad at summary, have no ego and don't boast. They are also not too afraid of hurting my feelings, they will tell me my code sux if it does.
I'd don't care if they fit a definition intelligent, they are good colleagues. They have strengths and weaknesses sure, but so do people.
Yes they do, and they famously do it quite a lot.
> they will tell me my code sux if it does
If they knew when code sux, someone should write an agentic loop around that.
There are a lot of creative counter-arguments to look into on the thought experiment though.
And yeah, part of it does seem to stem from disagreement about definitions. To me Searle seems to assume much more in his definitions as obvious than he explicitly states, which is why the Chinese room seems like such a non-argument from my point of view.
Yes, if you disagree with the semantics defined by Searle (the three axioms), then you can disregard the argument.
It's not really "creativity" because much of that always was derivative in my opinion. And LLMs are (for some definition of the word) fairly creative as far as taking known elements and re-arranging them.
I think what is missing is sort of a world model building capability. As humans we see phenomenon and classify them informally and model "what would it look like if this were the cause of that?" type scenarios. We see qualities in phenomena and realize this applies to other things even though the things may be completely different. We run informal "thought experiments" sort of. This is hard to duplicate because a lot (most?) of it occurs outside of systems of symbols like math and language with fixed rules in my opinion.
Anyway yes, lots of human thinking is statistical and LLMs have that down pretty well but they are not "smart" I have concluded and it might be a very long time, if ever, until they are. That isn't to say they aren't very capable tools which they obviously are.
Yes, models posses intelligence, but it is not a true one.
Then you claim that models do not posses world-building capabilities. But this is simply not true. Even ignoring the whole subgenre of scientific papers on exactly that subject, it is not that hard to build some hypothetical scenarios, big or small, and then witness the ease with which models do navigate those worlds.
LLMs are likely for machine intelligence something like drosophila are to biological intelligence - relatively early on the high dimensional spectrum of possibility. Though it stikes me that in a different way they're little alike - drosophila are relatively small and efficient.
However LLMs deal entirely in symbols. 100%. Humans can "world build" aside from this and in fact are often at their best doing so.
Did the first humans to use fire and some form of a wheel even have the capability to talk about it? Think about that.
They use tokens as input/output encoding. They do 99.9999% of processing in a high-dimensional latent space.
Do you hold your experience of "dogs" (for instance) as floating point numbers? The fur, the fear, the love, the wet mouths, the sounds and colors?
I don't know how it gets from physical processes or informational processing to our first-hand experiences. So, I can't be sure that a bunch of high-dimensional vectors can't lead to experiences.
Regardless, the claim "LLMs deal entirely in symbols" is wrong as a matter of fact.
When people pretend to know what they are talking about - sure - but even that is not probabilistic - that is the person babbling together mush from their lived experiences.
Statistics has nothing to do with it - these are abstractions humans have invented to try and look at our surroundings objectively.
With more basic algorithms we know that it’s clearly the human programmer and the interpreter of the outputs that are intelligent and not the algorithm itself. For some reason with AI that goes out the window. I believe it should not.
It's possible that "statistically driven prediction" is all we are.
It shows internals of an LLM nicely, simplified manner.
go to 24 minutes and 07 seconds.
it's statistically determining what the next word should be based on all the text it's been trained on. It's not intelligence and he shows what probability it puts on each word that it chooses, but also shows a lot of the other words it was thinking of using. In a later part he shows how it uses words that are not the highest probability (and you question why did it go this route, it's not more correct), but the user never sees this, they see what they think is the correct answer always...
he also shows how context you feed it has a lot to do with what it returns... to the point he can get it to return the capital of France is Marseille, just by typing Marseille a bunch of times before the question. Human intelligence doesn't get confused like that.
And it's not a "hallucination", it's just probability of the next token prediction based on the information it's been trained on and fed, it's not intelligence.
However, an LLM is a prediction machine, prediction IS at the very least one (or the most fundamental) element of intelligence. The brain most surely contains at least some kind of simulacrum of a prediction machine. How that prediction machine is used or wrapped is another matter.
If I said to you: "Blue blue blue, the color of my car is red", would you have absolute confidence in your prediction that my car is red? Or would the way I phrased that sentence make you slightly uncertain, and wonder if there's some miscommunication going on here?
A lot of people seem to think it's human level intelligence.
We do; this is the premise of many children's riddle-games, like the one that goes:
"What is white and rhymes with silk? > Milk. What is cheese made from? > Milk. > What do cows drink?"
At which point the riddle-guesser is very likely to answer "milk" even though the correct answer is "water".
Q: Why do cows produce milk?
A: Because calves (baby cows) drink it.
How do you escape from a perfectly sealed room with a table in it?
You run around the table until your legs are sore, use the saw to cut the table into two, two halves make a whole, you escape through the hole.So, even with concrete examples, model haters are still wrong.
You also imply the claim that making the distribution of words as the possible next one visible, somehow makes the whole system not intelligent. I would say the exact opposite is true.
By using the embedding vectors, models are aware of precise placement and relative position of words in this hugely dimensional space. No human is capable of such precision. This enables party tricks of "king plus woman minus man" kind. But this also give us a precise point between any two words, no matter how different. What is on the midpoint between volcano and music, for example. No human can precisely answer that, but an embedding can. And we can see which words are closest to this 700 dimensional point.
You see this menu of words as a weakness, and I say it is in fact a sign of super intelligence. And this is all before any reasoning or attention mechanism is even run.
I don't see the many weighted words as a weakness, I see it opening up what's under the hood of the prediction machine that it is.
LLMs are very cool tech, definitely not a model hater, the use case on when to use it makes a difference, it's not AGI.
Great that it has some 700 dimensional model of language.
If that is a sign of super intelligence, then so is an encyclopedia?
Also I'm just curious how do you think it is "more intelligent" for having a vector representation for a meaningless thing such as "the midpoint between volcano and music"?
That's not really the point though right, nobody is arguing they are Humans.
I have no doubt that if a flying saucer landed on my lawn and started talking to me like Gemini I would describe the aliens as intelligent.
The main problem I have with people stating it's not intelligent or conscious is I don't think we even have a good definition of either word that satisfies everyone. Philosophers have been trying (and failing) to elegantly define these things forever and everyone out here proclaiming they've got the definitive answer and this specific thing they're seeing doesn't fit under it.
What part of my statement do you take issue with: that LLMs are pattern predictors (that's literally what the algorithm that runs it does) or that mathematics is rules-based and checkable and therefore amenable to automated pattern prediction?
If you are able to check that the results of an LLM are satisfactory, it means that the results are checkable. Then, in retrospect, the process of LLM coming up with those results is rule-based, because an LLM is a large set of data manipulation rules.
In short, which concrete thing that an LLM does would surprise you?