The Social Edge of Intelligence: Individual Gain, Collective Loss
theideasletter.org
theideasletter.org
It assumes that the outputs are lacking because of a limit of ability.
I think there is a strong case to make that many of their limitations come from them doing what we have told them to do. Hallucinations are the stand out example of this. If you train it to give answers to questions, it will answer questions, but it might have to make up the answer to do so. This isn't not knowing that it does not know. This is doing the task given to it regardless of whether it knows or not.
If you were given the task of writing the script for a TV show with the criteria that it not offend any people whatsoever. You are told to make something that is as likeable as you can make it without anyone not-liking it at all. The options for what you can do are reduced to something that is okay-ish but rather bland.
That's what AI is giving us. OK but rather bland. It's giving it to us because that's what we've told it we want.
Are you asserting that an LLM could be NOT trained to answer when it knows it doesn’t know the answer, or if that’s not possible be trained to NOT answer when it knows it doesn’t know the answer?
If so, I would believe your thinking, but for some reason I have not yet seen a single LLM that behaves with that kind of self-knowledge.
I don't say this, because I know how, but because I see no reason why we will be unable to crack that problem. If our brains can do it, so will AI one day.
This is not a problem in the ability of the system, it is a problem of how to construct training for such a task.
To provide training examples where it answers it does not know the answer only when it does not know the answer. You need training examples where it says it doesn't know when it does not contain that knowledge, but it provides an answer when it does know the answer.
To create such an example, you need to know in advance what the model knows and what the model does not know. You can't just have a database of facts that it knows, because you also need to count things that it can readily infer.
Any model that can reliably give the sum of any two 10 digit integers should be able to answer so. You can't list every possible number that a model knows how to add. That is just the tiniest subset of the task you would have to do because you have to determine every inferrable fact, not just integers. Adding to the problem is that training on questions like this can add to the knowledge base to the model either from the question itself or by inductively figuring out the answer based upon the combination of the question and the fact that it was not expected to know the answer.
A completely different training system would have to be implemented. There is research on categorising patterns of activations that can determine a form of 'mental state' of a model. A dynamic training approach where the answer that the model is expected-to-give/rewarded-for-giving is partially dependent on the models own state could be achieved through this mechanism.
Do LLMs even have any kind of internal model of what they know or don't know? My understanding is that they don't.
There are things like https://arxiv.org/abs/2410.22071v2
But again things are not quite so simple. Detecting hallucinations might yield representations where it knew the answer but elected to hallucinate anyway because of some other obscure interaction.
Anthropics work on autoencoding activations for analysis has yielded a lot of information about the inner semantic information on models. I haven't seen a lot on bounds of knowledge there, but I wonder if that's something they hold back for competitive advantage.
I didn't mind that there was a typo.
~ intelligence
I see students obsess every day over their SAT scores, which to some is a measure of individual intelligence. But what SAT score would a pair of students working together on a single test get? Or a dozen students working together? Would it be higher or lower? What sort of strategies would maximize their ability to collaborate? What would be the effect of giving/removing access to a calculator on a student's score? Access to scratch paper? Access to textbooks? Access to a dictionary? Access to unlimited time?
If we want to claim to understand intelligence, these are the sort of questions we should be able to answer. Can we?
At that point, we can only define it as something that causes this observation. And that is not very useful.
Where as the other answer is simply "I can't say, but it's not [this]" and that isn't useful at all. It's simply opinion, which is literally the worst definition around. Personally, I don't define intelligence as "whatever qsera acknowledges as intelligence with no qualifying context"...
But here you go, https://news.ycombinator.com/item?id=47918103
Same for my dog.
Same for entire cultures, including adult humans (anumeric cultures).
Arguably, they're constantly doing calculus, because they can walk, throw, catch, etc - but none of them can count.
---
That said, to the point of my original comment - forcing a definition was absolutely useful, because it allows a more detailed and interesting conversation than "qsera doesn't believe an LLM is intelligent".
I agree that one is a poor definition, though. I also don't really think it rules out LLMs.
Is a robot that can walk, throw and catch is intelligent?
Last day there was a robot that can play ping pong? Is that intelligent?
We also see robots excel at chess and go, is that not intelligent? Why not? Given that's the mental opposite of the physical examples you're happy to discard (side note: I've done real world robotics development, I'd call walking/catching/throwing considerably more difficult activities than chess/go, and much closer to intelligence).
We see LLMs absolutely blow through the turing test, and that was literally the philosophical "gold standard" for machines that exhibit human-like intelligence for like 70 years.
So I really don't think I'm "getting it all very wrong" - I think this is a fundamental question that you're basically failing to honestly engage with because you've already made up your mind.
So again - define intelligence?
huh? In the comment above you said you would argue that it is. So is that not your argument?
>I'd call walking/catching/throwing considerably more difficult activities than chess/go, and much closer to intelligence
This is an even bizarre idea. I get that those things are harder to implement in a robot since it requires extremely fine sensors and motor control. But that making it closer to intelligence? That does not make a lot of sense..
Just because we cannot exactly define intelligence does not mean that you can call any arbitrary thing intelligent!
Why? It requires understanding a complex sense of self: where am I? How can I move? how do I fit within the environment? How can I change the environment?
Understanding, reasoning, self-awareness, and planning are all considered core aspects of intelligence, and those are all on display during navigation and locomotion.
I'd suggest those are all much closer to intelligence than the rules of chess. More complicated, too.
> Just because we cannot exactly define intelligence does not mean that you can call any arbitrary thing intelligent!
And just because intelligence appears alien in nature to you, doesn't mean it's not there... An ant only has ~250,000 neurons (far less than an LLM) are ants not intelligent? Not at all?
So if you define intelligence simply as "human", then sure - LLMs aren't that. But I also think that's an uninteresting and banal conclusion.
---
So my core point remains - define intelligent. I think it's surprisingly hard to do in a way that rules out LLMs, but doesn't also rule out large categories of things we do probably agree are intelligent.
These does not require any general intelligence or understanding and can be expressed as a bunch of fixed rules.
>I think it's surprisingly hard to do in a way that rules out LLMs,
May be. But we sure can say that LLMs are not "reasoning". Which is what I have done in the linked comment.
Wait, I meant light, yeah, photons! It's photons all the way down! And what are photons you ask? shit, no more questions. Got to go.
Statistically, an LLM finds a better optimal solution than one individual. But LLM solutions are more similar to each other than individuals’, so among many individuals, a few find a better optimal solution than the LLMs.
Hence why I think we need AI that responds more uniquely. Whether that be fine-tuned local models, one LLM that’s more influenced by its prompt and well-supports extremely long prompts, AI tools for humans, or something else. I’d love to see a local AI with continuous learning that generates complex UX to better interact with the user (more than prompts), but that may be far-fetched…
Very "intelligent" people can use those smarts to justify or rationalize all kinds of crazy stupid decisions.
Or were you looking for a different definition?
I fully expect our future to involve PhD factories where doctorates label AI output for the most competitive rates possible.
The majority of us will have to contend with an information environment that is polluted and overrun.
I’ll argue with that the internet pre social media was the “healthiest” in terms of our digital commons.
What can you do with a whole team comprised only of Expert Beginners?
Edit: for "not in the training data" yes, humans generally can't know what they can't know.
Its necessary if you want to replace such a worker, at least to have the corporate culture and knowledge in its training set.
Internet started it, hopefully LLMs will finish it.
Now with LLMs lowering the average through cognitive offloading and skill atrophy, prepare for it to get a whole lot worse.
If I read thousands of books that explain the details of another civilization in another galaxy, very thoroughly and consistently, but it it just happens to be all made up - did I gain knowledge? More importantly, does what I have in my brain now flip from being fiction to being knowledge if that civilization flipped from not existing to existing? How so, if nothing in my brain, or how I live out the rest of my life, changes in the least, if not a single atom in this galaxy changes (let's ignore that gravity has infinite reach and all that, for the sake of argument)?
If yes, how? What in your definition of knowledge makes that possible?
sounds a lot like math - made up entities that very thoroughly and consistently fit together.
Example it can easily do moderately advanced calculus, that is way better than the average human.
But this debate has parallels with self-driving cars; even if the numbers say that self-driving cars are not perfect but safer than human drivers, anything but perfection will be considered broken or outright illegal.
What's stopping us is that AI works by manipulating tokens and language but has no connection to reality as it exists. Einstein famously conceived of special relativity by imagining what it would be like to fly alongside a light wave[1]. This is a process that integrates spatial reasoning and imagination informed by living in the real universe where you can see objects moving or waves propagating in a pond. The language only comes later as a means of communicating these intuitions to others.
[1] https://sites.pitt.edu/~jdnorton/Goodies/Chasing_the_light/
Of course, the models are not intelligent. Their generated output reflects the statistical average. And in averaging more and more, you lose a lot of information.
The author argues that overreliance on AI will degrade the overall intelligence of human society, creating a negative feedback loop where future models train on increasingly degraded human data. I agree with this perspective to some extent. However, to definitively claim that human intelligence will only decline is overly simplistic. Rather, we might be about to witness a different facet—or the flip side—of what we have traditionally defined as intelligence.
Socrates once argued that the invention of writing would degrade the essence of human thought and memory. It is true that our capacity for raw memorization declined, but the act of recording enabled knowledge to be transmitted across generations. Couldn't LLMs represent a similar evolutionary trajectory?
It is undeniably true that LLMs atrophy certain cognitive muscles. However, I believe they catalyze development in other areas. In modern society, human discovery and knowledge are effectively monopolized by specific cliques. Without access to prestigious Western journals or incumbent tech giants, the barrier to entry is immense. The open-source community is no exception. For non-native English speakers, breaking into the open-source culture to access shared knowledge is notoriously difficult. But now, by spending a few dollars on an LLM, I can access the collective knowledge of that open-source ecosystem, translated seamlessly into my native language.
There is an old adage in the Korean Windows community: 'Linux is open, but it is not free.' And it’s true. To use Linux, you had to memorize arcane commands, and due to the lack of proper Korean documentation, the learning curve was vastly steeper than Windows. That very learning curve acted as a gatekeeping wall. LLMs explicitly dismantle that wall.
But this dismantling is a two-way street, and it exposes a fatal flaw in the author’s reliance on Shumailov’s 'Model Collapse' theory. The author claims AI compresses the tails of the data distribution, erasing minority viewpoints. What this ignores is that LLMs act as a conduit for cognitive diversity from the non-Western periphery. When a developer in South Korea or Brazil uses an LLM to translate their culturally embedded logic and problem-solving approaches into fluent English, they are injecting entirely new cognitive patterns into the global corpus. This does not compress the tails of the distribution; it actively thickens and extends them by capturing the 'social mind' of populations previously locked out of the internet's primary, English-dominated datasets.
Furthermore, LLMs function as a tool to re-evaluate things we've historically taken for granted—especially in areas that are too complexly intertwined, socio-politically loaded, or vast for the human mind to fully map. Take DeepMind's AlphaDev discovering a faster sorting algorithm as an example; it was a breakthrough achieved precisely because it reasoned from an alien, non-human perspective.
Human learning is fundamentally bottlenecked by environment and bias. Anyone who has interacted with academia knows it is riddled with pervasive prejudices and systemic inefficiencies. In South Korea, for instance, there is an entrenched bias that only researchers with US pedigrees are legitimate, and only papers in specific Western journals matter. This prejudice has prematurely killed countless promising research initiatives. It makes you wonder if the metrics we have long held up as 'superior' or 'correct' are actually deeply flawed. Modern society is too complex for the 'lone genius' model; paradigm shifts now require the intertwined research of multiple collectives. Yet, during this process, political interests often cause dominant groups to gatekeep and exclude others, completely regardless of scientific efficiency. In this context, an AI that lacks our inherent socio-political biases and optimizes purely based on probabilities can actually drive true breakthroughs.
Given all this, the absolute claim that AI unconditionally degrades human intelligence feels flawed. I seriously question whether the 'total sum' of human intelligence is actually experiencing a meaningful decline. Before making such claims, we desperately need to define what 'intelligence' actually means in this new context. The fatal flaw in current AI discourse is the complete lack of nuance—there is no middle ground. Everything is framed as a binary: either purely utopian or purely apocalyptic.
Speaking from personal experience, my cognitive muscle for writing raw code has atrophied because of AI. However, as a non-native English speaker, I used to struggle immensely with naming conventions. Now, my variable naming and overall architectural design capabilities have vastly improved. Conversely, I acutely feel my skills in manual memory layout management and granular code implementation degrading. The trade-off point will be wildly different for every individual.
Whenever I read doom-saying articles like the author's, I can't shake the feeling that they are simply projecting their own subjective anxieties and trying to pass them off as a universal conclusion
The decline brought by writing was not only in memorization. If in previous ages to understand something was to study deeply, eventually the definition may shift to having asked ChatGPT about it and skimmed through the response. This is discussed thorougly in Technopoly, which I'm still reading.
You should also consider that besides the effects of AI adoption in highly technical and scholarly people it will also affect a majority of average workers who may be more vulnerable to atrophy than others.
Not only that, but eventually AI will be native and people's perspective and usage will not be affected by previous generation habits. If people hardly bother to write their own emails, comments or essays then how will the AI-native generation approach that?
Although you make very solid points, I've been leaning to think that the AI effect on society will be shaped by the average user, not users such as yourself and the colleagues you observe in which case the doom-saying starts to make better sense.
We are all shaped by the cultural and social environments we come from, so my perspective is based on the context I’ve experienced. That difference itself is what I find interesting. These kinds of cultural and intellectual frictions make me think more deeply.
I also agree that asking good questions requires some level of foundational knowledge. But at the same time, it’s difficult to ask good questions without that foundation. As you mentioned, more experienced users will likely benefit more from AI.
Still, I think humans tend to find ways to adapt to these kinds of challenges, so it’s possible that new forms of education will emerge. For example, ways of learning how to use cognitive tools like LLMs effectively.
That said, I admit this is an optimistic view. In reality, as you pointed out, the gap will likely widen. People from more educated backgrounds will be better positioned to use these tools effectively, while others may fall further behind.
I agree with your perspective, and I appreciate your thoughts. It’s interesting to see so many different viewpoints—these differences are what make discussions like this engaging.
Have a great day.
The reason being, when taught how to effectively disagree all these counterfactual concepts that AI loses manifest, they are logically necessary. But if people are not taught how to explore the landscape of ideas, they become "fascists for the common" and literally create the hellscape civilization we are all trapped within.
This in itself is negative, but the ramifications are profound: the landscape of ideas is never realized by a material percentage of the students. And those who could have contributed worthwhile insights have been taught to not contribute.
You mean like teachers? Teachers bullies kids that answers wrong in class, so if you try to think for yourself rather than answer according to the book the teacher often makes fun of you and the rest of the kids laugh at you for not saying the right thing.
Just making teachers not do that would probably go a long way to changing the culture. Kids feeling fear of being asked a question in class is a strong signal we are doing something wrong.
Generating new training data from existing data will only generate patterns that already exist in the training data. It might help LLMs to capture it if has not already. But it can never generate novel patterns.
For example, Imagine some kind of neural network architecture that can do OCR. You might be able to generate variations of the letters it already know using some technique and use it to better train the recognition of already known letters.
But it would never be able to generate letters that it does not know.
Once again LLMs will have to be bound to a source of entropy or feedback of some sort as a limit. Sure you might be able to throw terawatts of cycles at say music production but without examples of what people already like or test audiences you cannot answer the question of whether it is any good.