Hacker News Reader

Comment by jshen | Hacker News Reader

> "The LLM AI technology generation is optimized to be fluently conversational and not to be factually correct all the time."

I always find this point a bit odd because humans aren't "optimized" to be correct either.

> "LLMs have no understanding of the underlying reality"

I struggle with this one because I see both sides of it. I was making a prompt the other day and gave a CSV file as an input and told the LLM if could only answer with values from one column and it did exactly as I asked. It's hard for me to see things like that and not believe it has an understanding at some level.

View on HN

>I always find this point a bit odd because humans aren't "optimized" to be correct either.

It depends which subsystem.

The older the system in the body the less likely it is to have a high error rate, otherwise we'd die from cancer at a much higher rate or injure ourselves far more often. Of course this also depends on the definition of 'correct', if the system never changed we'd never evolve.

>> "LLMs have no understanding of the underlying reality"

This statement has always bothered me because it really depends on what you mean by 'underlying reality'. How many layers are we talking about? What does understanding mean? Because at the end of the day, humans don't really understand 'underlying reality, we just have a particular set of input devices we take in information and do some transformations on it... Where is the understanding happening?

I'm talking about all of the cognitive science which seems to point to the fact that we often care more about fitting in with our group that being correct. Group cohesion seems to have been selected for.

Edit: there is also loads of evidence that our brains are wired to make incorrect decisions in many cases. Loss aversion is one example.

Humans have evolved to communicate based on some sort of shared mental model and shared intentionality—human speech generally has some sort of more-or-less coherent semantics by its very nature. While it's hard to reason about what goes on inside the neural network, there are too many examples of LLMs outputting not-even-wrong nonsense to conclude they have a similar internal model, even assuming that having an internal model at all would make sense in this context!

Have you read social media? I see an overwhelming volume of "not-even-wrong" nonsense from humans all the time.

> humans aren't "optimized" to be correct either

Humans at least have a concept of "being correct", even if we don't always set that as our primary goal when communicating.

LLMs don't even have a concept of "being correct". That would require having a concept of an "external world" that text refers to, which LLMs don't have. All they have is the text in their training data.

>LLMs don't even have a concept of "being correct"

They clearly do as plenty research indicates. You've just decided not to accept this. Pure confirmation bias in action.

> They clearly do as plenty research indicates.

I have seen plenty of researchers claim this. What I have not seen is actual support for such a claim.

There's plenty support. You just need to know how to read.

Please link to published papers that prove what you’re positing.

To tack on, I would love to read more research on this topic if there is relevant research to read.

I have not read much compelling evidence though I have seen a lot of researchers conjecturing and some inappropriately reaching (and anthropomorphising the living hell out of ai in the process sometimes).

There’s a lot to learn and I don’t think anyone can keep up with all the papers coming out so let’s keep things positive and work together to learn as much as we can og_kalu. Thanks for letting me tag this comment on the end of thread.

This is HN. We’re all nerds here anyways and nobody can keep up with everything in a field moving as fast at AI is these days. :)

I'd ask this question the other way. What research indicates that humans care primarily, or even to a high degree, about being objectively right versus deluding ourselves into believing we're right regardless of whether we are or not? I've seen a lot of evidence for the latter, but I'm not familiar with any for the former.

Humans do not _know_ the external world either. All humans have is perceptions from media like sounds, images and text! Internally we try to make a model of the world based on those perceptions. Any blanks are filled in by creativity ! Think about it, when we write or speak we start from some internalized model (which is mostly imperfect) and then we build text around it that feels fluent and consistent.

I would argue, human reasoning is conceptually not so much different from LLMs as one might think.

> Humans do not _know_ the external world either. All humans have is perceptions

Wrong. Human beings can act on the world as well as perceive it, and humans have rich internal models of the world that get updated continously as a result of those rich two-way interactions.

> I would argue, human reasoning is conceptually not so much different from LLMs as one might think.

I would argue that this is either an incredible over-estimate of LLMs, or an incredibly impoverished view of humans.

> Human beings can act on the world as well as perceive it

Conceptually this is the same for LLMs.

Input/perceive -> human/LLM -> output/act. You might say the acting part for LLMs is not autonomous but that is only dependent on how much agency is given to LLMs. LLMs outputs are massively used, hence their outputs lead to actions.

> and humans have rich internal models of the world that get updated continously as a result of those rich two-way interactions.

What is different from LLMs: rich, continuous. I would argue these are quantative and qualitative traits that fall outside of what I meant with "conceptually".

> Conceptually this is the same for LLMs.

No, it isn't. LLMs don't look at the results of their actions and update a world model. Humans do.

This is indeed a difference that I overlooked, there is limited feedback on LLMs output, other than a rater primitive survival model (models that aren't popular/correct/cheap/fast enough are not used anymore over time)

I love this point.

One of the things that has been on my mind of late is the idea of how much of what humans call “correct” or “right” is actually socially generated pressures to conform.

Berns and Asch’s various research on social conformity around what we normally would consider objectively correct answers seems to have some potential applications worth testing in AI land. AIs may be much more performant with AI “social inputs” (and of course all the long term memory improvements currently lacking required to use it as a tool). I am very much excited about the network AI and pipeline work (small teams!) currently coming into vogue. Feels like the “right direction” to make incredible progress.

https://www.psychologytoday.com/us/blog/am-i-right/201404/th...

Otherwise agree with your points exactly. Good comment.

> One of the things that has been on my mind of late is the idea of how much of what humans call “correct” or “right” is actually socially generated pressures to conform

And the only reason that is even a thing is that we humans have a concept of actually being right, which we can distinguish from "responding to socially generated pressures to conform".

I'm not sure I'd frame it that way. We seem to make decisions with our subconscious brains and our "rational" brains primarily rationalize a justification for that decision. Our brains are less concerned with being right than they are with justifying our subconscious behavior and fitting in with our tribe. Another example, our brains care far more about achieving and maintaining status than about being right.

We seem barely capable of objective thought at best.

> We seem to make decisions with our subconscious brains and our "rational" brains primarily rationalize a justification for that decision.

There is research that is claimed to show this, but it doesn't. All it actually shows is that the process of making a decision takes time, and involves many different parts of the brain. There is no separate "rationalization" that comes after the decision is made. It's all part of a single process that can't be decomposed into such separate parts.

In any case, all this is irrelevant to what I said. I said we humans have the concept of "actually being right". We don't always live up to that concept, but even making that observation, as you have, shows that we have the concept. If we didn't, we could not even make statements like this one of yours: "our brains care far more about achieving and maintaining status than about being right."

> We seem barely capable of objective thought at best.

I think you have a sadly impoverished view of human thought.

My view is based on the science.

Regarding the research that supports my claim, studies have been done to isolate the different parts of the brain where they tell one part to do something like get water, then they ask the other part of the brain why they got the water and it makes up a reason because it's not aware of what happened in the other part. This is eerily similar to LLMs hallucinating.

It’s certainly true that both genes and our culture, rather than just one or the other, provide an explanation. But it’s not one or the other, but one and the other.

This genetic and cultural evolution working together has shaped our brains to care for others, react to those who try to harm us, and to create moral rules that help us to live together successfully.

I’m not sure why responding to socialization pressures to conform to certain behaviors would preclude intrinsic or genetic morality. We know that babies exhibit intrinsic morality from a very young age, that moral behavior exists in non-human species, and that we have parts of the brain just to deal with moral judgement (though we are really in the infancy of understanding and mapping this from what I’ve read).

That’s not what I meant to imply or what research supports.

Humans aren't optimized to be correct, but they are optimized to survive and reproduce themselves in the real world. That's why we develop an understanding of the real world, while LLM can, at most, develop an understanding of human language and the constructs that human language can create. But without the napping that we have between that language and our experience of the world.

It gets very interesting when we start providing digital input of analog source data in film, photograph, sound and speech.

We recently started providing perception inputs, just like you’re talking about.

Obviously it’s nascent and not exactly a revolutionary statement, but it’s interesting to see the progression towards our experience.

I’m also interested in how the analog computing revolution is going to converge with AI in coming years, as we are making huge inroads into analog computing and that means cheap ubiquitous sensor inputs right in time for AI to start maturing.

LLMs develop a model of the world, experienced through human-generated text. The same as I've never been to China, but I have an internal model of China due to experiencing it through human-generated text.

There is a huge difference between your model and an LLM model in your example. Your model "knows" that China has rivers (of which you have direct experience), coastlines (same), trees, mountains etc. etc. An LLM has no tangible experience of any of the concepts included in the text.

Both my internal model and an LLM are built from sensory data. My model knows what a river is from having experienced it through sensors that an LLM doesn't currently have, but don't expect that limitation to last.

Also, given how many words humans have written describing every part of the experience, LLMs can generate a pretty good understanding of it as-is.

"[humans] are optimized to survive and reproduce ... in the real world."

I think about this difference a lot, and a fair bit of AI/ML attempts to reproduce this process to some degree.

There is a difference between general AI and generative AI. Humans can be creative and create vast fictional worlds with made up technology and magic. They can also optimize for facts and pass a BAR exam.

LLM like have some form of understanding of the world, not sure if its anywhere near comparable to our understanding though. But they are still not general enough to really focus on facts. They can get close but the nature of statistics means there isn't hard checks to truth.

I think the next step logically is to see how the human brain functions. We have a similar model where certain parts of our brain are hopelessly useless by themselves as they are so overly specialized in one thing, but work in harmony to create network effectiveness.

And we only just [mapped it for the first time](https://phys.org/news/2023-10-scientists-generate-single-cel...) so now much of what we are learning can potentially be applied to our “digital twins” in coming years.

I imagine a lot of progress with AI will involve similar networks, with governance providing evolutionary paths and guidance for multiple concurrent goals.

There are also lots of issues with training data and memory limits that make LLMs weak at continuity. Holes or weaknesses in the training data might “come through” in the behavior of hallucination.

Where do you think general AI is going right now that we should be looking at? I’m overwhelmed by how large this field has become in the last five years and am always interested in what others know.

Reply on news.ycombinator.com