Comment by thomastjeffery | Hacker News Reader

Comment by thomastjeffery | Hacker News Reader

Parent Full thread

I'm not convinced. The entire premise here is selection bias at its worst.

Here are the steps to create an LLM:

1. Tokenize the training corpus

2. Train the model using the tokens

3. Critique the result

4. Repeat steps 2 & 3 until satisfied

We skipped the most important step!

0. Write and curate the training corpus

There is an intense focus on the training steps, but that is entirely ignorant of the elephant in the room: the original training data itself. Where did it come from? What does it mean? What is it for? What does it know?

We don't train models on noise. We train them on coherent data. The very act of writing is where intelligence lies. A model can only present that intelligence. It cannot create it, and it is not aware of it. There is no objective-thought function. If there was, then the LLM could simply tell us about it.

A model is like a coloring book: it doesn't matter how incoherently you stumble around, as long as you stay between the lines. The picture was not invented by coloring it: it was invented by drawing the lines.

> A model can only present that intelligence. It cannot create it, and it is not aware of it. There is no objective-thought function.

That's just like, your opinion man.

I see both sides of this argument make statements like this that are clearly opinion and not backed up by facts. The truth is for now, we can't say yay or nay on whether the model has reasoning abilities, creative abilities, or a coherent internal world view. It's still a debate which is why there are academics like those in the OP trying to come up with theories that make predictions and then running experiments. It's still early days.

If you were to ask my opinion, I'd say these current models are probably closer to the stochastic parrot than an AGI, but there is enough there to start wondering, especially if we have years of improvements coming down the pipeline.

These aren't easy questions, and anyone espousing certainty now is doing themselves and others a disservice.

Not all of it is an opinion.

The LLM does not have the ability to tell you what it knows and what it doesn't know about a subject. In order to get information out of it, you need to present it with a pattern that matches with something it was trained on.

An example, is that an LLM can tell you how many programming languages it knows, how much it knows about them (which one it understands best), and how many it doesn't know. If you ask it a question like "how to do X in python", it will have been trained on something that roughly matches that and give a response, but if you ask "what things do you know how to do in python", and it simply is not capable of giving a response.

I'm not sure that this means it doesn't have some form of understanding of things, and I do believe that AGI is possible to create in general with sufficient technology, and I do realize that our brains may work in a similar way, but I think LLMs are missing something when it comes to understanding and creating that we have not yet figured out.

well... what we have an an LLM doesn't is state and feedback and restraint, a sense of time.

We have an internal state of some kind, an LLM cheats having a state by having every input be the entire conversation + one new input. making the context grow overtime making that state unsustainable.

The other thing an LLM is missing is proper feedback, it receives text chunk prompts and output text chunks. since it's output will never 'change the llm's internal state', it can not have a feedback loop. It can mimic one with the trick of refeeding messages as context, but that's untenable after a certain size input.

Another is restraint, we human's don't have to respond to everything. we can choose to ignore input entirely, leave people on read, etc... LLM's respond to each input regardless of it is appropriate, and they only respond to stimuli (maybe we do as well, just subtle real world effects).

last is a sense of time, which I don't see how the LLM can manage at all. It has no clock.

Give an LLM a rich state (okay, maybe rename it too), have it be 'always on', give it a camera and microphone feed for inputs, give it a robot arm to control (pass the output of that back into the model for feedback), give it a second brain to decided to talk or not, and give it a clock too... and what comes out would look leagues better than current AI.

I think that something - or at least one of them - might be objectivity.

An LLM has all of the what and none of the why.

> That's just like, your opinion man.

Yeah, that's why I literally said so. You are welcome to disagree with me.

My opinion isn't the whole point I intended to make here. I think there is too much focus on "wow look at the emergent behavior of this LLM!", and not enough focus on "where could this specific behavior come from, and why?" The second question may hold the key to both understanding what we have, and making something better.

Do you understand how these models work under the hood?

It's literally just picking the next most likely word, based on some fancy math. There is nothing that can create new insights or anything approaching intelligence.

I think this critique would only be fair if we more completely understood how we ourselves work under the hood.

The "fancy math" is a bunch of non-linear functions that encode something derived from series of tokens. I believe we're still exploring what the derived vector spaces "mean" for LLMs, with research inspecting smaller and simplified models to try to make sense of it all.

It's clear that LLMs can "remix" in interesting ways. "ChatGPT, write me a Shakespearean-style sonnet about how LLMs work." (https://pastebin.com/FwzqWJ5W) It's not clear to me that our creativity and intelligence is not the same or a similar process.

Beyond that, I'm not convinced there are definitions of meaning, knowledge, creativity, intelligence, etc, that are both useful and don't refer solely to the output. If I need a solution and none exists yet and something creates it, I will call that creative irrespective of process.

But I don't know anything about anything. :)

Your argument is centered on a lack of understanding.

How could anyone possibly argue that God and Heaven are fiction? No one can prove they don't exist! I prefer to approach such assertions with skepticism. There is no reason for me to believe the attestation is true, so I may as well believe it to be false.

People remix expressions all the time. Any good training corpus will contain countless examples. If a statistical model can recognize the general semantic patterns that are present in human-remixed expressions, then it should be able to repeat them. So that's a thing that both humans and LLMs can do. Does that mean that LLMs are interesting and useful? Yes! Does it mean we are equivalent? I don't think so.

If I could point out specifically what an LLM can't do, then I would be 99% of the way toward building something that can.

Yes, I understand how these models work under the hood as well as any lay person can. I worked through Andrej's video series and built my own toy model.

Your argument is missing the whole concept of emergence, which is what is at question here. How does emergence come about, under what circumstances, and what is possible in those circumstance? It's an open question.

Following your logic, one could argue that physics is basic and follows well understood rules. We can do some fancy math to predict what will happen in a given particle interaction. How would these basic particles ever give rise to life and minds with reasoning ability?

Well you and I are an existence proof that basic rules when followed on massive systems can give rise to emergent behavior.

Can you explain emergence then please? Because the way you're using it, and the way people tend to use it to present this argument is as a stand in for magic and another way of saying "if we don't understand it, then anything is possible, man!".

We can probably simplify it. If we have a system and move one level up it becomes a completely different thing. For example: one molecule in space has some properties like mass, velocity, position. But move a level up and you get pressure, flow, temperature, etc.. It's a completely different thing.

The same with models. On low level there are weights, gradients, matrices. But move up and (wow!) It's nothing like it. It's coherent text or actions. Commands and responses. Put together and you have image recognition and generation from description. It's something new with it's own laws and properties. Just a bit up and we'll get coherent behavior and social creatures. For them we need 'generic' component. It's like soul for life. And LLMs could be it. At least they are the best candidates today.

Consciousness may not be emergent, it could be like the panpsychists argue. Or other forms of consciousness dualism.

My argument is an attempt to explain where emergence itself comes from. Prove me right or wrong, and you have answered your question. Until then, as you said, the question is open. Why should I pretend otherwise, and claim that the LLM is definitely intelligent? Such a position is no more than vanity.

Following your logic, the million monkeys hammering randomly on typewriters that eventually produce the full works of Shakespeare, did so not through pure randomness, but by actually understanding the plight of Romeo and Juliet and the motivation behind Hamlet.

Edit since I can't reply: "I don't like your answer, therefore you aren't engaging in good faith." lol

> the million monkeys hammering randomly on typewriters that eventually produce the full works of Shakespeare, did so not through pure randomness, but by actually understanding the plight of Romeo and Juliet and the motivation behind Hamlet

The funny thing 'yes'. Million monkeys cannot do it by randomly typing. They cannot even randomly type at all. Each monkey is intelligent creature, they will follow some patterns. If they did manage to produce Shakspeare they definitely understood something.

Today it's easy to programmatically simulate 1m monkeys typing randomly 10 keys per second. Try it ;)

Then it's the thought experiment that needs re-framing. I'll give it a shot:

At nearly every point in the universe, there is a noise: Cosmic Microwave Background Radiation.

What if you listened to the CMBR at a specific point in space, converted that signal into Unicode text, and it just happened to write Romeo and Juliet? Is that intelligence? No. That's selection bias.

So why is ChatGPT able to quote Shakespeare when asked? Because the entire text of the play is in its training corpus, and the infinite permutations of nonsensical CMBR noise is not.

I am aware of the thought experiment you are referring to, but if I witnessed a group of monkeys write the full works of Shakespeare in my lifetime then yes, I would actually consider it something worthy of investigating.

However I can see you are not engaging in this conversation in good faith so I don't think there is anything else to discus with you.

> However I can see you are not engaging in this conversation in good faith so I don't think there is anything else to discus with you.

From my perspective, it seems pretty clear that you are the one arguing in bad faith. People are allowed to disagree with you. This person presented a thoughtful critique of your argument. It made plenty of sense to me.

> However I can see you are not engaging in this conversation in good faith...

What if this "categorize and ignore/forget" is causing humans to miss something extremely important in their midst? Bad faith is certainly possible here, but also possible is that this person genuinely believes what they say.

I find the "it's just picking the next most likely word" argument to be not very convincing. If anything it makes me wonder if that's not all there is to my own intelligence.

Hell, try writing a paragraph backwards and tell me that you're not "just picking the next most likely word" as well.

My saddest observation is that I don't even have control over what comes out next, especially when talking/typing fast. "I" observe myself speaking, more than anything. Sometimes, as I'm talking, I think "that's a good point!" because I didn't consciously make it, it was just part of this "magic stream" that I'm listening to as much as the person I'm speaking to is listening. I also find letting that stream flow is when I'm at my sharpest. Same with problem solving. I point my intention at something and answers come out, through "intuition" and "creativity" which are almost entirely autonomous/subconscious.

This makes me fear getting old, where this magic machine probably fades, almost certainly leaving me pointing my intention, and starting up sentences, without much happening after. I'm afraid I'll be left with my conscious self, which I don't find as entertaining.

It's able to perform that task because it has been fed examples. Essentially, it's the examples themselves that know how to do the task.

"Emergence occurs when a complex entity has properties or behaviors that its parts do not have on their own, and emerge only when they interact in a wider whole."

https://en.wikipedia.org/wiki/Emergence

That doesn't sound too dissimilar from what human brains do, if you replace "math" by "spiking neural network"

Not a whole lot different than what I do as a human. I can do it a lot faster, though, allowing more exploration into the second, third, and fourth most likely word at the same time. Nobody wants to wait years for ChatGPT to finally reply, so shortcuts have to be taken.

I have never discovered any new insights, but it is apparent that the single most important attribute in discovering a new insight is time spent focused on finding a new insight. Undoubtedly, insight simply comes from executing on even more permutations than regular Joes like me take the time for until the right combination finally lands.

"The statement that P=NP is [...]" (true or false)

The article uses the concept of irony and the word "because" as examples of fairly sophisticated next-word predictors. You can imagine others that require even more understanding.

I think one of the arguments here is that we do not know how thinking in the human brain works, so we can't say that something else doesn't do something similar.

These arguments are exhausting. Do you understand how neurons work under the hood? And yet, hook billions of them together in interesting ways, and look what happens.

Here are the steps to create a sentient primate: 1. Create some amino acids 2. Let the amino acids mix and recombine 3. Let the ones best at reproducing themselves outcompete the rest 4. Repeat steps 2 & 3 until satisfied

Of course, this ignores the elephant in the room. These are just _proteins_. What does it mean? What is it for? What does it know?

Clearly a mess of amino acids and derivatives can’t actually produce intelligence or understanding. At best, it can merely parrot the intelligence encoded in the base pairs.

The question we are asking is, "Where does the complexity come from?"

A complex result can come from a complex system, or a simple system with complex input (emergence). Human genetics is a complex system.

My argument is that an LLM is a simple system. The complexity of its output comes from the complexity of its input, not the system itself.

That complexity was created by human minds, not by AI. Without the human minds, there would be no complex LLM output.

Thanks - I think I understand you a bit better now.

I suppose I was arguing that genetics is a simple system with emergent complexity (and that perhaps LLMs could be too), but I suspect that betrays my limited understanding of genetics!

> We don't train models on noise. We train them on coherent data. The very act of writing is where intelligence lies. A model can only present that intelligence. It cannot create it, and it is not aware of it.

I think my main sticking point is the last sentence, which feels like a jump to me. For all the wisdom in your comments on complexity and invention, how does that link to awareness or understanding. Why does understanding require “ownership” of the complexity, so to speak?

To me, that’s not clear; could these be emergent properties of a system where the complexity is generated elsewhere, for instance?

Humans write. LLMs pretend to write. Really what they are doing is a statistical process called "continuation".

By training an LLM on human-written content, we are essentially restricting the problem domain from "any possible next character" to "something similar to what the humans wrote".

What the LLM does at the end of the day can be as simple and intelligent as blindly stumbling around whatever data it was given, because the domain it is stumbling around is made exclusively out of human-written content.

An LLM won't generate nonsense, because nonsense lies outside its domain: the LLM was never trained on nonsense, so there is no nonsense for it to stumble into.

An LLM can stumble into lies, offensive language, threats, incorrect arithmetic, logical fallacies, etc. These are generally presented as "limitations". The limitation is that we can't keep them out of the domain. It would be more clear to say those are limitations of the LLM creator. To the LLM itself, these limitations are a feature; though not the kind we are happy about.

Humans are able to think about our writing before we write it. That's a feature called objectivity, and LLMs don't have it. There's no obvious place to add that feature, either.

Hence the conclusion must be: clearly we have intelligence because God created us.

:)

That sounds about right. There is definitely nothing in the _design_ of the model or in the training procedure that is already and by itself intelligent, instead it is an intelligence extraction process. The intelligence is in the data and some of that apparently gets distilled into the model.

So we should stop thinking that the intelligence is the (untrained) brain and realize that it has always been the text.

I think it makes more sense to compare the current models to unconscious thought or intuition. So the result just kind of pops up as a result of a computation, just like unconscious thought pops up for humans as a result of a (hidden) computation. Humans do conscious reasoning by bringing their thoughts into memory and keeping them there, stretching the process out over time. In some sense LLMs being able to solve some things by being instructed to think step by steps reminds me of that. So perhaps adding some rumination will create a layer on top that appears closer to what we think of as thinking.

Right, it seems incorrect to claim that "The model can generate text that it couldn’t possibly have seen in the training data" when it's been trained on all manner of humans writing about all manner of things.

It seems likely that the training data included text involving " self-serving bias, metaphor, statistical syllogism and common-knowledge physics". So, when given a prompt asking for these things; the model is able to 'Stochastic Parrot' an appropriate response.

The graph theory work is clever, but the assumptions about the Models, and about Language, seem incorrect.

> We don't train models on noise. We train them on coherent data. The very act of writing is where intelligence lies. A model can only present that intelligence. It cannot create it, and it is not aware of it. There is no objective-thought function. If there was, then the LLM could simply tell us about it.

We don't train humans on noise. We train them on intangible and tangible things in the world. The very act of choosing and ordering the things is where intelligence lies. A human can only present that intelligence. It cannot create it, and it is not aware of it. There is no objective-thought function. If there was, then the human could simply tell us about it.

Turtles all the way down.

Let me put it this way: the text in the training corpus may very well contain the objective thought function. But to the LLM, that is indistinguishable from anything else. It's just another set of patterns between tokens.

From the LLM's perspective, there is only "what", and never "why". The "what" may even contain the "why"! Even so, it is always "what" to the LLM's eyes.

There is an objective difference between ideas, expression, and logic in my mind. I can't tell you precisely what that difference is made of, but I'm quite certain that LLMs are missing it.

And this is different from a baby how, other than that the baby takes several years to train? Do you train babies on noise? (There's a dystopian alternate-universe short story on training babies on noise to make them cattle).

Where did the first language come from?

Has anyone shown that you can leave a collection of LLMs in an empty environment, and watch them invent their own language or mathematics? I doubt that would work.

Hah, you would not believe it, but I am planning to use LLM to solve math problems by training it on the posts of my blog. Of course, my blog is silly nonsense, but it will be a fun exercise and address the elephant in the room as well. We will see what results we get for some fun prompts!

Marinate the nether rod

Power drill the yippee bog

Pressure wash the quiver bone

Tokenize the training corpus

Reply on news.ycombinator.com