Hacker News Reader

Comment by dagss | Hacker News Reader

"Writing code is probably something LLMs have excelled at the most. But how much of their ability to write code is due to their understanding of it vs. just them recalling code snippets which were present in their training dataset or their context window?"

This makes me think mostly about how humans work.

Programming is to humans like speaking a language. You have ideas for what to achieve and express, and if you are a proficient programmer then outputting those thoughts in a programming language is second nature.

Programming (like most things humans learn) takes practice. It is NOT enough to simply read a specification and then you know how to code (except for transferring skills between similar languages). You repeat, make mistakes, and learn from them.

I would be tempted to say that humans definitely "repeat snippets from their training dataset". We mainly recall and express from our training/experiences. Creativity is mostly recombination of things already seen.

View on HN

> Creativity is mostly recombination of things already seen.

To be clear, ‘mostly recombination’ is thankfully a lot more complex than next token prediction which is all LLMs can do.

Turns out the next token contains my job.

Have you used an LLM in the last 3 years?

Yeah, every day. Have you?

> "...how much of their ability to write code is due to their understanding of it vs. just them recalling code snippets which were present in their training dataset or their context window?"

I don't understand how people who've actually used a modern coding agent, or know anything about them, can still be asking this question.

The most basic feature of LLMs is that they predict tokens based on their training data and input (context window.) But "predicting tokens" is not the same thing as copying snippets. None of this would work nearly as well if what they were doing was copying snippets.

And even if this were a question worth studying, the article's approach doesn't in any way help answer the question, because apparently the goal wasn't in fact to answer that question but rather to create a language that would be intrinsically difficult for LLMs. It'd also be pretty difficult for humans. That's what obfuscation does.

This would be better if it were just framed as a fun exercise, like any other obfuscated language, not as having some real purpose.

Yeah, no. LLMs do fantastic on problems they can copy out of their training set, but they fall apart on actually novel tasks. That's the entire reason we have to keep inventing new benchmarks, and why scores on new benchmarks are so low.

LLMs copy paste from their training set. That's plainly obvious when you try to do real work with them.

Oh cool, where did they copy the solutions to hundreds of open math questions from?

Similar math work that no human had connected to that problem

Remixing existing work into something novel is ~ the definition of creativity.

That's not an exhaustive definition, creating also means adding new stuff to the mix often involving leaps of imagination. Or we would have been stuck with the same words, music, tools, programming languages from the beginning. It's hard to see we would have even gotten started without those leaps and new ideas, genres, stories, programming paradigms coming about. Humans have been creating new things from the start.

When I look at the history of human culture I really see a lot of remixing and "trends" and the leaps are rather small.

What sticks out is really the lack of people doing something totally new. Look back on history and it seems like humans can't do anything except as in response to some input that we re-mix. (Although it is across fields -- a change in economic conditions or scientific outlook may affect music and painting; a small improvement in factory production methods may ripple down to experiments validating general relativity, etc)

Why didn't Mozart create a piece of jazz or rock to mix things up? Why are there no medieval cubistic painters? Why didn't someone understand $scientific_topic before?

Most such questions can be answered by some kind of "the social environment and culture at that time didn't prepare people to do it". I.e., humans couldn't remix it yet from what was already there.

Yes there are leaps but they seem to always be within some small "box" that prior knowledge creates for us. We can jump one inch at the time but no further..

In the "I wonder if this is in its training set"...

FRACTRAN. Given 2^x return 3^(x^3).

https://chatgpt.com/share/6ad58a36-5278-83ea-b1f5-22534915e1...

I entered two prompts to it:

    Describe the language FRACTRAN.


    Write a FRACTRAN program that takes 2^x and finishes with the value of in the 3^y of y = x^3.
    4 (2^2) -> 6561 (3^8).  Other values may be in other power registers.

And it did https://chatgpt.com/share/6ad58a36-5278-83ea-b1f5-22534915e1...

The program that it came up with was:

38/17 1495/38 29/19 19/23 62/377 37/29 29/31 41/185 89/37 3913/82 47/41 41/43 106/611 59/47 47/53 61/413 37/59 9581/122 71/61 61/67 146/923 79/71 71/73 249/869 59/79 79/83 97/178 1/89 89/97 17/2

You can paste that into https://tjwei.github.io/Fractran/ and run it. Put the input of 2^3 and I recommend a 1000x speedup. It's output was 3^27

---

One can certainly go adversarial with their tasks... though I know junior programmers that fall apart on things they can't look up on Stack Overflow either.

For tasks that I've worked with it on, I've been reasonably impressed by it. Though I'll also note that a lot of the work that we do isn't new.

There are register machines out there. When I search google for "register machine cube of number" it comes up with one. The question of how do you apply the register machine model for that (if you're claiming that is in its training data rather than searching for it like I did) and then compile that into FRACTRAN is about chaining together tools reasonably... it appears to have written a compiler in Python for a model register machine to FRACTRAN.

Lets ask a different question... if I gave you the task of writing that program in FRACTRAN, would you attempt to solve it in the same way?

I don't think this is entirely true. It feels almost like a no true Scotsman: if even parts of a task can be related to anything in the training data, then clearly it's not novel enough for the sake of this argument.

It feels like language models are becoming more capable of picking novel problems apart, identifying strategies for the pieces, and then executing the pieces to come up with a complete solution. This is part of what makes CoT reasoning pay off so well.

I am interested to see what comes after CoT reasoning (perhaps the latent-space reasoning [0] that recently hit the front page?), and how it'll affect which problems can be effectively solved with a model. But I feel like it's not very useful to say models completely fall apart on novel tasks.

Maybe they fall apart when the solution includes completely novel strategies, but sometimes models engineer new strategies too, so even that's not completely off the table.

But then again, you could still say that if it's possible to just engineer a solution out of the blue by recycling patterns from the training data, was the problem ever even novel in the first place?

Also consider that models don't only directly learn patterns from the training data. There's a limited degree of generalization that enables the model to reproduce the patterns of the training data, but not necessarily only the patterns of the training data. The way loss is modeled in training is designed to allow the model to plausibly output the training data, not to ensure the model outputs nothing else.

[0]: https://news.ycombinator.com/item?id=50145337

> I don't think this is entirely true.

That's a very polite way to say that it's complete bullshit. You can't really have a discussion with people who aren't engaging with what the technology actually does, but rather with their biased, quasi-religious belief about what it can't possibly do - namely, compete with humans on intellectual tasks. When the reality is that we're all outclassed by these models, in many respects.

> That's a very polite way to say that it's complete bullshit.

It's not the worst rule of thumb I've seen, and it does explain some things. It's just not entirely accurate, as I said. There's some nuance it doesn't capture, but it does capture some of the picture.

"Complete bullshit" would be more than a somewhat-accurate, yet incomplete understanding to me. It would be... well, complete bullshit. And I would probably call it such.

You are both talking about this claim?

    LLMs do fantastic on problems they can copy out of their training set, but they fall apart on actually novel tasks.

I mean, I give Claude what I consider novel problem solving work all the time.

You can make the statement true if you raise the bar for "novel task" to a point where most humans also can't do them. In these threads the goal posts for words like "understanding", "creative", "novel" etc gets moved for LLMs to way past their previous definitions, in fact way past a large % of the worlds human software developers.

To just take one example: Ask for a security review and you get back a list of all the things you would have found yourself after some days of pouring over the code (if you are senior developer well versed in security issues), plus the ones you would have overlooked, but in 15 minutes instead of in several days.

Before LLMs came along this would obviously have been "novel work". A shining example of a human using their intellect! But now that LLMs can do it it is just "copy and paste work"...

For me, that claim indeed. I believe they mean novel types of task as opposed to novel instances of existing problems

I had conversations in Morse code with like gpt 3 or 4. I guarantee there were no snippets of what we were chatting about, in morse code, in its training set.

LLMs are clearly able to apply seemingly arbitrary transformations to symbol streams in ways that generalize well beyond the training set. It's getting into metaphysics whether that is "novel" but it's definitely not "copy paste from the training set"

This post is literally about a language the author invented to be hard for LLMs to parse and isn't in their training set and yet Opus 5 scored 68.8% on it.

You may agree with the author (that it doesn't mean it understands the language) but it's pretty hard to argue that this is copied form the training set.

It might not be copied from its training set (yet), but it is copied from its context window.

The code it produces is a clear copy of the spec or the existing code that is fed to it as part of the task. It is just following the spec instructions, and some models follow it so closely that they get caught out by the lies/adversarial traps in the spec. Opus 5 is better at seeing through the lies, but it still very much copies from its context window.

I mean this is a pretty unusual standard to hold something to. How is it supposed to operate if you can't tell it how to operate?

Humans look at the docs too! Is there any difference?

Yes, humans can learn which parts of the spec are true and which aren’t. LLMs can too, but only through their context window (or long term through retraining)

You think you don't have a context window? Try coming back to your own uncommented code after half a year.

In-context learning is more than "copying" though. It's generalization.

> Yeah, no. LLMs do fantastic on problems they can copy out of their training set, but they fall apart on actually novel tasks.

Do you believe humans are any different, in general? What "actually novel task" have you recently aced?

> LLMs copy paste from their training set.

This doesn't follow, and isn't consistent with even the most shallow understanding of how a neural network works.

If we critique your own thought process here in the same way, we'd have to conclude that you're not a very useful model.

> That's plainly obvious when you try to do real work with them.

What's plainly obvious is that you've never tried to do real work with them.

It's easy to tell when you're dealing with posers: they describe their brief aborted experience from a year or more ago as though it's universal.

I’m getting a little tired of these pedantic games of gotcha undermining any attempts at serious engagement with the subject by people who say things that are even slightly skeptical of the capabilities of LLMs.

Back in college we called it “philosopher sniping”, and it wasn’t a positive term. It was used to denote the general bad faith vibes that people got from folks who would force everyone trying to participate in a serious discussion to be grindingly precise in their choice of words, because any time someone used any sort of verbal shorthand they would interpret it in the least charitable way possible and use that as a pretense to start belittling and mansplaining.

This is not gotcha, this is fundamental reality.

> any attempts at serious engagement with the subject

If you believe the OP article fits that description, we have nothing further to discuss.

> force everyone trying to participate in a serious discussion to be grindingly precise in their choice of words

This is no way, shape, or form about "word choice". The quote I posted reflects an objectively incorrect understanding of what LLMs do.

> ...start belittling and mansplaining.

Pointing out objectively true facts is not "mansplaining".

Do you not think "serious engagement with the subject" requires at least some familiarity with the subject? That's not in evidence here or in your comments. What sort of serious engagement do you believe you're achieving by coming at this from a position of near-total ignorance?

This reply is so perfect, I would almost think it's satire.

No satire intended. I'm just extremely sick of seeing people - on a site called "Hacker News", no less - react to the most amazing technology humans have ever invented with, basically, tribal fear and wilful ignorance. (They could learn about it if they wanted to.) It's just sad. I suppose they're just not capable of anything better.

And to talk about "serious engagement with the subject" while doing that is just salt in the wound. Yes, let's fucking engage seriously with the subject! Where is that happening, though?

The satire was apparently lost on you.

And claiming that it is the "most amazing technology humans have ever invented" is telling. Also claiming that a contrarian opinion is immediately rooted in tribal fear and ignorance. Ever bothered to stop and think that there might be good reasons to reject LLMs as a technology the way it is currently being used or is YOUR view of that maybe wilful ignorance and fueled by tribal fear?

But the response you gave was just perfectly playing exactly into the point the parent comment was making.

> there might be good reasons to reject LLMs as a technology the way it is currently being used

Yes, absolutely. But denying their obvious capabilities is not a very effective way to do that.

I don't see much thought-through analysis from the anti-AI crowd. That seems to have a lot to do with the fact that they're not actually very familiar with the technology, from either a practical or theoretical perspective.

This means they rely on what they assume will be an effective tactic, trying to claim that the machines that are smarter than they are in many respects are somehow... not. Which puts them in very obvious conflict with reality. Counterfactuals are not a great basis for a protest movement.

As for the parent comment, it was talking about "pedantic games of gotcha." I reject the idea that that's what my comment was doing. The idea that LLMs are copy-pasting snippets is ludicrous delusion. But maybe you don't realize that either?

Also, point of order: "The satire was apparently lost on you" makes no sense, since I wrote the comment and it wasn't intended as satire.

I honestly don’t think you’re engaging in good faith, so I won’t waste much of my time, but here’s a common thought, given to you by someone very familiar with the technology:

LLMs and AI derivates are interesting technology with many possible applications. However their externalities, including but not limited to the further consolidation of power in the hands of a few, rising inequality, increased propaganda, spam, scams, surveillance, uses in war, loss of critical thought and skills, cementing loneliness and driving suicide, devaluing of art, consumption of resources, increase in tech prices, etc, etc, etc, etc aren’t worth it.

That’s just it. It isn’t worth it. LLMs are interesting, but not that interesting, and that should be clear to anyone with the ability to have some empathy and look beyond their own nose.

What does this have to do with the obfuscated language in the OP?

Is it your goal to derail every post related to AI like this?

> What does this have to do with the obfuscated language in the OP?

It has to do with your post that I replied to. I’ll remind you of your own words:

> I don't see much thought-through analysis from the anti-AI crowd. That seems to have a lot to do with the fact that they're not actually very familiar with the technology, from either a practical or theoretical perspective.

You made a general comment, and that’s what I replied to. You then made a more targeted comment (“As for the parent comment”), which I did not reply to.

> Is it your goal to derail every post related to AI like this?

What an absurd accusation. Indeed reinforcing you’re not arguing in good faith and this is a waste of time.

> Yes, absolutely. But denying their obvious capabilities is not a very effective way to do that.

So you imply that anyone anti-AI is not actually familiar with the technology? So only a true Scotsman is able to appreciate the beauty of AI?

> the machines that are smarter than they are

Depends on your definition of smart.

> Also, point of order: "The satire was apparently lost on you" makes no sense, since I wrote the comment and it wasn't intended as satire.

I'll just let that stand for itself and encourage you to read the thread again if it's not beyond your context window.

> I'm just extremely sick of seeing people - on a site called "Hacker News", no less - react to the most amazing technology humans have ever invented with, basically, tribal fear and wilful ignorance.

Personally, I’m extremely sick of seeing people—especially on this website—spewing that tired fallacious talking point.

If you believe LLMs to be “the most amazing technology humans have ever invented”, I think you lack imagination. Medicine, agriculture, computers themselves… There are entire fields of inventions which are more amazing and important than LLMs. You’re just used to them so pay them no mind.

The fact you think opposing arguments are “tribal fear and wilful ignorance” shows you pay no attention to them and dismiss them without engaging in good faith.

All those technologies you mention were created by intelligences. None of them are intelligences themselves. That’s why I said what I did.

Also, I didn’t say all opposing arguments fit the description I gave. “Good faith” is not sufficient here - you need to engage with intelligence and reading comprehension, as well.

But this is another trend that’s getting annoying: accusing people with a nuanced - not necessarily even negative, just not 100% positive - opinion of the tech of being a hater.

I would submit that your accusation of willful thinking smells like a case of projection.

And before we go there, no, I don’t like it when people who don’t like the tech start calling everyone who hates it less than they do a booster, either.

> the most amazing technology humans have ever invented

Ah, yes, nothing subjective here, no sir.

> > ...start belittling and mansplaining.

> Pointing out objectively true facts is not "mansplaining".

Different axis entirely, actually, but also tends towards being about objectively true facts. Mansplaining is a sexist term that's entirely about correcting a misconception, but only applies when talking to a woman. "A condescending explanation" is a non-sexist one-to-one replacement for it.

Welcome to the internet.

> Creativity is mostly recombination of things already seen. Doesn't this apply to any creative work produced by humans?

Yes that is what I say so we agree.

> But how much of their ability to write code is due to their understanding of it vs. just them recalling code snippets which were present in their training dataset or their context window?

Given that a small model that was trained only on pre-computer-era texts can write simple python programs after having python explained to it in context I'd say that LLMs must be pretty good at understanding coding.

This is a very interesting datum.

Can you point me to a source?

https://talkie-lm.com/introducing-talkie

Thank you, much appreciated!

> I would be tempted to say that humans definitely "repeat snippets from their training dataset". We mainly recall and express from our training/experiences. Creativity is mostly recombination of things already seen.

Isn’t this Skinner (Operant Conditioning)? This worldview is certainly intuitive to many.

Reply on news.ycombinator.com