[sleepytree]

If I write a book that contains Einstein's theory of relativity by virtue of me copying it, did I create the theory? Did my copying of it indicate anything about my understanding of it? Would you be justified to think the next book I write would have anything of original value?

I think what he is trying to say is that LLMs current architecture seems to mainly work by understanding patterns in the existing body of knowledge. In some senses finding patterns could be considered creative and entail reasoning. And that might be the degree to which LLMs could be said to be capable of reasoning or creativity.

But it is clear humans are capable of creativity and reasoning that are not reducible to mere pattern matching and this is the sense of reasoning that LLMs are not currently capable of.

[TeMPOraL]

> If I write a book that contains Einstein's theory of relativity by virtue of me copying it, did I create the theory? Did my copying of it indicate anything about my understanding of it? Would you be justified to think the next book I write would have anything of original value?

No, but you described a `cp` command, not an LLM.

"Creativity" in the sense of coming up with something new is trivial to implement in computers, and has long been solved. Take some pattern - of words, of data, of thought. Perturb it randomly. Done. That's creativity.

The part that makes "creativity" in the sense we normally understand it hard, isn't the search for new ideas - it's evaluation of those ideas. For an idea to be considered creative, it has to match a very complex... wait for it... pattern.

That pattern - what we call "creative" - has no strict definition. The idea has to be close enough to something we know, so we can frame it, yet different enough from it as to not be obvious, but still not too different, so we can still comprehend it. It has to make sense in relevant context - e.g. a creative mathematical proof has to still be correct (or a creative approach to proving a theorem has to plausibly look like it could possibly work); creative writing still has to be readable, etc.

The core of creativity is this unspecified pattern that things we consider "creative" match. And it so happens that things matching this pattern are a match for pattern "what makes sense for a human to read" in situations where a creative solution is called for. And the latter pattern - "response has to be sensible to a human" - is exactly what the LLM goal function is.

Thus follows that real creativity is part of what LLMs are being optimized for :).

[sleepytree]

> For an idea to be considered creative, it has to match a very complex... wait for it... pattern.

If we could predefine what would count as creativity as some specific pattern, then I'm not sure that would be what I would call creative, and certainly wouldn't be an all-inclusive definition of creativity. Nor is creativity merely creating something new by perturbing data randomly as you mentioned above.

While LLMs might be capable of some forms of creativity depending on how you define it, I think it remains to be seen how LLMs' current architecture could on its own accomplish the kinds of creativity implicit in scientific progress in the Kuhnian sense of a paradigm shift or in what some describe as a leap of artistic inspiration. Both of these examples highlight the degree to which creativity could be considered both progress in an objective sense but also be something that is not entirely foreshadowed by its precursors or patterns of existing data.

I think there are many senses in which LLMs are not demonstrating creativity in a way that humans can. I'm not sure how an LLM itself could create something new and valuable if it requires predefining an existing pattern which seems to presuppose that we already have the creation in a sense.

[corimaith]

>Creativity" in the sense of coming up with something new is trivial to implement in computers, and has long been solved. Take some pattern - of words, of data, of thought. Perturb it randomly. Done. That's creativity.

Formal Proof Systems aren't even nearly close to completion, and for patterns we don't have a strong enough formal system to fully represent the problem space.

If we take the P=NP problem, that likely can be solved formally that a machine could do, but what is the "pattern" here that we are traversing here? There is a definitely a deeper superstructure behind these problems, but we can only glean the tips, and I don't think the LLMs with statistical techniques can glean further in either. Natural Language is not sufficient.

[daveguy]

> Take some pattern - of words, of data, of thought. Perturb it randomly. Done. That's creativity.

This seems a miopic view of creativity. I think leaving out the pursuit of the implications of that perturbation is leaving out the majority of creativity. A random number generator is not creative without some way to explore the impact of the random number. This is something that LLM inference models just don't do. Feeding previous output into the context of a next "reasoning" step still depends on a static model at the core.

[glenstein]

>If I write a book that contains Einstein's theory of relativity by virtue of me copying it, did I create the theory? Did my copying of it indicate anything about my understanding of it? Would you be justified to think the next book I write would have anything of original value?

If you, after copying the book, could dynamically answer questions about the theory, it's implications, and answer variations of problems or theoretical challenges in ways that reflect mainstream knowledge, I think that absolutely would indicate understanding of it. I think you are basically making Searle's chinese room argument.

>But it is clear humans are capable of creativity and reasoning that are not reducible to mere pattern matching and this is the sense of reasoning that LLMs are not currently capable of.

Why is that clear? I think the reasoning for that would be tying it to a notion "the human experience", which I don't think is a necessary condition for intelligence. I think nothing about finding patterns is "mere" insofar as it relates to demonstration of intelligence.

[vouwfietsman]

> But it is clear humans are capable of ...

Its not though, nobody really knows what most of the words in that sentence mean in the technical or algorithmical sense, and hence you can't really say whether llms do or don't possess these skills.

[southernplaces7]

>nobody really knows what most of the words in that sentence mean in the technical or algorithmical sense

And nobody really knows what consciousness is, but we all experience it in a distinct, internal way that lets us navigate the world and express ourselves to others, yet apparently some comments seem to dismiss this elephant of sensation in the room by pretending it's no different than some cut and dried computational system that's programmed to answer certain things in certain ways and thus "is probably no different from a person trained to speak". We're obviously, evidentially more than that.

[mrcsd]

Words are not reducible to technical statements or algorithms. But, even if they were, then by your suggestion there's not much point in talking about anything at all.

[Rury]

> LLMs current architecture seems to mainly work by understanding patterns in the existing body of knowledge ...

>But it is clear humans are capable of creativity and reasoning that are not reducible to mere pattern matching and this is the sense of reasoning that LLMs are not currently capable of

This is not clear at all. As it seems to me, it's impossible to imagine or think of things that are not in someway tied to something you've already come to sense or know. And if you think I am wrong, I implore you to provide a notion that doesn’t agree. I can only imagine something utterly unintelligible, and in order to make it intelligible, would require "pattern matching" (ie tying) it to something that is already intelligible. I mean how else do we come to understand a newly-found dead/unknown language, or teach our children? What human thought operates completely outside existing knowledge, if not given empirically?

[rileymat2]

Why can’t creativity be taking the works, a bunch of works, finding a pattern then randomly perturbing a data point/concept to see if there are new patterns.

Then cross referencing that new random point/idea to see if it remains internally consistent with the known true patterns in your dataset.

This is how humans create new ideas often?