chpatrick

I think people misunderstand when they see that it's a "statistical model". That just means that out of a range of possible answers, it picks in a humanlike way. If the logical answer is the humanlike thing to say then it will be more likely to sample it.

In the same way a human might produce a range of answers to the same question, so humans are also drawing from a theoretical statistical distribution when you talk to them.

It's just a mathematical way to describe an agent, whether it's an LLM or human.

The question wasn't can it be trusted, it was does it think.

Again, I'm not claiming that LLMs can think like people (I don't know that). I just don't like that people confidently claim that they can't, just because they work differently from biological brains. That doesn't matter when it comes to the Turing test (which they passed a while ago btw), just what it says.

I'm not saying that LLMs do anything, just that it's rich to confidently say they don't do something when we don't even understand how humans do it.

It's like we're pretending cognition is a solved problem so we can make grand claims about what LLM's aren't really doing.

Sometimes LLMs hallucinate or bullshit, sometimes they don't, sometimes humans hallucinate or bullshit, sometimes they don't. It's not like you can tell a human to stop being delusional on command either. I'm not really seeing the argument.

I disagree. Creativity is coming up with something out of the blue. Thinking is using what you know to come to a logical conclusion. LLMs so far are not very good at the former but getting pretty damn good at the latter.

LLMs are absolutely not "fully understood". We understand how the math of the architectures work because we designed that. How the hundreds of gigabytes of automatically trained weights work, we have no idea. By that logic we understand how human brains work because we've studied individual neurons.

And here's some more goalpost-shifting. Most humans aren't capable of novel mathematical thought either, but that doesn't mean they can't think.

Me neither but that's why I don't like arguments that say LLM's can't do X because of their substrate, as if that was self-evident. It's like the aliens saying surely humans can't think because they're made of meat.

Who says what degree of complexity is enough? Seems like deferring the problem to some other mystical arbiter.

In my opinion AI slop is slop not because AIs are basic but because the prompt is minimal. A human went and put minimal effort into making something with an AI and put it online, producing slop, because the actual informational content is very low.

I think plenty of people have problems with the second one but you wouldn't say that means they can't think.

Do you need to "think like a human" to think? Is it only thinking if you do it with a meat brain?

Humans are also trained on data made by humans.

> it could never, for example, invent a game like chess or a social construct like a legal system. Those require motivations like "boredom", "being social", having a "need for safety".

That's creativity which is a different question from thinking.

They don't follow "grammatical rules", they process inputs with an incredibly large neural net. It's like saying humans aren't really thinking because their brains are made of meat.

Have you tried asking a modern LLM to invent a programming language?

Actually general-purpose LLMs are pretty decent at playing chess games they haven't seen before: https://maxim-saplin.github.io/llm_chess/

Again the problem is that what "thinking" is totally vague. To me if I can ask a computer a difficult question it hasn't seen before and it can give a correct answer, it's thinking. I don't need it to have a full and colorful human life to do that.

Actually LLMs crunch through half a terabyte of weights before they "speak". How are you so confident that nothing happens in that immense amount of processing that has anything to do with thinking? Modern LLMs are also trained to have an inner dialogue before they output an answer to the user.

When you type the next word you also put a word that fits some requirement. That doesn't mean you're not thinking.

Could a human code the solution if they didn't learn to code from someone else? No. Could they do it if someone didn't tell them the rules of towers of hanoi? No.

That doesn't mean much.

Are any of those required for thinking?

It's more that "thinking" is a vague term that we don't even understand in humans, so for me it's pretty meaningless to claim LLMs think or don't think.

There's this very cliched comment to any AI HN headline which is this:

"LLM's don't REALLY have <vague human behavior we don't really understand>. I know this for sure because I know both how humans work and how gigabytes of LLM weights work."

or its cousin:

"LLMs CAN'T possibly do <vague human behavior we don't really understand> BECAUSE they generate text one character at a time UNLIKE humans who generate text one character a time by typing with their fleshy fingers"

[citation needed]

People constantly assert that LLMs don't think in some magic way that humans do think, when we don't even have any idea how that works.

Novo Nordisk has a market cap of 250 billion last I checked.

Depends on if you think we're at the end of AI development or the beginning.

Isn't the reason cancer? Eventually the DNA copying errors add up.

But for the analysis use cases you mentioned, can't you just ask an LLM to read the text and output the answer as JSON, and you're done? Is it just because running LLMs is expensive?

Incredible.

Is there any use case for "traditional" NLP in the age of LLMs?

You can upgrade your old Framework 16 to this. Framework 13 wouldn't work anyway because it's a different chassis.

Ok then where is the amazing non Linux deployment solution that everyone uses instead?