Post

And even when quantum computing enters the field it's not "reasoning" You can potentially use "quantum tunneling" to get the most optimized solution for deeply complex problems but that is also not "reasoning." It's using physics tricks to more efficiently drive computation.

Have seen really gripping proof on this.

Have you actually taken the remotest amount of time to learn about how LLMs work.

“Artificial cleverness”

I don't think it even makes sense to say that there's a distinction between "reasoning" and "mimicking reasoning". You either reason or you don't.

It’s not debatable. It’s only a “debate” if you fundamentally don’t understand the tech.

Well if you struggled with less you probably wouldn’t be so infatuated with AI.

I have another chart that assigns statistics to what charts I should use when multiple options exist. The charts are updated frequently. My messages back to you make sense to you, but again, I dont know what any of those characters mean. Am I sentient? Do I speak your language? Can I reason?

I think you are gullible

Again, if it could "reason", being fed every possible way to calculate the product of X and Y from all the stolen crap fed into it, being fed a question of "what is X times Y" it would use one of those methods to arrive at the conclusion. It doesn't.

how do you square this with many llms coming up with complex mathematical counterexamples that are not in the literature (nor anything close). for the record i think the downstream consequences suck pretty bad and also llms have no real world model but they can do math its hard to deny at this point

The most worrying thing that LLM development has shown us is how many people are completely lacking awareness of their own thought processes.

I agree, but it's probably more than you think.

Regardless of what it's doing... How does something imitating you change you? It's not your sense of self or nature you're worried about, I don't think. Maybe it's something else.

A Macaw could do mimicry of a sonnet but an intelligent person would not claim it can recite Shakespeare

Novel to whom? My niece laughed because "you know what? Chicken butt!" was novel to her. That doesn't make it creative or new

Its pure mimicry because it has no executive function

it's really easy to understand, in concept, The Chinese Room. But when faced with how many people just refuse the notion that whoever is in the room is clearly not fluent in Chinese, and the fact that the language we use around it implies they are fluent... it's a wild thing

bsky.app/profile/matt...

You simply can't have reasoning when you're a stateless machine. Come back when that changes.

Youre blending me the other guy. I only cool my phone with oil

Because you’re fucking dumb

We don’t understand everything about how human reasoning works, but we do know exactly how LLMs work and we know it’s definitely not doing anything most people would ever consider “reasoning.”

It IS understood how it works. And it is NOT reasoning, not as we humanly define it. That's not to say it's of no value; that would be an incorrect statement.

Why do you assume that your lack of understanding means that nobody understands? That's a staggeringly arrogant point of view.

or, I think - fail to do so and show how difficult it is to encode meaning in language

Alright, have an LLM make something without any training data.

It doesn’t think tho. It is simply just tying together multiple ideas which it has scraped from the internet. There isn’t any actual thinking or perception going on.

ok

Which is a charitable word for capitalist fantasy!

Yes, and I’m being very generous to the computers by putting them in company of our feathered friends.

Funny enough, there's a ruby-crowned Kinglet attacking the reflection in our parked car's driver's side mirror right now

No. Next question.

Good one

Exactly!

Jesus Christ do people actually think its reasoning? Calling it artificial intelligence was a mistake (it's not) and allowing the people who think movies are reality to spearhead its marketing was also a mistake.

It’s a program. It doesn’t do anything except when prompted. You might as well say an IBM computer from the 70’s is reasoning because it can do math.

The fact that you're able to make that argument shows you have no idea how LLMs actually work. This is the same technology that predicts the next word you're going to type on your phone, but on steroids. All they're doing is probabilistically stringing words together. There's no THOUGHT there.

If you grant that LLMs can pattern match and perform templated substitutions, then they can evaluate predicate logic.

In the end, it all comes down to assign, compare and jump. Fantastic we've come this far with those building blocks, but here we are. We're starting to repeat history in that what we don't understand surely must be divine.

I think the specifics are too charged to have a great discussion but what you describe as Self I think of as language; the fact that LLMs work as well as they do given the underlying model is revealing about language to me; insofar as one considers language part of self, your argument should hold.

This is not an essential limitation of the tech, Claude can scan poetry, count syllables and rhyme. But the logic inherent in language is indeed enough to produce logical statements without needing to reason.

AI doesn't change the 'nature of humanity" if we become differently introspective about what it means to be human and what it means to think, the underlying nature is still the same

Language makes sense by itself in a network of references. We have known that for over a hundred years. The computer doesn't need to actually "think" to be able to use the map of the English language it was able to produce from training

If you told someone a decade ago that in ten years giant statistical models of human language would be capable of hacking computers with very little in the way of human direction, people would ask if you were high.

I mean, post-structuralist linguistics started in the 1960s

its always funny to me that saying LLMs are not doing independent reasoning actually makes LLMs even weirder, and the impact on how we view humanity even bigger than saying that they are capable of independent reasoning

It wouldn't change human nature but it would definitely be a blow to people who consider human cognitive capabilities to be fundamentally unique/special.

that's because that's what it's been trained to do. If we trained in motivation and independent reasoning, that's what we'd get

why should motivation be considered determinative as to intelligence or sapience or capacity for reasoning? As for independent reasoning, again, unclear why that should be determinant of intelligence and also, if anything LLMs have too much of that due to cursed stochasticity

And a human performing an equivalent reasoning task hasn't ingested vast swathes of all human knowledge up to a certain date, which is then near-instantly indexable by their mind, to perform that task. If the machine is thinking, it's still not clear these modes of cognition are similar IMO.

Dude, there is no magic quotient in them. They are mimicking bots trained by assholes to scam you into thinking you are talking to something. They do a great job at it because all of the money in the world has been poured into it! It sucks!

Language is the interface, and as such there is interop. If the what is reason and the claim is that "reasoning" must use the same method to arrive at said answer, that's fine, but that's almost an ancient question, never to be solved.

It doesn't, though. It puts words (or pixels) together in a statistically likely combination. It has the same relation to reasoning as a funny looking cloud does to Abraham Lincoln.

Machines are never going to have a novel though, Paul. It scares the ever-loving shit out of me that so many of you tech bros are incapable of understanding that.

You're kidding, right? Please tell me you're kidding. Because otherwise you have no idea how this stuff works.

I would accept that they are simulacra. I'd accept that they can occasionally be very good simulacra, but are far more often mediocre or downright bad at it. but that's what they are & forgetting that is the pathway to madness both on a personal and societal level.

You tricked yourself into believing an algorithm is alive because you aren't used to having actual conversations with people. It isn't thinking, it's barely guessing correctly.

if AI can reason and think like you say it can then corporations own thinking entities and people use them for labour and entertainment without any sort of benefit or compensation to the entity, we have a name for this phenomena and you're not gonna like it!

I think this articulates it well. IMO the root of it is that there is more than one way to answer questions, and LLMs do this a new way, with maybe high-order correlations between all the questions and answers and text they've seen or something. Not the human way, but still a way to do it.

also how would I know if all I was doing was mimicry? What if all I’ve ever known is mimicry? People pretending to act like people doing people things occasionally bumping into a novel sequence of information.

Because if you think your LLM is reasoning you either have an exceedingly narrow definition of reasoning or you don’t understand how LLMs work.

So you agree that what llms do isn’t reasoning?

Like how does one even tell the difference between "mimicking" and "reasoning" even if you have technical skill and access to look under the proverbial hood?

Would it? If a person only replied to others in sentences they had determined, based on other stuff they'd heard and read people say, on what they had calculated to be a likely string of words to make sense after that prompt, would we call that "reasoning"? Or would we call it, idk, me at a party

I would suggest that this is necessarily so: any process that works even poorly at coming at truths must be an approximation of Bayes. But they e.g. share the same sort of features like dual-process structure, quirks that suggests they're doing things like subitizing, etc (arxiv.org/abs/2607.26179)

The explicit claim (threat) from manufacturers that AI will take everybody's job, the push of AI into domains where it is harmful (teaching of humanities, mass surveillance, war), the sociopathic behaviour of some of the owners, all those don't help the case for AI.

That and the plagiarism and the lies and the wasted resources...

See I feel like this thread of the AI discourse should have ended with "is artificial intelligence artificial"

LLMs definitely "learn" differently than traditional video game AI (through machine learning over vast quantities of source material). But I'm still looking for evidence of different reasoning.

I agree with that. But I don't think we've seen genuine emergent reasoning. The models are just applying general language understanding to novel problems in a linguistic manner. They are basically all cynical Wittgenstein followers. For that reason, I'm still skeptical LLMs are a path to AGI.

And I say that in as neutral a way as I can; it is tbd exactly whether that generalizes indefinitely or whether the costs of doing it outweigh the benefits etc. But it is geniunely doing something genuinely /different/ to the "expert systems" we used to deal with

Crucially, this is because people reason and LLMs don't. A woodworker and an injection mold can both craft a chair. Why do we not claim the injection mold uses reason when it creates the same output as the woodworker?

okay but if I have to accept it, you have to accept that my hand actually turns into a fox when I do shadowplay

“Look at me so much smarter than a machine I can’t even come up with a unique response to new technology”

somewhere in this debate about "reasoning" is a more interesting question, to me, of how these probabilistic outputs are turning the user's cognitive capacity into grey goo. that doesn't strike me as a metaphysical question, but it does strike me as a philosophy-of-mind question?