LLMs can’t reason — they just crib reasoning-like steps from their training data
LLMs can’t reason — they just crib reasoning-like steps from their training data
Did someone not know this like, pretty much from day one?
Not the idiot executives that blew all their budget on AI and made up for it with mass layoffs - the people interested in it. Was that not clear that there was no “reasoning” going on?
there’s a lot of people (especially here, but not only here) who have had the insight to see this being the case, but there’s also been a lot of boosters and promptfondlers (ie. people with a vested interest) putting out claims that their precious word vomit machines are actually thinking
so while this may confirm a known doubt, rigorous scientific testing (and disproving) of the claims is nonetheless a good thing
No they do not im afraid, hell I didnt even know that even ELIZA caused people to think it could reason (and this worried the creator) until a few years ago.
Well, two responses I have seen to the claim that LLMs are not reasoning are:
So I think this research is useful as a response to these, although I think "fuck off, promptfondler" is pretty good too.
A lot of people still don't, from what I can gather from some of the comments on "AI" topics. Especially the ones that skew the other way with its "AI" hysteria is often an invite from people who know fuck all about how the tech works. "Nudifier" or otherwise generative images or explicit chats with bots that portray real or underage people being the most common topics that attract emotionally loaded but highly uninformed demands and outrage. Frankly, the whole "AI" topic in the media is so massively overblown on both fronts, but I guess it is good for traffic and nuance is dead anyway.
Indeed, although every one of us who have seen a tech hype train once or twice expected nothing less.
PDAs? Quantum computing. Touch screens. Siri. Cortana. Micropayments. Apps. Synergy of desktop and mobile.
From the outset this went from “hey that’s kind of neat” to quite possibly toppling some giants of tech in a flash. Now all we have to do is wait for the boards to give huge payouts to the pinheads that drove this shitwagon in here and we can get back to doing cool things without some imaginary fantasy stapled on to it at the explicit instruction of marketing and channel sales.
We suspect this research is likely part of why Apple pulled out of the recent OpenAI funding round at the last minute.
Perhaps the AI bros “think” by guessing the next word and hoping it’s convincing. They certainly argue like it.
🔥
When you ask an LLM a reasoning question. You're not expecting it to think for you, you're expecting that it has crawled multiple people asking semantically the same question and getting semantically the same answer, from other people, that are now encoded in its vectors.
That's why you can ask it. because it encodes semantics.
did you ask a LLM for a post to make here? that might explain this mess of a comment
Rooting around for that Luke Skywalker "every single word in that sentence was wrong" GIF....
thank you for bravely rushing in and providing yet another counterexample to the “but nobody’s actually stupid enough to think they’re anything more than statistical language generators” talking point
because it encodes semantics.
if it really did so, performance wouldn't swing up or down when you change syntactic or symbolic elements of problems. the only information encoded is language-statistical
Paraphrasing Neil Gaiman, LLMs don't give you information; they give you information shaped sentences.
They don't encode semantics. They encode the statistical likelihood that each token will follow a given sequence of tokens.
because it encodes semantics.
Please enlighten me on how? I admit I don't know all the internals of the transformer model, but from what I know it encodes precisely only syntactical information, i.e. what next syntactical token is most likely to follow based on a syntactical context window.
How does it encode semantics? What is the semantics that it encodes? I doubt they have denatotational or operational semantics of natural language, I don't think something like that even exists, so it has to be some smaller model. Actually, it would be enlightening if you could tell me at least what the semantical domain here is, because I don't think there's any naturally obvious choice for that.
so.... a stochastic parrot?
What if I told you 90% of humans do that.
It's always funny to see this because you think that you're part of the smart 10% with original thoughts while actually you're the insufferable 10% whose only thought is that of superiority with nothing to back it up.
My cat has more original thoughts than that and he's currently stuck head-first in a cereal box.
it’s not shocking because we’ve seen worse, but it is remarkable how fascist the implications of this “most people don’t possess cognition” idea are
it’s also very funny how many of these presumed cognition-havers have come to this thread and our instance in general with effectively the same lazy, shitty, thoughtless take on the nature of humanity
You guys always come crawling out from whatever rock you’re hiding under for these posts as if someone saying LLMs aren’t smart makes your spider senses tingle.
It would be almost impressive if it weren’t so stupid.
What if I told you I have the power to ban you from the forum because you're terminally boring?
People keep saying this, but I'm not convinced our own brains are doing anything more.
Let the haters hate.
Despite the welcome growth of atheism, almost all humans at one level or another cling to the idea that our monkey brains are filled with some magic miraculous light that couldn't possibly be replicated. The reality is that some of us only have glimmers of sapience, and many not even that. Most humans, most of the time, are mindless zombies following a script, whether due to individual capacity, or a civilization that largely doesn't reward metacognition or pondering the questions that matter, as that doesn't immediately feed individual productivity or make anyone materially wealthier, that maze doesn't lead to any yummy cheese for us.
AI development isn't finally progressing quickly and making people uncomfortable with its capability because it's catching up to our supposedly transcendental superbrains (that en masse spent hundreds of thousands of years wandering around in the dirt before it finally occurred to any of them that we could grow food seasonally in one place). It's making a lot of humans uncomfortable because it's demonstrating that there isn't a whole hell of a lot to catch up to, especially for an average human.
There's a reason pretty much everyone immediately discarded the Turing Test and basically called it a bullshit metric after elevating it for decades as a major benchmark in the development of AI systems... The moment a technology and design that could readily pass it became available. That's the blind hubris of man on grand display.
Creationists: We don't understand the brain so it must be the work of god.
AI Worshipers: We don't understand the brain so it must work exactly like LLMs.
The reality is that some of us only have glimmers of sapience, and many not even that. Most humans, most of the time, are mindless zombies following a script
It’s a funny thing, that there are certain kinds of people who are assured of their own cleverness and so alienated from society that they think that echoing the same dehumanising blurb produced by so many of their forebears is somehow novel or informative, rather than just following a script.
(the irony of responding with an xkcd is not lost on me)
Much like the promptfondlers proudly claiming they are stochastic parrots, flaunting your inability to recognise intelligence in other humans isn’t a great flex.
of course somebody prompted up a LessWrong-specific chatbot