Thinking can not be separated from motivation. It's really simple. Humans and other organisms fundamentally think to replicate their DNA. Until AI has a similar incentive structure driving it, it won't be thinking. There is no human behavior or thought that can not be explained by evolutionary drives. It is really perplexing to me how people think "intelligence" is some kind of concrete thing that just magically emerges from a certain degree of computational complexity. I argue instead that intelligence is an adaptive behavior emerging from evolutionary drives interacting with the real world. World models are not prerequisite but consequent of such molded apparatus. Machines won't become intelligent until it is adaptive for them to do so. There is no magic just evolutionary drives and physical possibility. Our current top down approach of "pre-training" LLMs is bound to fail because it does not allow for real time emergence of adaptive behaviors such as general intelligence. Mimicking intelligence through predicting the next word is no more intelligence than a photograph of something is an actual thing. Training a combinatorial network to interpolate images and words is not the same thing as adaptive self modifying behavior in the real world of physics such as organisms engage with through the set of behaviors that we call intelligence.
Bag of words, have mercy on us
331–340 of 362 posts
Re: Bag of words, have mercy on us
#332I think a better metaphor is the Library of Babel. A practically infinite library where both gibberish and truth exist side by side. The trick is navigating the library correctly. Except in this case you can’t reliably navigate it. And if you happen to stumble upon some “future truth” (i.e. new knowledge), you still need to differentiate it from the gibberish. So a “crappy” version of the Library of Babel. Very impre…
This is where I sit too. Obviously language is an expression of thought but the Library of Babel is a great example that language without intent is just garbage. You got me thinking of reading before the internet. You'd grab a book and internalize the subject, later refining over time with more books, experiments and other forms of conversation. That journey of developing your own model is undervalued in understanding. That first book could of be absolute shit but you couldn't know that.
I've been learning more about roses lately and the amount of information on them varies so much because the world roses live in is equally varied. LLMs make for a better search engine but you still need to develop your own internal models, worse yet - if LLMs continue to be refined off of cul-de-sac conclusions then all the wisdom of the journey is lost both to the consumer and the LLM itself.
Re: Bag of words, have mercy on us
#333Earlier quoted context omitted.
Hmm, seems unlikely. They are not sounded out part is true, sure, but I question whether 'abstract thoughts' can be so easily dismissed as mere words. edit: come to think of it and I am asking this for a reason: do you hear your abstract thoughts?
Different people have different levels of internal monologuing or none at all. I don't generally think with words in sentences in my head, but many people I know do.
Internal monologue is a like a war correspondent's report of the daily battle. The journalist didn't plan or fight the battle, they just provided an after-the-fact description. Likewise the brain's thinking--a highly parallelized process involving billions of neurons--is not done with words.
Play a little game of "what word will I think of next?" ... just let it happen. Those word choices are fed to the monologue, they aren't a product of it.
Re: Bag of words, have mercy on us
#334Earlier quoted context omitted.
Are you a stream of words or are your words the “simplistic” projection of your abstract thoughts? I don’t at all discount the importance of language in so many things, but the question that matters is whether statistical models of language can ever “learn” abstract thought, or become part of a system which uses them as a tool. My personal assessment is that LLMs can do neither.
LLMs and human brains are both just mechanisms. Why would one mechanism a priori be capable of "learning abstract thought", but no others? If it turns out that LLMs don't model human brains well enough to qualify as "learning abstract thought" the way humans do, some future technology will do so. Human brains aren't magic, special or different.
Thermometers and human brains are both mechanisms. Why would one be capable of measuring temperature and other capable of learning abstract thought?
> If it turns out that LLMs don't model human brains well enough to qualify as "learning abstract thought" the way humans do, some future technology will do so. Human brains aren't magic, special or different.
Google "strawman".
Re: Bag of words, have mercy on us
#335Earlier quoted context omitted.
Human brains aren’t magic in the literal sense but do have a lot of mechanisms we don’t understand. They’re certainly special both within the individual but also as a species on this planet. There are many similar to human brains but none we know of with similar capabilities. They’re also most obviously certainly different to LLMs both in how they work foundationally and in capability. I definitely agree with the mat…
When someone says "AIs aren't really thinking" because AIs don't think like people do, what I hear is "Airplanes aren't really flying" because airplanes don't fly like birds do.
That's a fallacy of denial of the antecedent. You are inferring from the fact that airplanes really fly that AIs really think, but it's not a logically valid inference.
Re: Bag of words, have mercy on us
#336Earlier quoted context omitted.
As capable as they get, I still don't see a lot of uses for these things, myself, still. Sometimes if I'm fundamentally uninspired I'll have a model roll the dice, decide what I do or don't like about where it went to create a sense of momentum, but that's the limit. There's never any of its output in my output, even in spirit unless it managed to go somewhere inspiring, it's just a way to let me warm up my generatio…
Yeah, that makes sense. If people don't see uses for AI, they shouldn't use it. But going out of the way to imply that people who use AI cannot think is pretty stupid in itself imo. I am not sure how to put this, but maybe to continue with your example, I like a lot of indie stuff as well, but I don't think anyone who watches, say, Fast and Furious, cannot think or is stupid, unless they explicitly make it the case b…
I actually do think that people who prefer content of fidelity over content of intent are making a mistake, yes. I don't think they're incapable of thinking, I don't care to apply any virtue labels to this preference, but they are literally preferring not to think.
LLMs can only produce things by and for people who prefer not to do the work the LLMs are doing for them. Most of the time I do not prefer this.
Like, there was a 2-panel comic that went around the RPG community a bit back where it was something like "Game Master using LLM to generate 10 pages of backstory for his campaign setting from a paragraph" in the first panel and "Player using LLM to summarize the 10 page backstory into a paragraph" in the second. Neither of these people care for the filler (because they didn't produce or consume it) so it's turned the two-LLM system into a game of telephone.
Re: Bag of words, have mercy on us
#337Earlier quoted context omitted.
When someone says "AIs aren't really thinking" because AIs don't think like people do, what I hear is "Airplanes aren't really flying" because airplanes don't fly like birds do.
That's a fallacy of denial of the antecedent. You are inferring from the fact that airplanes really fly that AIs really think, but it's not a logically valid inference.
Yeah at that point, just arguing semantics
Re: Bag of words, have mercy on us
#338Earlier quoted context omitted.
LLMs are compression and prediction. The most efficient way to (lossfully) compress most things is by actually understanding them. Not saying LLMs are doing a good job of that, but that is the fundamental mechanism here.
Where’s the proof that efficient compression results in “understanding”? Is there a rigorous model or theorem, or did you just make this up?
It's the other way around. Human learning would appear to amount to very efficient compression. A world model would appear to be a particular sort of highly compressed data set that has particular properties.
This is a case where it's going to be next to impossible to provide proof that no counterexamples exist. Conversely, if what I've written there is wrong then a single counterexample will likely suffice to blow the entire thing out of the water.
Re: Bag of words, have mercy on us
#339Earlier quoted context omitted.
I am a stream of words - I have even ran out of tokens while speaking before :) But raising kids, I can clearly see that intelligence isn't just solved by LLMs
> But raising kids, I can clearly see that intelligence isn't just solved by LLMs Funny, I have the opposite experience. Like early LLMs kids tend to give specific answers to the questions they don't understand or don't really know or remember the answer to. Kids also loop (give the same reply repeatedly to different prompts), enter highly emotional states where their output is garbled (everyone loves that one), etc.…
Same failure modes, but not a general solution to intelligence.
Re: Bag of words, have mercy on us
#340Earlier quoted context omitted.
A model of language is a model of a tiny specialized part of the world: language. And if anybody gets annoyed that my comment is tautological, get annoyed by the people that made the comment necessary.
When you ask an LLM a question about cars, it needs an inner representation of what a car is (how imperfect it may be) to answer your question. A model of "language" as you want to define it would output a grammatically correct wall of text that goes nowhere.
A map of how concepts relate in language is not a model of the world, except on the extremely limited sense that languages are part or the world.
And yeah, that wasn't clear before people created those machines that can speak but can't think. But it should be completely obvious to anybody that interacts with them for a small while.