I feel like the logic of your question is actually inverted from reality.
It’s kind of like you’re saying “prove god doesn’t exist” when it’s supposed to be “prove god exists.”
If a problem isn’t documented LLMs simply have nowhere to go. It can’t really handle the knowledge boundary [1] at all, since it has no reasoning ability it just hallucinates or runs around in circles trying the same closest solution over and over.
It’s awesome that they get some stuff right frequently and can work fast like a computer but it’s very obvious that there really isn’t anything in there that we would call “reasoning.”
I don't want to address directly your claim about lack of generalization, because there's a more basic issue with the GP statement. Even though I will say, today's models do seem to generalize quite a bit better than you make it sound.
But more importantly, you and GP don't mention any evidence for why that is due to specifically using next token prediction as a mechanism.
Why would it not be possible for a highly generalizing model to use next token prediction for its output?
That doesn't follow to me at all, which is why the GP statement reads so weird.
> you and GP don't mention any evidence for why that is due to specifically using next token prediction as a mechanism.
Again, inverted burden of proof. We don’t have to prove that next token prediction is unable to do things that it currently cannot do and has no compelling roadmap that would lead us to believe it will do those things.
It’s perhaps a lot like Tesla’s “we can do robocars with just cameras” manifesto. They are just saying that they can do it because humans use eyes and nothing else. But they haven’t actually shown their technology working as well as even impaired human driving, so the burden of proof is on them to prove naysayers wrong. Put up or shut up, their system is approaching a decade late from their promises.
To my knowledge Tesla is still failing simple collision avoidance tests while their competitors are operating revenue service.
This other article critical of the test methodology actually still points out (defends?) the Tesla system by saying that it’s not reasonable to expect Tesla to train the system on unrealistic scenarios:
That really gets back to my exact point: AI implemented the way it is today (e.g. next token prediction) can’t handle anything it has no training data for while the human brain is amazingly good at making new connections without taking a ton of time to be fed thousands of examples of that new discovery.
> Why would it not be possible for a highly generalizing model to use next token prediction for its output?
The issue is that it uses next token prediction for its training, it doesn't matter how it outputs things but it matters how its trained.
As long as these models are trained to be next token predictors you will always be able to find flaws with it that are related to it being a next token predictor, so understanding that is how they work really makes them much easier to use.
So since it is so easy to get the model to make errors due to it being trained to just predict tokens people argue that is proof they aren't really thinking. Like, any extremely common piece of text when altered slightly will typically still output the same follow-up as the text it has seen millions of times even though it makes no logical sense. That is due to them being next token predictors instead of reasoning machines.
You might say its unfair to abuse their weaknesses as next token predictors, but then you admit that being a next token predictor interferes with their ability to reason, which was the argument you said you don't understand.
This is like watching a Jurassic Park movie and proclaiming “if nobody can tell the difference between a real dinosaur and a CGI dinosaur…” when literally everyone in the theater can tell that the dinosaur is CGI.
If I order Chinese takeout, but it gets made by a restaurant that doesn't know what Chinese food tastes like, then is that food really Chinese takeout?
Not exactly. When "intelligence" is like your frozen Walmart fries, the analogy works a bit better. Some people are arguing that yes, you can buy some frozen intelligence from your local (internet) store.
> I am so floored that at least half of this community, usually skeptical to a fault, evangelizes LLMs so ardently. Truly blows my mind.
...
> I’m open to them becoming more than a statistical token predictor, and I think it would be really neat to see that happen
I'm more shocked that so many people seem unable to come to grips with the fact that something can be a next token predictor and demonstrate intelligence. That's what blows my mind, people unable to see that something can be more than the sum of its parts. To them, if something is a token predictor clearly it can't be doing anything impressive - even while they watch it do I'm impressive things.
> I'm more shocked that so many people seem unable to come to grips with the fact that something can be a next token predictor and demonstrate intelligence.
Except LLMs have not shown much intelligence. Wisdom yes, intelligence no. LLMs are language models, not 'world' models. It's the difference of being wise vs smart. LLMs are very wise as they have effectively memorized the answer to every question humanity has written. OTOH, they are pretty dumb. LLMs don't "understand" the output they produce.
> To them, if something is a token predictor clearly it can't be doing anything impressive
Shifting the goal posts. Nobody said that a next token predictor can't do impressive things, but at the same time there is a big gap between impressive things and other things like "replace very software developer in the world within the next 5 years."
I think what BoiledCabbage is pointing out is that the fact that it's a next-token-predictor is used as an argument for the thesis that LLMs are not intelligent, and that this is wrong, since being a next-token-predictor is compatible with being intelligent. When mikert89 says "thinking machines have been invented", dgfitz in response strongly implies that for a for thinking machines to exist, they must become "more than a statistical token predictor". Regardless of whether or not thinking machines currently exist, dgfitz argument is wrong and BoiledCabbage is right to point that out.
> an argument for the thesis that LLMs are not intelligent, and that this is wrong,
Why is that wrong? I mean, I support that thesis.
> since being a next-token-predictor is compatible with being intelligent.
No. My argument is by definition that is wrong. It's wisdom vs intelligence. Street-smart vs book smart. I think we all agree there is a distinction between wisdom and intelligence. I would define wisdom as being able to recall pertinent facts and experiences. Intelligence is measured in novel situations, it's the ability to act as if one had wisdom.
A next token predictor by definition is recalling. The intelligence of a LLM is good enough to match questions to potentially pertinent definitions, but it ends there.
It feels like there is intelligence for sure. In part it is hard to comprehend what it would be like to know the entirety of every written word with perfect recall - hence essentially no situation is novel. LLMs fail on anything outside of their training data. The "outside of the training" data is the realm of intelligence.
I don't know why it's so important to argue that LLMs have this intelligence. It's just not there by definition of "next token predictor", which is at core a LLM.
For example, a human being probably could pass through a lot of life by responding with memorized answers to every question that has ever been asked in written history. They don't know a single word of what they are saying, their mind perfectly blank - but they're giving very passable and sophisticated answers.
> When mikert89 says "thinking machines have been invented",
Yeah, absolutely they have not. Unless we want to reducto absurd-um the definition of thinking.
> they must become "more than a statistical token predictor"
Yup. As I illustrated by breaking down the components of "smart" into the broad components of 'wisdom' and 'intelligence', through that lens we can see that next token predictor is great for the wisdom attribute, but it does nothing for intelligence.
>dgfitz argument is wrong and BoiledCabbage is right to point that out.
Why exactly? You're stating apriori that the argument is wrong without saying way.
When you type you're also producing one character at a time with some statistical distribution. That doesn't imply anything regarding your intelligence.
Wait-- are you claiming that AI is a bigger technological change than the development of computing devices and a networking infrastructure for those devices?
Well, is the computer revolution bigger than the electricity revolution? They just build on each other. But it might be interpreted as the next new abstraction that causes major changes in the industry.
and they don't have to revolutionize the world to be revolutionary in our industry. it might be that the use-cases unlocked by this new technology won't move the needle in an industrial revolution sense but it's nonetheless a huge leap for computer science and the kinds of tasks that can be done with software.
i don't understand people who
seem to have strongly motivated reasoning to dismiss the new tech as just a token predictor or stochastic parrot. it's confusing the means with the result, it's like saying Deep Blue is just search, it's not actually
playing chess, it doesn't understand the game—like
that matters to people playing against it.
I personally don't dismiss or advocate for AI/LLMs, I just take what I actually see happening, which doesn't appear revolutionary to me. I've spent some time trying to integrate it into my workflow and I see some use cases here and there but overall it just hasn't made a huge impact for me personally. Maybe it's a skill issue but I have always been pretty effective as a dev and what it solves has never been the difficult or time consuming part of creating software. Of course I could be wrong and it will change everything, but I want to actually see some evidence of that before declaring this the most impactful technology in the last 100 years. I personally just feel like LLMs make the easy stuff easier, the medium stuff slightly more difficult and the hard stuff impossible. But I personally feel that way about a lot of technology that comes along though, so it could just be I'm missing the mark.
I think this is the observation that's important right now. If you're an expert that isn't doing a lot of boilerplate, LLMs don't have value to you right now. But they can acceptably automate a sizeable number of entry-level jobs. If those get flushed out, that's an issue, as not everyone is going to be a high-level expert.
Long-term, the issue is we don't know where the ceiling is. Just because OpenAI is faltering doesn't mean that we've hit that ceiling yet. People talk about the scaling laws as a theoretical boundary, but it's actually the opposite. It shows that the performance curve could just keep going up even with brute force, which has never happened before in the history of statistics. We're in uncharted territory now, so there's good reason to keep an eye on it.
I'm starting to learn that AI progress is just really hard to talk about.
On the one hand, I completely agree with you. I've even said before, here on Hacker News, that AI is underhyped compared to the real world impact that it will have.
On the other, I run into people in person that seem to think dabbing a little cursor on a project will suddenly turn everyone into 100x engineers. It just doesn't work that way at all, but good luck dealing with the hypemeisters.