Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

And yet It Understands

borretti.me

Post reply on HN
Print view View original

161–170 of 231 posts

Re: And yet It Understands

#161

Post by esailija »

_dain_ wrote: post #110
ivanbakel wrote: post #105

Earlier quoted context omitted.

Because it's a purely statistical explanation that doesn't require understanding. Put differently, it's possible that GPT doesn't "understand" language itself as a concept, and instead tokens in the same language are just highly-correlated when it comes to prediction. When affecting weights between tokens, it wouldn't be surprising that those weights have effects across languages, much in the same way they work withi…

No, you're the one who's begging the question. Why can't a "purely statistical" process have an understanding? If you a priori assume it can't, then nothing could ever persuade you GPT understood anything, no matter how it performed. And again, this magical word "just". "Just highly correlated", "just probabilities". Putting the word "just" in front of something doesn't mean you've explained it.

> Why can't a "purely statistical" process have an understanding?

Once you understand long division, you can do it on infinite numbers without ever having seen the specific numbers. You can get this full understanding just from a handful of examples, no need for terabytes of them.

No matter how many examples of long division examples you fed to statistical model like GPT, there will always be infinite amount of numbers you can tell it where it will give the wrong answer*, unless you cheated and actually hard coded the understanding into the model.

If it cannot understand long division just from few examples it cannot ever understand it. The very reason it needs ridiculous amounts of data is precisely because it cannot understand. If you think it understands you simply aren't trying very hard to confirm otherwise.

* in a way that reveals there is no understanding of long division, obviously a human would also give wrong answer after being awake 100 hours writing numbers on paper

Re: And yet It Understands

#162

Post by famouswaffles »

civilized wrote: post #155
famouswaffles wrote: post #150

Earlier quoted context omitted.

Are humans not generally intelligent ?. Since when has been the answer to "not good at math" been chuck a textbook at it ? You would have limited success doing this with people. Do people not explain things they don't fully understand? Understanding is not binary. This is kind of problem I keep seeing. Expectations and post shifting have grown so much that a significant chunk of the human population wouldn't even pas…

I learned arithmetic laws from a textbook and practice problems. I didn't need a teacher gesticulating or any fancy multimodal stuff. It's symbolic manipulation. What is the machine missing that I was given? There's no post shifting. The research community has been setting itself realistically attainable benchmarks. Now that the research community has made a lot of progress against its benchmarks, we have hype, claim…

I don't care what you did. All due respect, you are one person. I care about what can observed by people in general.

Dunno what to tell you other than textbook and practice problems is far from the solution you think it is for a big chunk of the population.

Re: And yet It Understands

#164

Post by YeGoblynQueenne »

marcosdumay wrote: post #133
_xnmw wrote: post #127

I still hold that it doesn't "understand". Even if it answered all questions perfectly, stopped making mistakes, and produced fully working programs better than the best crack developer teams, that still doesn't mean it "understands". "Understanding" is not an output, it's a process, that is sometimes (but not always) measured by its output.

If it stopped making mistakes and produced complete fully working programs, no, there would be no way to say it doesn't understand. Yes, "understanding" is a process, but it's not well defined. And anyway, if it's a requirement for those things, and the AI did those things, then the only possibility is that the AI has this process in some way. But well, our current AIs do not produce complete programs, nor fully work…

>> If it stopped making mistakes and produced complete fully working programs, no, there would be no way to say it doesn't understand.

There exist program synthesis systems that always produce "complete fully working programs" and that in fact cannot make any mistakes because they are based on algorithms with strong theoretical guarantees (like mathematical proofs) of their correctness. You give them a set of examples and some "background knowledge" and they spit out a program that's consistent with the examples and the background knowledge, no ifs and buts and maybes.

I could point you to several systems like that, if you wish, but first I want to make sure that we both agree what you are saying: a program synthesis system that never makes mistakes "understands". Is that right? Could you also please clarify what such a system "understands"?

Then I'm happy to link you to some systems like the ones I'm talking about.

Re: And yet It Understands

#165

Post by anonyfox »

Animats wrote: post #149

There's more of a model inside large language models than was previously thought. How much of a model? Nobody seems to know. There was that one result where someone found what looked like an Othello board in the neuron state. Someone wrote, below: > We know the basic architecture of large language models, but hardly anything about how they calculate anything specific. That’s the mystery. It will take research, not ca…

In quantum physics, we also don’t really understand anything („shut up and calculate“) still people build awesome stuff that works.

Humans learned how to use and create fire looooong before understanding what fire actually is! Just a few centuries ago, people believed that fire is its own element!

Feels kinda similar to people searching for „consciousness“ that „understands“ things as if it would be something special/magic… when it’s probably more like naturally emerging behaviors when scaling up neural networks?

Re: And yet It Understands

#166

Post by glenstein »

im3w1l wrote: post #128
glenstein wrote: post #65

Earlier quoted context omitted.

>means that even chairs and rockets share some attributes (helping you get higher), the difference between them is still qualitative, not quantitve. I don't think so, because the needed 'quality' is the ability to traverse space. So I don't think I agree that the qualitative piece is missing. Perhaps the moon example is helpful here because the real solution, a rocket ship, uses propulsion rather than sheer mass, and…

The quote isn't talking about physical distance. It's talking about "close" in the sense of progress. You haven't made any progress towards walking on the moon even if the pile of chairs is as big as Mount Everest. There isn't just the question of toppling either, the chairs would start physically crushing each other and then they are no longer chairs.

I think this misunderstands my point across the board. Physical distance is progress, and not only that, it's the most fundamental kind of progress for this type of problem. Chairs run up against practical limitations but not in-principle limitations, and people invoke the piling-up-chairs argument because they are confused about which kinds of arguments are about practicality and which are about illustrating a conceptual principle.

>You haven't made any progress towards walking on the moon even if the pile of chairs is as big as Mount Everest.

I literally gave an example where this is exactly the thing that does, in fact, happen.

>There isn't just the question of toppling either, the chairs would start physically crushing each other and then they are no longer chairs.

I guess this means you actually read to the end of my comment, so that's good, but, this introduces JV debate team metaphysics that has nothing to do with anything, and if it did matter, you could substitute out chairs for just about any other physical material where the property of chair-ness is moot and the point would be the same. Gravel? Let's go with gravel.

But I shouldn't even have to make the point because the fundamental insight is that the nature of the problem is about the traversibility of space.

Again, this quote and this argument have been around for decades now, and it has been used to support an a whole range of arguments, some of which are now being abandoned because they seem increasingly untenable with recent advancements.

If the people who used this example in the 1960s knew that, come 2023, people were no longer making the in-principle argument, and were even denying that it was ever ever about the possibility in principle, they would wonder what the hell was happening in 2023 that rendered this position of theirs no longer respectable.

Re: And yet It Understands

#167

Post by foldr »

_dain_ wrote: post #126
ivanbakel wrote: post #114

Earlier quoted context omitted.

>No, you're the one who's begging the question. Why can't a "purely statistical" process have an understanding? This is a related, but fundamentally different thing to the point I replied to in your original comment. You asked: > I don't see how it can [apply training across languages] unless it really has some kind of understanding. I provided a potential explanation that is in line with how we think GPT works, and…

What would convince you? Give an example of something that would show beyond a shadow of a doubt that it understands.

What would convince you that it didn’t?

Re: And yet It Understands

#168

Post by civilized »

famouswaffles wrote: post #162
civilized wrote: post #155

Earlier quoted context omitted.

I learned arithmetic laws from a textbook and practice problems. I didn't need a teacher gesticulating or any fancy multimodal stuff. It's symbolic manipulation. What is the machine missing that I was given? There's no post shifting. The research community has been setting itself realistically attainable benchmarks. Now that the research community has made a lot of progress against its benchmarks, we have hype, claim…

I don't care what you did. All due respect, you are one person. I care about what can observed by people in general. Dunno what to tell you other than textbook and practice problems is far from the solution you think it is for a big chunk of the population.

We have a machine that is dedicated to symbolic manipulation (prediction of new symbols based on past ones). It seems reasonable to compare it to a person that is good at symbolic manipulation. Why would we compare the machine to average people who mostly don't think that way?

Why would training a neural network architecture that lives and breathes symbols somehow yield an entity with intelligence like that of an average human? Most humans learn primarily from completely different sources.

Re: And yet It Understands

#169

Post by PoignardAzur »

YeGoblynQueenne wrote: post #151

Guys guys! Stop talking about LLMs a minute and look at this! I gave my phone's calculator app this very hard multiplication problem and it got it right! Look! 2398794857945873 * 10298509348503 = 2.47040112696963e+28 My calculator can do arithmetic! But only humans can do arithmetic! Therefore, my calculator must understand arithmetic! And I bet it always gets it right, too! That means it must understand arithmetic b…

Wow, being sarcastic about AI! I bet you're the first person to come up with that idea!

Man, I wish we could all be so smart and clear-seeing.

Re: And yet It Understands

#170

Post by glenstein »

datatrashfire wrote: post #158
glenstein wrote: post #65

Earlier quoted context omitted.

>means that even chairs and rockets share some attributes (helping you get higher), the difference between them is still qualitative, not quantitve. I don't think so, because the needed 'quality' is the ability to traverse space. So I don't think I agree that the qualitative piece is missing. Perhaps the moon example is helpful here because the real solution, a rocket ship, uses propulsion rather than sheer mass, and…

They are qualitatively different because the rocket can traverse that amount of space scalably. The chairs cannot.

Six stages of acceptance:

1. We'll never land on the moon.

2. It's hard to land on the moon.

3. I never said we can't land on the moon, you just can't do it with pile of chairs.

4. I was never denying that chairs illustrate a fundamental principle that illustrates why the problem is indeed solvable in principle.

5. I was never saying distinctions about what is or isn't 'qualitative' was intended to show that we can't in fact land on the moon.

6. I know, I've been saying we'll land on the moon this whole time! I've always disagreed with people who said otherwise despite coming to defense of arguments that were used for decades to make the point that I supposedly am not endorsing!

Post reply on HN