[camjohnson26]

This is a popular take, but does it hold up to reality? From what I’ve seen most people have long expected AI to solve standardized tests, even more free form ones like the LSAT. LLMs’ new abilities are mostly just because of faster and cheaper training and huge amounts of data, but I don’t see anything it can solve that doesn’t use pattern matching.

There are many things that pattern matching over large amounts of data can solve, like eventually we can probably get fully generated movies, music compositions, and novels, but the problem is that all of the content of those works will have to have been formalized into rules before it is produced, since computers can only work with formalized data. None of those productions will ever have an original thought, and I think that’s why GPT-3’s fiction feels so shallow.

So it boils down to a philosophical question, can human thought be formalized and written in rules? If it can, no human ever has an original thought either, and it’s a moot point.

[burlesona]

I agree with your take, but will emphasize that the recent wave of AI progress has me questioning how much of human intelligence just reduces to pattern matching. There's certainly a lot of things, like painting, that most people wouldn't have called "pattern matching" a few years ago and now seem to clearly fall into that category.

[oliveshell]

This reminds me of how I felt when I was 14 years old and I discovered what oxytocin was on an episode of Boston Legal.

The fact that feelings of love and closeness could be prompted by a mere chemical was deeply saddening to me. It wrecked my worldview.

"Love is just the result of some chemical? Then it's not even real!" I thought to myself.

Fast-forward ~20 years later, and that's proven to be an obvious— and massive— and useless— oversimplification.

Of course love isn't "just a reaction caused by a chemical." It's a fantastically complex emergent property of our biological system that we still absolutely do not understand.

It's the same with thinking: are parts of it analogous to pattern matching? Sure! Is this the whole story? Not even close.

[chairhairair]

Is love just a (complicated) biochemical reaction? Of course not! But also yes, of course it is.

[somenameforme]

There's one rather extreme difference. Humanity went from a domain where there was literally no such thing as painting, to the Mona Lisa. Once there is an extremely large and well established body of course one can create,in literally any field, solely by mimicry, but "intelligence" is what enables us to go from nothing to something. And that remains completely absent in any any sort of "AI" of today.

[pnt12]

Contrarian view: I think you need to be critical about which patterns to match. Eg if my inputs are a book on astronomy and one of conspiracy theories, how do I answer "Is the Earth flat?".

Now contrarian to the contrarian view: many of us live in bubble echos and go for the popular opinion instead of critical thinking, so maybe that's a bar too high even for humans.

[gowld]

> you need to be critical about which patterns to match

and how do you do that? By pattern-matching on "high-quality source"

[danaris]

The difference is, every human is capable of critical thinking, whether or not they have been educated to do so or choose to make use of it.

LLMs do not have that capability, fundamentally.

[ryandvm]

I agree. Try formulating a sentence backwards in your head and you'll realize that most of the speaking that HUMANS do is just figuring out the next token.

[bobthepanda]

Making existing art, or art similar to existing art, might be pattern matching.

Making totally new innovations in art, particularly ones that people end up liking, is a whole different ball game.

[mlsu]

I mean, the data has to come from somewhere.

Look at something like [Luncheon on the Grass](https://en.wikipedia.org/wiki/Le_D%C3%A9jeuner_sur_l%27herbe)

This painting was revolutionary. When it was first exhibited in Paris, people were shocked. It was rejected from the Salon (the most prominent art exhibition at the time). Yet, 10 years later, every painting in the Salon resembled it. And you can draw a line from this painting, to Monet, from which you can draw a line to Picasso, from which you can draw a line to Pollock....

Obviously, none of these are totally new innovations, they all came from somewhere. Pattern making.

The only difference between this and these language models is that Manet and artists like him use their rich sensory experience obtained outside of painting to make new paintings. But it's all fundamentally pattern matching in the end. As long as you can obtain the patterns, there's no difference between a human and a machine in this regard.

[bobthepanda]

Sure, in hindsight those things have a line between them, but a lot of art is also based on rejection of existing patterns.

A urinal and some soup cans are very mundane objects, and yet were the start of some notable art movements and careers.

[mlsu]

Duchamp, quoted on why he wrote what he wrote on fountain:

> Mutt comes from Mott Works, the name of a large sanitary equipment manufacturer. But Mott was too close so I altered it to Mutt, after the daily cartoon strip "Mutt and Jeff" which appeared at the time, and with which everyone was familiar. Thus, from the start, there was an interplay of Mutt: a fat little funny man, and Jeff: a tall thin man... I wanted any old name. And I added Richard [French slang for money-bags]. That's not a bad name for a pissotière. Get it? The opposite of poverty. But not even that much, just R. MUTT.

Why did he choose "Mutt" after reading the strip, and not before? Why did he make the piece after moving to the US, and not before? Why was fountain made only a few short years after economies were industrialized, and not before (or 100 years later?)

[bobthepanda]

The point is, can an AI point out novel things well? All these little things add up to make it novel, and the search space for all the possible combinations of little things is infinite, when only a select few will click with the public at any given time.

[Rant423]

>is a whole different ball game.

I was thinking the same: can a (future) model be like Leonardo or Beethoven, and actually innovate?

Assuming that what Beethoven did is not "just" making music similar to pre-existing music.

And yes, I'm aware the bar was raised from "average human" to Beethoven.

[tjr]

I remember reading the biography of a 20th century musician/composer, who said something to the effect of -- "Sure, I can sit down and write 4-part cantatas like Bach did, but that doesn't mean that I'm as great of a composer as Bach. What made Bach so great was that he was the one who figured out how to put these things together in the first place. Once he did that, copying the approach is no big deal."

It seems to me we're at a similar place now with AI tools. If you provided an AI tool with all music written _prior to_ Bach, would that tool take those inputs and create something new along the lines of what Bach did?

Or if provided input of all music up through the 1920s, would it create bebop? Or if provided music through the 1940s, would it create hard bop? Or if provided music through the 1970s, would it create music like Pat Metheny?

On one hand, being able to create more of the same sort of music that already exists is a very respectable thing, and what today's AI tools can do is utterly amazing. It takes human composers time and effort to be able to learn to write music that is certainly not innovative, but just matching the state of the art. And there's certainly a commercial market for churning out more of the same.

But in terms of asking, how close are these tools to human intelligence?, I think this is one legitimate area to bring up.

[camjohnson26]

Granted these are exceptional humans, but they are extreme examples of a capability that all humans have, but no machine has, which is coming up with something new.

People underestimate the impact that innovations, true ones not the Silicon Valley buzz words, have had on the world. Einstein’s theories were not inevitable, neither was Plato, democracy, or most of the other big impactful ideas of history. But we’re all conditioned to accept the lie of inevitable scientific progress, without justifying why things must always get better and more advanced. On the contrary, the collapse of many great civilizations shows that things often get much worse, quickly.

[slingnow]

Can you explain how this is a whole different ballgame?

It seems to me that making art that people like is a combination of pattern matching, luck, the zeitgeist, and other factors. However it doesn't seem like there's some kind of unknowable gap between "making similar art" and "making innovations in art that people like". I'm of the opinion that all art is in some sense derivative in that the human mind integrates everything it has seen and produces something based on those inputs.

[bobthepanda]

Luck and the zeitgeist are pretty important. Without those, you have a lot of noise and are basically throwing things at the wall until it sticks.

A urinal, and some supermarket soup cans, represent pretty pivotal art movements. It’s not clear what makes those two things more art than others, and even to people at the time it wasn’t super clear.

[chordalkeyboard]

"Good artists copy, great artists steal" -Picasso

All art is derivative.

[jimbokun]

> but I don’t see anything it can solve that doesn’t use pattern matching.

Do you have evidence that human brains are not just super sophisticated pattern matching engines?

Humans read novels, listen to compositions, watch movies, and make new ones similar in some ways and different in other ways. What is fundamentally different about the process used for LLMs? Not the current generation necessarily, but what's likely to emerge as they continue to improve.

[camjohnson26]

If you’re looking for proof you’re begging the question, asking for a formal proof of something that by definition can’t be proven, which only makes sense if your philosophical basis is that reality is a formal system. Other people have other philosophical bases, and while they may not be formally probable, they can be supported with other evidence that is equally strong, pointing to the non determinism of quantum physics or the infinitely recursive question of “what caused the first cause”.

The strongest evidence I have is that people are notoriously difficult to predict, individually.

[gowld]

Do pattern matching engines get out of bed in the morning and make breakfast?

[tspike]

If they have a body, and needs that they recognize they need to fill, sure.

[joshuahedlund]

Humans can ask questions and seek out information. LLMs can only respond to questions.

[camjohnson26]

LLMs can ask questions too.

[javajosh]

We are about to test the tests, so to speak, and discover whether an agent that aces a test is capable of doing "real work". Meaning information work you would normally pay a human to do. Paperwork stuff, managing accounts, but also programming and social media marketing. Anything mediated by a computer.

If so it means the union of all human expertise is a few gigabytes. Having seen both a) what we can do in a kilobyte of code, and b) a broad range of human behavior, this doesn't seem impossible. The more interesting question is: what are humans going to do with this remarkable object, a svelte pocket brain, not quite alive, a capable coder in ALL languages, a shared human artifact that can ace all tests? "May you live in interesting times," indeed.

[WastingMyTime89]

> but the problem is that all of the content of those works will have to have been formalized into rules before it is produced, since computers can only work with formalized data.

Clearly the key takeaway from GPT is that given enough unstructured data, LLM can produce impressive results.

From my point of view, the flaw in most discussion surrounding AI is not that people underestimate computers but overestimate how special humans are. At the end of day, every thoughts are a bunch of chemical potentials changing in a small blob of flesh.

[loandbehold]

Sounds like Chinese Room argument. Maybe human intelligence is just a pattern matching?

[tspike]

What would be an alternative explanation for our capabilities? It was once controversial (and still is in some circles) to say that humans are animals simply because it took away some of our sense of being "special."

[sirsinsalot]

We might consider certain humans to have had innovative or original thoughts.

It is probably true that at a given point many many people had the same or very similar ideas.

Those who execute or are in the right place and time to declare themselves the originator are the ones we think innovated.

It isn't true. Or rarely is true. History is written by the victor (and their simps)

[kordlessagain]

> can human thought be formalized and written in rules

No, and I think it's because human thought is based on continuous inferencing of experience, which gives rise to the current emotional state and feeling of it. For a machine to do this, it will need a body and the ability to put attention on things it is inferencing at will.

[luxcem]

The embodied cognition is still a theory, can consciousness appears in a simulated brain without a physical body? Maybe. What seems to be a limiting factor for now it's that current models don't experience existence, they don't have memory and don't "think" outside of the prompt. They are just instances of code launched and destroyed as soon as their task is done.

Right now it's possible to simulate memory with additional context (eg system prompt) but it doesn’t represent existence experienced by the model. If we want to go deeper the models need to actually learn from their interaction, update their internal networks and have some capabilities of self reflection (ie "talking to themselves").

I'm sure that's highly researched topic but it would demands extraordinary computational power and would cause lot of issues by letting such an AI in the wild.

[kordlessagain]

Embeddings via ada-002 give us a way to update the model in real time. Using Weaviate, or another dense vector engine, it is possible to write "memories" to the engine and then search those with concepts at a subsequent inferencing step. The "document models" that the engine stores can be considered a "hot model".

[bobvanluijt]

Yeah - it will become available as a multi2vec Weaviate module as well in due time.