this post was submitted on 30 Sep 2024

196 points (93.0% liked)

Technology

35123 readers

276 users here now

This is the official technology community of Lemmy.ml for all news related to creation and use of technology, and to facilitate civil, meaningful discussion around it.

Ask in DM before posting product reviews or ads. All such posts otherwise are subject to removal.

Rules:

1: All Lemmy rules apply

2: Do not post low effort posts

3: NEVER post naziped*gore stuff

4: Always post article URLs or their archived version URLs as sources, NOT screenshots. Help the blind users.

5: personal rants of Big Tech CEOs like Elon Musk are unwelcome (does not include posts about their companies affecting wide range of people)

6: no advertisement posts unless verified as legitimate and non-exploitative/non-consumerist

7: crypto related posts, unless essential, are disallowed

founded 5 years ago

MODERATORS

[email protected]

196

Don’t believe the hype: AGI is far from inevitable (www.ru.nl)

submitted 2 months ago by [email protected] to c/[email protected]

47 comments fedilink hide all child comments

Will AI soon surpass the human brain? If you ask employees at OpenAI, Google DeepMind and other large tech companies, it is inevitable. However, researchers at Radboud University and other institutes show new proof that those claims are overblown and unlikely to ever come to fruition. Their findings are published in Computational Brain & Behavior today.

all 48 comments

sorted by: hot top controversial new old

[–] [email protected] 52 points 2 months ago (2 children)

It's a classic BigTech marketing trick. They are the only one able to build "it" and it doesn't matter if we like "it" or not because "it" is coming.

I believed in this BS for longer than I care to admit. I though "Oh yes, that's progress" so of course it will come, it must come. It's also very complex so nobody else but such large entities with so much resources can do it.

Then... you start to encounter more and more vaporware. Grandiose announcement and when you try the result you can't help but be disappointed. You compare what was promised with the result, think it's cool, kind of, shrug, and move on with your day. It happens again, and again. Sometimes you see something really impressive, you dig and realize it's a partnership with a startup or a university doing the actual research. The more time passes, the more you realize that all BigTech do it, across technologies. You also realize that your artist friend did something just as cool and as open-source. Their version does not look polished but it works. You find a KickStarter about a product that is genuinely novel (say Oculus DK1) and has no link (initially) with BigTech...

You finally realize, year after year, you have been brain washed to believe only BigTech can do it. It's false. It's self serving BS to both prevent you from building and depend on them.

You can build, we can build and we can build better.

Can we build AGI? Maybe. Can they build AGI? They sure want us to believe it but they have lied through their teeth before so until they do deliver, they can NOT.

TL;DR: BigTech is not as powerful as they claim to be and they benefit from the hype, in this AI hype cycle and otherwise. They can't be trusted.

[–] [email protected] 9 points 2 months ago (2 children)

It's one thing to claim that the current machine learning approach won't lead to AGI, which I can get behind. But this article claims AGI is impossible simply because there are not enough physical resources in the world? That's a stretch.

[–] [email protected] 6 points 2 months ago (1 children)

I haven't seriously read the article for now unfortunately (deadline tomorrow) but if there is one thing that I believe is reliable, it's computational complexity. It's one thing to be creative, ingenious, find new algorithms and build very efficient processors and datacenters to make things extremely efficient, letting us computer things increasingly complex. It's another though to "break" free of complexity. It's just, as far as we currently know, is impossible. What is counter intuitive is that seemingly "simple" behaviors scale terribly, in the sense that one can compute few iterations alone, or with a computer, or with a very powerful set of computers... or with every single existing computers... only to realize that the next iteration of that well understood problem would still NOT be solvable with every computer (even quantum ones) ever made or that could be made based on resources available in say our solar system.

So... yes, it is a "stretch", maybe even counter intuitive, to go as far as saying it is not and NEVER will be possible to realize AGI, but that's what their paper claims. It's a least interesting precisely because it goes against the trend we hear CONSTANTLY pretty much everywhere else.

[–] [email protected] 5 points 2 months ago

PS: full disclosure, I still believe self-hosting AI is interesting, cf my notes on it https://fabien.benetou.fr/Content/SelfHostingArtificialIntelligence but that doesn't mean AGI can be reached, even less that it'd be "soon". IMHO AI itself as a research field is interesting enough that it doesn't need grandiose claims, especially not ones leading to learned helplessness.

[–] [email protected] 1 points 2 months ago (1 children)

Maybe if they keep using digital computers. What they need is an analogue system. It's much more efficient for this kind of work.

[–] [email protected] 1 points 2 months ago

Saw a great video about this (project is still ongoing).

[–] [email protected] 7 points 2 months ago (1 children)

And the big tech companies also stand to benefit from overhyping their product to the point of saying it will take over the world. They look better for investors and can justify laws saying they should be the only arbiters of this technology to "keep it out of criminal hands" while happily serving the criminals for a fee.

[–] [email protected] 3 points 2 months ago

Indeed, AKA the OpenAI playbook.

[–] [email protected] 25 points 2 months ago (3 children)

You do all this on three pounds of wet meat powered by cornflakes.

The idea we'll never recreate it through deliberate effort is absurd.

What you mean is, LLMs probably aren't how we get there. Which is fair. "Spicy autocorrect" is a limited approach with occasionally spooky results. It does a bunch of stuff people insisted would never happen without AGI - but that's how this always goes. The products of human intelligence have always shown some hard-to-define qualities which humans can eventually distinguish from our efforts to make a machine produce anything similar.

Just remember the distinction got narrower.

[–] [email protected] 7 points 2 months ago (1 children)

I agree. Very few people in industry are claiming that LLMs will become AGI. The release of o1 demonstrates that even OpenAI are pivoting from pure LLM approaches. It was always going to be a framework approach that utilizes LLMs.

[–] [email protected] 1 points 2 months ago

I had hopes for recurrent systems becoming kinda... Dixie Flatline. Maybe not general enough to learn, but spooky enough to evaluate claims.

[–] [email protected] 2 points 2 months ago

"Spicy auto~~correct~~assume"

Ftfy

[–] [email protected] 19 points 2 months ago

The paper: https://link.springer.com/content/pdf/10.1007/s42113-024-00217-5.pdf

[–] [email protected] 18 points 2 months ago (2 children)

Obviously those claims are overblown lol, AIs literally cannot think. They are currently LLMs. They are impressive, sure, but anyone knows the technology knows that this is NOT AGI, and it is entirely possible we will never get AGI. It's also possible we will get AGI, but this ain't it. lol

[–] [email protected] 5 points 2 months ago* (last edited 2 months ago)

I think regardless of how realistic it is, it's definitely interesting how the owning class really likes the idea of a future with a class of sapient beings with no legal rights.

[–] [email protected] 14 points 2 months ago* (last edited 2 months ago) (2 children)

This is a silly argument:

[..] But even if we give the AGI-engineer every advantage, every benefit of the doubt, there is no conceivable method of achieving what big tech companies promise.’

That’s because cognition, or the ability to observe, learn and gain new insight, is incredibly hard to replicate through AI on the scale that it occurs in the human brain. ‘If you have a conversation with someone, you might recall something you said fifteen minutes before. Or a year before. Or that someone else explained to you half your life ago. Any such knowledge might be crucial to advancing the conversation you’re having. People do that seamlessly’, explains van Rooij.

‘There will never be enough computing power to create AGI using machine learning that can do the same, because we’d run out of natural resources long before we'd even get close,’ Olivia Guest adds.

That's as shortsighted as the "I think there is a world market for maybe five computers" quote, or the worry that NYC would be buried under mountains of horse poop before cars were invented. Maybe transformers aren't the path to AGI, but there's no reason to think we can't achieve it in general unless you're religious.

EDIT: From the paper:

The remainder of this paper will be an argument in ‘two acts’. In ACT 1: Releasing the Grip, we present a formalisation of the currently dominant approach to AI-as-engineering that claims that AGI is both inevitable and around the corner. We do this by introducing a thought experiment in which a fictive AI engineer, Dr. Ingenia, tries to construct an AGI under ideal conditions. For instance, Dr. Ingenia has perfect data, sampled from the true distribution, and they also have access to any conceivable ML method—including presently popular ‘deep learning’ based on artificial neural networks (ANNs) and any possible future methods—to train an algorithm (“an AI”). We then present a formal proof that the problem that Dr. Ingenia sets out to solve is intractable (formally, NP-hard; i.e. possible in principle but provably infeasible; see Section “Ingenia Theorem”). We also unpack how and why our proof is reconcilable with the apparent success of AI-as-engineering and show that the approach is a theoretical dead-end for cognitive science. In “ACT 2: Reclaiming the AI Vertex”, we explain how the original enthusiasm for using computers to understand the mind reflected many genuine benefits of AI for cognitive science, but also a fatal mistake. We conclude with ways in which ‘AI’ can be reclaimed for theory-building in cognitive science without falling into historical and present-day traps.

That's a silly argument. It sets up a strawman and knocks it down. Just because you create a model and prove something in it, doesn't mean it has any relationship to the real world.

[–] [email protected] 9 points 2 months ago (2 children)

This is a gross misrepresentation of the study.

That's as shortsighted as the "I think there is a world market for maybe five computers" quote, or the worry that NYC would be buried under mountains of horse poop before cars were invented.

That's not their argument. They're saying that they can prove that machine learning cannot lead to AGI in the foreseeable future.

Maybe transformers aren't the path to AGI, but there's no reason to think we can't achieve it in general unless you're religious.

They're not talking about achieving it in general, they only claim that no known techniques can bring it about in the near future, as the AI-hype people claim. Again, they prove this.

That's a silly argument. It sets up a strawman and knocks it down. Just because you create a model and prove something in it, doesn't mean it has any relationship to the real world.

That's not what they did. They provided an extremely optimistic scenario in which someone creates an AGI through known methods (e.g. they have a computer with limitless memory, they have infinite and perfect training data, they can sample without any bias, current techniques can eventually create AGI, an AGI would only have to be slightly better than random chance but not perfect, etc...), and then present a computational proof that shows that this is in contradiction with other logical proofs.

Basically, if you can train an AGI through currently known methods, then you have an algorithm that can solve the Perfect-vs-Chance problem in polynomial time. There's a technical explanation in the paper that I'm not going to try and rehash since it's been too long since I worked on computational proofs, but it seems to check out. But this is a contradiction, as we have proof, hard mathematical proof, that such an algorithm cannot exist and must be non-polynomial or NP-Hard. Therefore, AI-learning for an AGI must also be NP-Hard. And because every known AI learning method is tractable, it cannor possibly lead to AGI. It's not a strawman, it's a hard proof of why it's impossible, like proving that pi has infinite decimals or something.

Ergo, anyone who claims that AGI is around the corner either means "a good AI that can demonstrate some but not all human behaviour" or is bullshitting. We literally could burn up the entire planet for fuel to train an AI and we'd still not end up with an AGI. We need some other breakthrough, e.g. significant advancements in quantum computing perhaps, to even hope at beginning work on an AGI. And again, the authors don't offer a thought experiment, they provide a computational proof for this.

[–] [email protected] 3 points 2 months ago* (last edited 2 months ago) (2 children)

Hey! Just asking you because I'm not sure where else to direct this energy at the moment.

I spent a while trying to understand the argument this paper was making, and for the most part I think I've got it. But there's a kind of obvious, knee-jerk rebuttal to throw at it, seen elsewhere under this post, even:

If producing an AGI is intractable, why does the human meat-brain exist?

Evolution "may be thought of" as a process that samples a distribution of situation-behaviors, though that distribution is entirely abstract. And the decision process for whether the "AI" it produces matches this distribution of successful behaviors is yada yada darwinism. The answer we care about, because this is the inspiration I imagine AI engineers took from evolution in the first place, is whether evolution can (not inevitably, just can) produce an AGI (us) in reasonable time (it did).

The question is, where does this line of thinking fail?

Going by the proof, it should either be:

That evolution is an intractable method. 60 million years is a long time, but it still feels quite short for this answer.
Something about it doesn't fit within this computational paradigm. That is, I'm stretching the definition.
The language "no better than chance" for option 2 is actually more significant than I'm thinking. Evolution is all chance. But is our existence really just extreme luck? I know that it is, but this answer is really unsatisfying.

I'm not sure how to formalize any of this, though.

The thought that we could "encode all of biological evolution into a program of at most size K" did made me laugh.

[–] [email protected] 8 points 2 months ago* (last edited 2 months ago) (1 children)

If producing an AGI is intractable, why does the human meat-brain exist?

Ah, but here we have to get pedantic a little bit: producing an AGI through current known methods is intractable.

The human brain is extremely complex and we still don't fully know how it works. We don't know if the way we learn is really analogous to how these AIs learn. We don't really know if the way we think is analogous to how computers "think".

There's also another argument to be made, that an AGI that matches the currently agreed upon definition is impossible. And I mean that in the broadest sense, e.g. humans don't fit the definition either. If that's true, then an AI could perhaps be trained in a tractable amount of time, but this would upend our understanding of human consciousness (perhaps justifyingly so). Maybe we're overestimating how special we are.

And then there's the argument that you already mentioned: it is intractable, but 60 million years, spread over trillions of creatures is long enough. That also suggests that AGI is really hard, and that creating one really isn't "around the corner" as some enthusiasts claim. For any practical AGI we'd have to finish training in maybe a couple years, not millions of years.

And maybe we develop some quantum computing breakthrough that gets us where we need to be. Who knows?

[–] [email protected] 2 points 2 months ago

Ah, but here we have to get pedantic a little bit: producing an AGI through current known methods is intractable.

I didn't quite understand this at first. I think I was going to say something about the paper leaving the method ambiguous, thus implicating all methods yet unknown, etc, whatever. But yeah, this divide between solvable and "unsolvable" shifts if we ever break NP-hard and have to define some new NP-super-hard category. This does feel like the piece I was missing. Or a piece, anyway.

e.g. humans don't fit the definition either.

I did think about this, and the only reason I reject it is that "human-like or -level" matches our complexity by definition, and we already have a behavior set for a fairly large n. This doesn't have to mean that we aren't still below some curve, of course, but I do struggle to imagine how our own complexity wouldn't still be too large to solve, AGI or not.

Anyway, the main reason I'm replying again at all is just to make sure I thanked you for getting back to me, haha. This was definitely helpful.

[–] [email protected] 1 points 2 months ago

That's a great line of thought. Take an algorithm of "simulate a human brain". Obviously that would break the paper's argument, so you'd have to find why it doesn't apply here to take the paper's claims at face value.

[–] [email protected] 0 points 2 months ago

There's a number of major flaws with it:

Assume the paper is completely true. It's just proved the algorithmic complexity of it, but so what? What if the general case is NP-hard, but not in the case that we care about? That's been true for other problems, why not this one?
It proves something in a model. So what? Prove that the result applies to the real world
Replace "human-like" with something trivial like "tree-like". The paper then proves that we'll never achieve tree-like intelligence?

IMO there's also flaws in the argument itself, but those are more relevant

[–] [email protected] 3 points 2 months ago

but there's no reason to think we can't achieve it

They provide a reason.

Just because you create a model and prove something in it, doesn't mean it has any relationship to the real world.

What are we science deniers now?

[–] [email protected] 13 points 2 months ago (1 children)

The problem is when your boss believes in hype and makes layoffs (already happenning)

[–] [email protected] 5 points 2 months ago

It's literally insane that they are doing this even though they don't even have the replacement. It really shows their colour.

[–] [email protected] 12 points 2 months ago (1 children)

Of course, as everywhere else, the LLMentalist comment is here!

[–] [email protected] 7 points 2 months ago

Read few months ago, warmly recommended. Basically on self selection bias and sharing "impressive" results while ignoring whatever does not work... then claiming it's just the "beginning".

[–] [email protected] 6 points 2 months ago (1 children)

Elon Musk was Steve Jobs, Thomas Edison was Nikola Tesla, more examples I’m sure, and Sam Altman IS ELON MUSK.

To paraphrase Göring: smarty-man hype & their promises work the same in every decade

[–] [email protected] 2 points 2 months ago (1 children)

And I say this disbelief as a loony-sort who believes that insects maybe have language as complex as humans’, and that AGI will probably happen someday & potentially while people living now are alive.

But I just look at they hypesters who think they could ever control such a mind, who obviously plan to, & they just seem like goofy carnival types playing at summoning a god, & when the real thing shows up it is NOT happy these ants were so presumptuous. Most of us aren’t, any AGI who might eventually be evaluating us, Star-Trek-Q-style!

[–] [email protected] 1 points 1 month ago

update, I no longer really believe that people alive now will live to see AGI

[–] [email protected] 4 points 2 months ago

To be honest I really think that an AI surprising human brain in many ways is a matter of time, but what people don't tend to talk about is whether or not we are slowly approaching the limit what we can do with technology, because I already see tech progress slowing down in some areas.

[–] [email protected] 2 points 2 months ago

I'm sure AGI is far off and AGI is impossible is exactly what AGI wants us to think.

[–] [email protected] 1 points 2 months ago (1 children)

Man I dunno if we’ve ever been here before. Nope.

[–] [email protected] 6 points 2 months ago (1 children)

Instead of linking to a jpeg hosted on a non-HTTPS website for a weird investments scam you could just link wikipedia:

https://en.wikipedia.org/wiki/Gartner_hype_cycle

[–] [email protected] 1 points 2 months ago

Sure thanks for finding a better link. Here’s a cookie. 🍪

[–] [email protected] 1 points 2 months ago* (last edited 2 months ago)

It better not be inevitable...

[Loads bolter with religious intent]

[–] [email protected] 0 points 2 months ago (1 children)

I liked SCI better

[–] [email protected] 1 points 2 months ago (1 children)

SCI?

[–] [email protected] 3 points 2 months ago (3 children)

SCI

I looked this up because it's new to me. AGI is what you think it is, and superintelligent collective intelligence is a collection that can perform tasks. Instead of 1 LLM or 1 AGI doing all the work, you have a team of agents and humans who can talk to each other. AGI seems like far off space tech and SCI is more like a next gen pursuit.

[–] [email protected] 2 points 2 months ago (1 children)

I meant like King's Quest 5

[–] [email protected] 1 points 2 months ago (1 children)

Lol did you really?

[–] [email protected] 1 points 2 months ago (1 children)

Yes but you should really ask me about LOOM

[–] [email protected] 1 points 2 months ago

What's loom?

[–] [email protected] 2 points 2 months ago

You may be interested in some history of computer-aided collaboration: https://www.quora.com/Who-invented-the-modern-computer-look-and-feel/answer/Harri-K-Hiltunen

[–] [email protected] 1 points 2 months ago

Cool... unfortunately my search is not turning up much. SCI.ai a science geared LLM was about 30% of my search results over Google and DDG. The other 70% is about Sierra's Creative Interpreter, and a moisturizer additive had 2 hits.

Glad you gave me the synopsis, apparently I'm incapable of finding that info myself, regardless of what combination of AI/AGI/SCI/DIFFERENCES etcetcetc and nauseum.