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

Why I'm still bearish on LLMs after Navier-Stokes

dank.systems

Post reply on HN
Print view View original

351–360 of 646 posts

Re: Why I'm still bearish on LLMs after Navier-Stokes

#351

Post by sobellian »

hackinthebochs wrote: post #346
sobellian wrote: post #83

Earlier quoted context omitted.

I tested both myself and a weak bot against Astra xhigh, https://lichess.org/study/27lCQqDa . It's still pretty bad at chess, though it takes longer to devolve into illegal moves.

So you weren't giving it an updated board state after every move? If you want to compare apples to apples, it should give an updated board state for each move, or you should play blindfolded.

I can play blindfolded. I am expert OTB (though I haven't played in a while). The game was like 18 moves of theory in the Maroczy Bind.

Re: Why I'm still bearish on LLMs after Navier-Stokes

#352

Post by nimbleal »

Planktonne wrote: post #254
nimbleal wrote: post #199

Earlier quoted context omitted.

Maybe practically it doesn’t matter? Perhaps AGI is not the model but the model plus everything it’s got access to. If we’re modelling intelligence in the way we seem to have to to have any coherent definition of AGI, it seems to me is always going to be more “intelligent” than alone.

That would mean we should consider any human with coding knowledge a chess grandmaster, which is obviously not the case.

My points is more that, while we have a strong intuition about where, as an entity, a human's boundaries are (i.e. where the person begins and ends), philosophically it' not immediately obvious that the analogy applies to the a model in the same way. Why should that be the line drawn that says this is the "thing" and this other stuff is external to the thing? It feels somewhat arbitrary.

Of course this is a difficult question with humans too, hence my reliance on intuition above. We don't have the same cultural/biological framework to fall back on with AI.

Re: Why I'm still bearish on LLMs after Navier-Stokes

#353

Post by hackinthebochs »

lelanthran wrote: post #265
hackinthebochs wrote: post #189

Earlier quoted context omitted.

We're not talking about learning the rules of chess here, but playing a competent game from just being shown the rules. Why is it so hard for people to keep track of the thread of discussion?

> We're not talking about learning the rules of chess here, but playing a competent game from just being shown the rules. Okay, lets go with that: it's the "shown the rules" bit that we are arguing about. The argument is that a human may play maybe a dozen games after learning the rules, after which they won't be inadvertently attempting illegal moves. What we are observing with SOTA models is that, even after seeing…

This is not good reasoning. Humans need at least dozens if not hundreds of reinforcement sessions to only make legal moves, and still occasionally fail (consider pins, discovered check, failing to respond to check). LLMs must one-shot a competent game after imbibing a mass of disconnected units of information about chess. Nothing about the two are similar.

See my comment here for more: https://news.ycombinator.com/item?id=49725306

Re: Why I'm still bearish on LLMs after Navier-Stokes

#354

Post by danpalmer »

_superposition_ wrote: post #323
yshklarov wrote: post #209

Great article, but the lack of sentence capitalization makes it unnecessarily difficult to read. Apologies if this comment is off-topic, but it really is quite egregious, and since the article was submitted by the author I presume they are open to the feedback.

Threw me off too. Like why???

It's a tech bro thing. Altman does it too, and I've worked with people in the past who do it.

I read it as "I'll take literally any conscience for myself no matter how minor, at any cost for you no matter how big".

Re: Why I'm still bearish on LLMs after Navier-Stokes

#355

Post by killerstorm »

"Language Models are Few-Shot Learners" - 2020, the GPT-3 paper.

It have been demonstrated that in-context learning is a very powerful mechanism. There's no evidence that models of the size of GPT-6 are bad at in-context learning. In fact, ARC-AGI-3 score might indicate they are good at it.

There's no evidence that a bespoke RL environment is required for each new skill - quite likely a good demonstration is sufficient.

Re: Why I'm still bearish on LLMs after Navier-Stokes

#356

Post by simianwords »

minraws wrote: post #142
modulus1 wrote: post #124

Earlier quoted context omitted.

I agree w/ this perspective. An agent with a harness that can run programs can solve a lot more than one without the harness. The AI system includes the harness, and it's not clear to me that AGI requires more than LLMs + code generation & execution are capable of.

So AI is AGI in fields where code can't solve anything? Is code omnipotent, I have been in software all my life and I would hard agree here. Sure stuff LLMs can do with being good at parts of code reproduction is incredible. And honestly it's the new way to do a lot of things but I have not see an iota of proof that it can scale across the board. For instance Maths is just code with different symbols and slightly les…

> AI is the best invention at figuring out or walking the search space and directionally doing logically computation over general software adjacent stuff.

So.. like chess?

Anyway, do you have any prediction on what LLM's can or can't do in a few years?

Re: Why I'm still bearish on LLMs after Navier-Stokes

#357

Post by hi_im_greg_h »

intrasight wrote: post #258
dotdi wrote: post #211

Earlier quoted context omitted.

The whole point of this post was that it's questionable what can be achieved without huge investments into oversight and steering, because navier-stokes was a topic with an unusual level of specification. The problem itself was a specification. Such situations are rare in real-world scenarios. AI agents are good at solving well-specified tasks, not at solving problems. They do well in fields where the cost/effort of…

> a topic with an unusual level of specification Solving cancer also has an unusual level of specification. Many real world problems have that characteristic.

This is so untrue its actually insane to me that anyone could believe this.

The real world is so messy, and specifically cancer/biology is insanely messy and certainly not well specified.

You should close chatgpt and read a book sometime.

Re: Why I'm still bearish on LLMs after Navier-Stokes

#358

Post by danpalmer »

keeda wrote: post #73

The premise in the very first point seems off: > the frontier labs are priced according to the narrative that they have produced or will in the very near future produce a fully automated drop-in replacement for most knowledge workers... Even assuming this is how the AI companies are being valued (they're not), the numbers are off. The "value" of most knowledge workers -- based on what enterprises currently pay for th…

> It's reasonable to assume that if AI drop-in-replaced all those knowledge workers, AI companies could credibly charge somewhere in that order of magnitude, because that's what the market is already bearing

This assumes you don't change the market, but at the scale of (checks notes...) "all knowledge work", that just doesn't hold.

For example if you put 1bn people out of work, you now need some sort of safety net to bail out much of that workforce, a truly unprecedented change. You also lose tens of trillions of dollars of tax revenue.

One solution might be to recoup that cost and lost tax revenue from businesses by raising corporation tax. If corporation tax went from low tens of percent to high tens of percent, would those businesses be able to afford all that AI? No. Same order of magnitude? I doubt it.

There are many possible futures there, but the simplification made in the parent comment is completely unrealistic. The article is right in calling out the valuations as crazy.

Re: Why I'm still bearish on LLMs after Navier-Stokes

#359

Post by Topfi »

hackinthebochs wrote: post #346
sobellian wrote: post #83

Earlier quoted context omitted.

I tested both myself and a weak bot against Astra xhigh, https://lichess.org/study/27lCQqDa . It's still pretty bad at chess, though it takes longer to devolve into illegal moves.

So you weren't giving it an updated board state after every move? If you want to compare apples to apples, it should give an updated board state for each move, or you should play blindfolded.

Blindfolded flex by OP aside (I can barely play when seeing the board), considering reasoning traces and their nature, if we want to be fair, a person would have to get the moves, but be allowed to write them down or draw up a board in their notepad. My working memory can barely handle five chunks, a models reasoning tokens are masses of written text in comparison.

Re: Why I'm still bearish on LLMs after Navier-Stokes

#360

Post by webern777 »

bluegatty wrote: post #114

"are priced according to the narrative that they have produced or will in the very near future produce a fully automated drop-in replacement for most knowledge workers, " No, they're really not. They're priced in a way that would imply AI will be universal form of compute, alongside traditional deterministic systems - which it will be. And that they will capture most of that ... which they won't. The Frontier Labs ar…

They are pricing as a first mover AGI monopoly that makes no sense.

I think the core mistake is this partial-equilibrium reasoning. Take the new technology and then hold everything else fixed.

$40 trillion of knowledge work routed unchanged through a new toll booth. Profit. This has nothing to do with reality.

Nvidia on the other hand does have the CUDA monopoly so their toll booth is printing money but that will get routed around or broken at some point.

Post reply on HN