Comment by intenex | Hacker News Reader

Comment by intenex | Hacker News Reader

I'm surprised I'm not seeing anyone bring up the fact that the internal models used by the frontier labs are light years ahead of the public models we have - in the latest METR report, we're three generations behind - there's still an unreleased model from July!

And the latest model that they measured, which is already behind the newest model they're now working on, has an insane jump in capabilities: https://www.reddit.com/r/singularity/comments/1xorf5c/metr_r...

This is easy for us to empirically verify with some back of the envelope math - OpenAI's internal model that solved Navier-Stokes ~2 months ago required 10,000 agents running for 88 hours, or 880,000 total hours, to solve it. A month later, when they dropped the 722 math proofs with some truly stunning results like UGC in there, it took just a single agent an average of 3 hours to get those results. Granted, the results aren't on the same caliber, but some of them genuinely are insane - and even if you knock off multiple digits off the 300,000x speedup there, you're still getting some insane acceleration in capabilities.

And then of course ETH was disproved, and Anthropic claimed it took just 14 million output tokens to do this with their internal research model.

Put simply, the models the frontier labs are using to speed up R&D autonomously that are making them believe RSI is imminent look nothing like the outdated models we have publicly available to us. Saying that the models today don't look like they would do autonomous R&D that well says nothing about how the internal models would perform on similar tasks - in the same way that someone looking at older LLMs and saying they don't code very well autonomously would be mind-blown to see how autonomous coding has turned out today.

No they are not. If they were really getting good at that kind of speed Ai of Meta, Google and Grok etc would have also gotten very good at least matching the current frontier as all the ai dev engineers have moved from oen of these companies to another or build their own companies.

Aside from everything else, you can't just add together the hours from an agent swarm, because swarms become less cost-effective at scale. There's a nice study of that here: https://www.tobyord.com/writing/swarm-scaling

With the most optimistic scaling law from that source, a swarm of 10,000 only gives you a ~500x improvement in performance, and with the most pessimistic only a ~90x improvement. The point of a swarm is to get the same result faster, but more expensively (which OpenAI wanted because of the race dynamics around Navier-Stokes).

This alone knocks 1-2 OOMs off your supposed improvement, and the rest may easily be the different inherent difficulty of the constructions, which is all over the place -- in the big dump, one result (Quasi-RH) took more effort than the rest, and already with Astra people are finding crazy results. Right now, it's easier to judge problem difficulty in terms of the AI time needed to solve them (And whether they can be solved! Most still can't!) than to judge AIs based on which problems they can solve.

If it’s not public, it’s PR.

I’m totally willing to believe the people making these claims are honest, but until it can be replicated, it’s as good as a lie.

Then why haven't they built an open source replacement for MacOS and iOS yet?

What would be the point of that?

Show-off to investors.

I think solving Millenium problems is more compelling, and arguably less work for OpenAI than a feature parity iOS clone.

Except everyone can see and feel an operating system, unlike the exotic math 'proofs' that may or may not be verified at some point.

> I'm surprised I'm not seeing anyone bring up the fact that the internal models used by the frontier labs are light years ahead of the public models we have

I'm surprised anyone actually believes it.

Why?

Because we see the products they release even after the breathless "we do 30000 PRs per day" marketing posts.

No, the labs' models are not light-years ahead. That's what labs claim with literally no independent verification of these claims. They want you (and investors, and politicians) to believe these models are light-years ahead.

What they do have is unlimited and unrestricted access to their own models.

OpenAI sat on GPT-4 for 8 months before actually releasing it and that was a major stride over 3. It's kind of silly not to realize they do indeed have access to models months ahead of the public at minimum. Nothing you have access to is the true frontier. It only takes some common sense to acknowledge that. It's not unbelieveable, you just don't want to believe it.

> OpenAI sat on GPT-4 for 8 months before actually releasing it and that was a major stride over 3.

OpenAI said they sat on it for 8 months

> It's kind of silly not to realize they do indeed have access to models months ahead of the public at minimum

Yes, they do. Nothing in this sentence implies that what they have access to is "light years ahead of the public models".

>OpenAI said they sat on it for 8 months

They sat in it. We had concrete details ('rumors' that turned out surprisingly accurate) months before the public release.

>Yes, they do. Nothing in this sentence implies that what they have access to is "light years ahead of the public models".

"Light years" is subjective, but the accomplishments of those internal models we're priven to should be enough to give you some indication. You can fire up as many Astra agents as you like you're not going to get the performance of Open AI's internal model in math. Just as you can fire as many GPT-4 instances and you're not going to get anything resembling Opus 5.5.

> They sat in it. We had concrete details ('rumors' that turned out surprisingly accurate) months before the public release.

So all you had was rumours then "confirmed" by OpenAI marketing

> "Light years" is subjective

Yes. And this subjective description invokes a certain image that I objected to.

> but the accomplishments of those internal models we're proven to should be enough to give you some indication.

Yes. The indication is: there are some improvements, with diminishing returns [1]. A lot of perceived improvements now come from tooling, harnesses etc.

And the people working at the labs have access to unrestricted models and unlimited compute that will not flag you for "bio research" and will not stop running after 5 hours, which adds to the marketing myths.

> You can fire up as many Astra agents as you like you're not going to get the performance of Open AI's internal model in math.

Of course I won't. I don't have the same access.

Doesn't mean any of their models are light years ahead. And as for maths, again, you literally have no idea what they were actually doing besides their marketing materials. But even their marketing materials let slip some details that amount to "commandeered a full cluster for 20 000 hours and let it rip".

Edit: Moreover. We know they are not light years ahead, or substantially better than what users want. While they let some of their researchers do cool shit with unlimited resources, you can see the rest of what these labs produce: from their websites, to their harnesses, to their apps. Buggy incoherent mess. Doesn't scream "light years ahead" to me.

[1] Subjectively the models haven't moved an inch since February this year, and regressed in quite a few areas (code duplication is worse, Claude-ish permeates everything, context compaction is even more random and training quickly overrides context etc.)

No it wasn't. GPT-4 was indistinguishable from 3.5.

It's really easy to understand, what they have are the same model with less restrictions and gaurdrails, they also have infinite amount of usage and no worry about cost, they can burn a million dollars for an article in the news.

Reply on news.ycombinator.com