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

GPT-5 is behind schedule

wsj.com

Post reply on HN
Print view View original

211–220 of 1001 posts

Re: GPT-5 is behind schedule

#211

Post by devit »

SpicyLemonZest wrote: post #129
bwhiting2356 wrote: post #120

I want AI to help me in the physical world: folding my laundry, cooking and farming healthy food, cleaning toilets. Training data is not lying around on the internet for free, but it's also not impossible. How much data do you need? A dozen warehouses full of robots folding and unfolding laundry 24/7 for a few months?

Laundry folding is an instructive example. Machines have been capable of home-scale laundry folding for over a decade, with two companies Foldimate and Laundroid building functional prototypes. The challenge is making it cost-competitive in a world where most people don't even purchase a $10 folding board. I would guess that most cooking and cleaning tasks are in basically the same space. You don't need fine motor co…

I think the problem of those is that they are special purpose, and probably too expensive and bulky for that single purpose.

A single general-purpose robot that can do everything would be much easier to sell.

Re: GPT-5 is behind schedule

#212

Post by imiric »

glenstein wrote: post #207
imiric wrote: post #197

Earlier quoted context omitted.

It's reasonable to ask for sources when an opinion is phrased as a fact, as GGP did. I don't see how you got that it was _unmistakably_ an opinion from that comment. There is no way to deduce by intuition alone that GPT-5 == GPT-4o. So either that person has some information the rest of us aren't privy to, or it's an opinion phrased as a fact. In either case, it deserves clarification.

On a second read I see that the comment notes that it is intended as speculation, but still it seems rather confident in its own accuracy and I am not even sure it's wrong, but just looking for something that warrants the confidence.

Ah, fair enough. I missed the speculation bit.

Re: GPT-5 is behind schedule

#213

Post by famouswaffles »

riku_iki wrote: post #200
famouswaffles wrote: post #167

Earlier quoted context omitted.

Not really. o3-low compute still stomps the benchmarks and isn't anywhere that expensive and o3-mini seems better than o1 while being cheaper. Combine that with the fact that LLM inference has reduced orders of magnitudes in cost the last few years and hampering over the inference costs of a new release seems a bit silly.

If you are talking about ARC benchmark, then o3-low doesn't look that special if you take into account there are plenty of finetuned models with much smaller resources achieved 40-50% results on private set (not semi-private like o3-low).

- I'm not just talking about ARC. On frontier Math, we have 2 scores, one with pass@1 and another with consensus vote with 64 samples. Both scores are much better than previous Sota.

- Also apparently, ARC wasn't a special fine-tune but rather some of the training set in the corpus for pre-training.

Re: GPT-5 is behind schedule

#215

Post by simonw »

xanderlewis wrote: post #206
simonw wrote: post #146

Earlier quoted context omitted.

No, I'm complaining that just because GPT-4 is called GPT-4 doesn't mean it's the fourth LLM from OpenAI. Off the top of my head: GPT-2, Codex, GPT-3 in three different flavors (babbage, curie, davinci), GPT-3.5. Suggesting that GPT-4 was "fourth" simply isn't credible. Just the other day they announced a jump from o1 to o3, skipping o2 purely because it's already the name of a major telecommunications brand in Europ…

It’s somehow funny to hear a British company being described as ‘in Europe’, but I suppose you’re technically correct…

https://en.wikipedia.org/wiki/O2_(brand) - "O2 (typeset as O2) is a global brand name owned by the Spanish telecommunications company Telefónica"

Re: GPT-5 is behind schedule

#216

Post by danielmarkbruce »

exhaze wrote: post #160
simonw wrote: post #140

Earlier quoted context omitted.

The new idea is inference-time scaling, as seen in o1 (and o3 and Qwen's QwQ and DeepSeek's DeepSeek-R1-Lite-Preview and Google's gemini-2.0-flash-thinking-exp). I suggest reading these two pieces about that: - https://www.aisnakeoil.com/p/is-ai-progress-slowing-down - best explanation I've seen of inference scaling anywhere - https://arcprize.org/blog/oai-o3-pub-breakthrough - François Chollet's deep dive into o3 I'…

I think the wildest thing is actually Meta’s latest paper where they show a method for LLMs reasoning not in English, but in latent space https://arxiv.org/pdf/2412.06769 I’ve done research myself adjacent to this (mapping parts of a latent space onto a manifold), but this is a bit eerie, even to me.

It's just concept space. The entire LLM works in this space once the embedding layer is done. It's not really that novel at all.

Re: GPT-5 is behind schedule

#217

Post by MichaelDickens »

dheera wrote: post #202
simonw wrote: post #113

"OpenAI’s is called GPT-4, the fourth LLM the company has developed since its 2015 founding." - that sentence doesn't fill me with confidence in the quality of the rest of the article, sadly.

Articles these days are probably written by ChatGPT

I doubt it, if you ask ChatGPT whether GPT-4 is OpenAI's fourth LLM, it gives the correct answer. That's the sort of thing GPT-2 might have said.

Re: GPT-5 is behind schedule

#218

Post by staunton »

KTibow wrote: post #188
viraptor wrote: post #127

Earlier quoted context omitted.

Editing mistakes that AI wouldn't make is the new "proof of human input".

I've been messing around with base (not instruction tuned) LLMs; they often evade AI detectors and I wouldn't be surprised if they evade this kind of detection too, at least with a high temperature

> with a high temperature

More like: with the right prompting

Re: GPT-5 is behind schedule

#219

Post by layer8 »

kevingadd wrote: post #192
layer8 wrote: post #183

Earlier quoted context omitted.

Counterexample: Everyone uses dishwashers. Yet I don’t think we’ll have a robot doing the dishes human-style, or even just filling up and clearing out a dishwasher, within the next decade or two, regardless of price.

Part of the tradeoff there is efficiency. I like my dishwasher because it's as good at getting things clean as I am but it does it using less water and less soap, and at scale, it takes less time too. It's just a great use case for machine automation because you can do clever stuff w/a dishwasher that's hard to replicate outside of that closed environment. I struggle to imagine a scenario where a 1-2 person household…

I think it would have to be a general-purpose robot, and doing the laundry would just be one of many things it can do, similar to how running a particular program is only one of many things a computer can do. More than that, I believe it would actually require a general-purpose robot to handle all contingencies that can arise in doing laundry.

As someone who does laundry about twice a week, it would certainly be nice. But it’s a pie in the sky at this time even just on the technological side.

Re: GPT-5 is behind schedule

#220

Post by ramesh31 »

t_serpico wrote: post #208

One fundamental challenge to me is that if each training run because more and more expensive, the time it takes it to learn what works/doesn't work widens. Half a billion dollars for training a model is already nuts, but if it takes 100 iterations to perfect it, you've cumulatively spent 50 billion dollars... Smaller models may actually be where rapid innovation continues simply because of tighter feedback loops. O3…

But if the scaling law holds true, more dollars should at some point translate into AGI, which is priceless. We haven't reached the limits yet of that hypothesis.
Post reply on HN