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

OpenAI declares 'code red' as Google catches up in AI race

theverge.com

Post reply on HN
Print view View original

671–680 of 960 posts

Re: OpenAI declares 'code red' as Google catches up in AI race

#671

Post by yalok »

nodja wrote: post #514
MikeTheGreat wrote: post #444

(My apologies if this was already asked - this thread is huge and Find-In-Page-ing for variations of "pre-train", "pretrain", and "train" turned up nothing about this. If this was already asked I'd super-appreciate a pointer to the discussion :) ) Genuine question: How is it possible for OpenAI to NOT successfully pre-train a model? I understand it's very difficult, but they've already successfully done this and they…

> Genuine question: How is it possible for OpenAI to NOT successfully pre-train a model? The same way everyone else fails at it. Change some hyper parameters to match the new hardware (more params), maybe implement the latest improvements in papers after it was validated in a smaller model run. Start training the big boy, loss looks good, 2 months and millions of dollars later loss plateaus, do the whole SFT/RL sheba…

add to it multiple iterations of having to restart pretraining from some earlier checkpoint when loss plateaus too early or starts increasing due to some bugs…

Re: OpenAI declares 'code red' as Google catches up in AI race

#672

Post by tomp »

brazukadev wrote: post #528
tomp wrote: post #499

Earlier quoted context omitted.

I don't know man. My code runs in 0.11s Gemini's code runs in 0.5s. Boss wants an explanation. ¯\_(ツ)_/¯

0.11s is faster than 0.5s

Yeah that’s the point. Now instead of just writing good code, I’m also supposed to debug shitty AI code.

Re: OpenAI declares 'code red' as Google catches up in AI race

#673

Post by curioussquirrel »

This is probably not a core concern for most HN readers, but at work we do multilingual testing for synthetic text data generation and natural language processing. Emphasis on multilingual. Gemini has made some serious leaps from 1.5 to 2.5 and now 3.0, and is actually proficient in languages that other models can only dream of. On the other hand, GPT-5 has a really mixed performance in a lot of categories.

Re: OpenAI declares 'code red' as Google catches up in AI race

#674

Post by chroma205 »

energy123 wrote: post #472
spiralpolitik wrote: post #361

Earlier quoted context omitted.

There is no moat in the models. The moat is in the UX. The problem is that OpenAI is far away from where the user is and not going to get there anytime soon. Google meanwhile is exactly where the user is. OpenAI IMHO is a dead company at this point. They are overvalued relative to the fundamentals and don't appear to have any way of getting the numbers to work in the timeframe that their investors will expect. They a…

> There is no moat in the models. I guess that's mostly true, but why does Jane Street get to have a moat in models but LLM companies can't? It feels like a structurally similar situation. The critical mass of research talent is somewhat of a moat in itself.

> I guess that's mostly true, but why does Jane Street get to have a moat in models but LLM companies can't?

Common misconception by people outside quant trading.

Modern “alpha” in trading does not come from better models but rather business connections with exchanges and regulators for preferential fees and/or revenue agreements.

If you are a “lead market maker” like Jane Street for ETFs, you can effectively skip the FIFO queue that retail traders and large passive index funds (VTI, VOO) must wait in.

Citadel has exclusive contracts to execute PFOF trades with e.g. Schwab. Even a simple exponential moving average model can be profitable with such a business arrangement.

OpenAI and Sam Altman tried to cut a deal (threaten?) with the US government, but looks like US government called Sam’s bluff.

Re: OpenAI declares 'code red' as Google catches up in AI race

#677

Post by fragmede »

shridharxp wrote: post #658
felixfurtak wrote: post #324

Earlier quoted context omitted.

This is pretty much all that OpenAI is at the moment. Mozilla is a non-profit that is only sustained by the generous wealthy benefactor (Google) to give the illusion that there is competition in the browser market. OpenAI is a non-profit funded by a generous wealthy benefactor (Microsoft). Ideas of IPO and profitability are all just pipe dreams in Altmans imagination.

Few months ago, the founder was talking about "AGI" and ridiculous universal basic compute. At this point, I don't even know whom to believe. My first hand experience tells ChatGPT and even ClaudeCode are no where near the expertise they are touted to be. Yet, the marketing by these companies is so immense that you get washed away, you don't know who are agents and who are putting their true opinions.

> My first hand experience tells ChatGPT and even ClaudeCode are no where near the expertise they are touted to be

Not doubting you, but where specifically have the latest models fallen short for you?

Re: OpenAI declares 'code red' as Google catches up in AI race

#678

Post by krackers »

mvkel wrote: post #550
jdoliner wrote: post #274

I've seen a rumor going around that OpenAI hasn't had a successful pre-training run since mid 2024. This seemed insane to me but if you give ChatGPT 5.1 a query about current events and instruct it not to use the internet it will tell you its knowledge cutoff is June 2024. Not sure if maybe that's just the smaller model or what. But I don't think it's a good sign to get that from any frontier model today, that's 18 m…

It's not a rumor, it's confirmed by OpenAI. All "models" since 4o are actually just optimizations in prompting and a new routing engine. The actual -model- you are using with 5.1 is 4. Nothing has been pre-trained from scratch since 4o. Their own press releases confirm this. They call 5 their best new "ai system", not a new model https://openai.com/index/introducing-gpt-5/

I can believe this, Deepseek V3.2 shows that you can get close to "gpt-5" performance with a gpt-4 level base model just with sufficient post-training.

Re: OpenAI declares 'code red' as Google catches up in AI race

#679

Post by a_victorp »

TheOtherHobbes wrote: post #473
littlestymaar wrote: post #360

Earlier quoted context omitted.

I'd be a little bit more nuanced: I think there's something off with their plans right now: it's pretty clear at this point that they can't own the technological frontier, Google is just too close already and from a purely technological PoV they are much better suited to have the best tech in the medium term. (There's no moat and Google has way more data and compute available, and also tons of cash to burn without de…

Altman's main interest is Altman. ChatGPT will be acquihired, most people will be let go, the brand will become a shadow of its former self, and Altman will emerge with a major payday and no obvious dent in his self-made reputation as a leading AGIthinkfluenceretc. I don't think ads are that easy, because the hard part of ads isn't taking money and serving up ad slop, it's providing convincing tracking and analytics.…

This! Most people that don't work on adtech have no idea how hard it is to: 1. Build a platform that offers new advertising inventory that advertisers can buy 2. Convince advertisers to advertise on your platform 3. Show advertisers that their advertising campaigns in your platform are more successful than in the several other places they can advertise

Re: OpenAI declares 'code red' as Google catches up in AI race

#680

Post by ares623 »

teaearlgraycold wrote: post #602
dwa3592 wrote: post #32

why couldn't GPT5.1 improve itself? Last I heard, it can produce original math and has phd level intelligence.

I believe it was Sam Altman that said software engineers wouldn't have jobs by the end of the year. They still have a few weeks to make good on that.

I bet my retirement funds on that promise!
Post reply on HN