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

521–530 of 960 posts

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

#521

Post by MikeTheGreat »

encomiast wrote: post #457
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…

I’m not sure what ‘successfully’ means in this context. If it means training a model that is noticeably better than previous models, it’s not hard to see how that is challenging.

Ah. Thanks for posting - this makes a lot of sense.

I can totally see how they're able to pre-train models no problem, but are having trouble with the "noticeably better" part.

Thanks!

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

#522

Post by AdieuToLogic »

scotty79 wrote: post #475
danielodievich wrote: post #469

Last week there we had a customer request that landed in our support on a feature that I partially wrote and wrote a pile of public documentation on. Support engineer ran customer query through Claude (trained on our public and internal docs) and it very, very confidently made a bunch of stuff up in the response. It was quite plausible sounding and it would have been great if it worked that way, but it didn't. While…

Yeah, LLMs are not really good about things that can't be done. At some point you'll be better off with implementing features they hallucinated. Some people with public APIs already took this approach.

>> Support engineer ran customer query through Claude (trained on our public and internal docs) and it very, very confidently made a bunch of stuff up in the response.

> Yeah, LLMs are not really good about things that can't be done.

From the GP's description, this situation was not a case of "things that can't be done", but instead was the result of a statistically generated document having what should be the expected result:

  It was quite plausible sounding and it would have been 
  great if it worked that way, but it didn't.

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

#523

Post by MikeTheGreat »

octoberfranklin wrote: post #454
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…

You don't train the next model by starting with the previous one. A company's ML researchers are constantly improving model architecture. When it's time to train the next model, the "best" architecture is totally different from the last one. So you have to train from scratch (mostly... you can keep some small stuff like the embeddings). The implication here is that they screwed up bigly on the model architecture, and…

Huh - I did not know that, and that makes a lot of sense.

I guess "Start software Vnext off the current version (or something pretty close)" is such a baseline assumption of mine that it didn't occur to me that they'd be basically starting over each time.

Thanks for posting this!

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

#524

Post by rprend »

rglullis wrote: post #311
bgirard wrote: post #290

Earlier quoted context omitted.

It's a fun trope to repeat but that's not what OpenAI is doing. I get a ton of value from ChatGPT and Codex from my subscription. As long as the inference is not done at a lost this analogy doesn't hold. They're not paying me to use it. They are generating output that is very valuable to me. Much more than my subscription cost. I've been able to help setup cross app automation for my partner's business, remodel my ho…

> They're not paying me to use it. Of course they are. > As long as the inference is not done at a loss. If making money on inference alone was possible, there would be a dozen different smaller providers who'd be taking the open weights models and offering that as service. But it seems that every provider is anchored at $20/month, so you can bet that none of them can go any lower.

They do make money on inference.

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

#525

Post by loloquwowndueo »

tomp wrote: post #499
danielodievich wrote: post #469

Last week there we had a customer request that landed in our support on a feature that I partially wrote and wrote a pile of public documentation on. Support engineer ran customer query through Claude (trained on our public and internal docs) and it very, very confidently made a bunch of stuff up in the response. It was quite plausible sounding and it would have been great if it worked that way, but it didn't. While…

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

As long as the explanation is going to come out being wrong, I’m sure you can whip something up in 0 seconds.

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

#526

Post by PantaloonFlames »

bgirard wrote: post #290
stack_framer wrote: post #247

Earlier quoted context omitted.

I can generate $20 billion in ARR this year too! I just need you to give me $100 billion and allow me to sell each of your dollars for 0.2 dollars.

It's a fun trope to repeat but that's not what OpenAI is doing. I get a ton of value from ChatGPT and Codex from my subscription. As long as the inference is not done at a lost this analogy doesn't hold. They're not paying me to use it. They are generating output that is very valuable to me. Much more than my subscription cost. I've been able to help setup cross app automation for my partner's business, remodel my ho…

You are mostly missing the point. You’re saying you get value out of what OpenAI is offering you. Thats not at issue here.

The question is, does OpenAI get value out of the exchange?

You touched on it ever so briefly: “as long as inference is not done at a loss”. That is it, isn’t it? Or more generally, As long as OpenAI is making money . But they are not.

There’s the rub.

It’s not only about whether you think giving them your money is a good exchange. It needs to be a good exchange for both sides, for the business to be viable.

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

#527

Post by CSMastermind »

CamperBob2 wrote: post #405
alecco wrote: post #317

Earlier quoted context omitted.

SemiAnalysis said it last week and AFAIK it wasn't denied. https://newsletter.semianalysis.com/p/tpuv7-google-takes-a-s...

That is.... actually a seriously meaty article from a blog I've never heard of. Thanks for the pointer.

Semianalysis is great, def recommend following

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

#528

Post by brazukadev »

tomp wrote: post #499
danielodievich wrote: post #469

Last week there we had a customer request that landed in our support on a feature that I partially wrote and wrote a pile of public documentation on. Support engineer ran customer query through Claude (trained on our public and internal docs) and it very, very confidently made a bunch of stuff up in the response. It was quite plausible sounding and it would have been great if it worked that way, but it didn't. While…

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

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

#529

Post by pcchristie »

throwacct wrote: post #80
OutOfHere wrote: post #79

Earlier quoted context omitted.

I strongly advise never using Google's Drive storage. They're known to scan all content, and to disable all access if even a single file is "problematic", often misclassified by a bot. If you do use the storage, do backup all your files, and be ready to lose access at any time, with no way to reach any intelligent human.

Which product do you recommend? OneDrive? Dropbox?

Filen is quite good, is E2E encrypted and currently offering (final round of) lifetime plans for Black Friday.

They are not super mature yet (though have been around for several years) so the product still has some improvements to be made, but I like it.

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

#530

Post by embedding-shape »

rahimnathwani wrote: post #509
alecco wrote: post #317

Earlier quoted context omitted.

SemiAnalysis said it last week and AFAIK it wasn't denied. https://newsletter.semianalysis.com/p/tpuv7-google-takes-a-s...

Dylan Patel joined Dwarkesh recently to interview Satya Nadella: https://www.dwarkesh.com/p/satya-nadella-2

And this is relevant how? That interview is 1.5 hours, not something you just casually drop a link to and say "here, listen to this to even understand what point I was trying to make"
Post reply on HN