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

OpenAI says it has evidence DeepSeek used its model to train competitor

ft.com

Post reply on HN
Print view View original

651–660 of 1001 posts

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#651

Post by johndhi »

tempeler wrote: post #479
Imnimo wrote: post #320

I think there's two different things going on here: "DeepSeek trained on our outputs and that's not fair because those outputs are ours, and you shouldn't take other peoples' data!" This is obviously extremely silly, because that's exactly how OpenAI got all of its training data in the first place - by scraping other peoples' data off the internet. "DeepSeek trained on our outputs, and so their claims of replicating…

On another subject, if it belongs to OpenAI because it uses OpenAI, then doesn't that mean that everything produced using OpenAI belongs to OpenAI? Isn't that a reason not to use OpenAI? It's very similar to saying that you used Google and searched; now this product belongs to Google. They couldn't figure out how to respond; they went crazy.

to be clear, their terms of service are pretty clear that the USER owns the outputs.

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#652

Post by pertymcpert »

hmottestad wrote: post #641
Imnimo wrote: post #504

Earlier quoted context omitted.

I think it would cast doubt on the narrative "you could have trained o1 with much less compute, and r1 is proof of that", if it turned out that in order to train r1 in the first place, you had to have access to bunch of outputs from o1. In other words, you had to do the really expensive o1 training in the first place. (with the caveat that all we have right now are accusations that DeepSeek made use of OpenAI data -…

At the pace that DeepSeek is developing we should expect them to surpass OpenAI in not that long. The big question really is, are we doing it wrong, could we have created o1 for a fraction of the price. Will o4 cost less to train than o1 did? The second question is naturally. If we create a smarter LLM, can we use it to create another LLM that is even smarter? It would have been fantastic if DeepSeek could have come…

The whole discussion is about whether or not the second case of using o1 outputs to fine tune R1 is what allowed R1 to become so good. If that's the case then your assertion that DeepSeek will surpass OpenAI doesn't really make sense because they're dependent on a frontier model in order to match, not surpass.

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#654

Post by tasuki »

glenstein wrote: post #550
tasuki wrote: post #364

Earlier quoted context omitted.

I understand they just used the API to talk to the OpenAI models. That... seems pretty innocent? Probably they even paid for it? OpenAI is selling API access, someone decided to buy it. Good for OpenAI! I understand ToS violations can lead to a ban. OpenAI is free to ban DeepSeek from using their APIs.

Sure, but I'm not interested in innocence. They can be as innocent or guilty as they want. But it means they didn't, via engineering wherewithal, reproduce the OpenAI capabilities from scratch. And originally that was supposed to be one of the stunning and impressive (if true) implications of the whole Deepseek news cycle.

Nothing is ever done "from scratch". To create a sandwich, you first have to create the universe.

Yes, there is the question how much ChatGPT data DeepSeek has ingested. Certainly not zero! But if DeepSeek has achieved iterative self-improvement, that'd be huge too!

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#655

Post by roblabla »

sangnoir wrote: post #606
worik wrote: post #431

[flagged]

> Banned in the USA. Only. The US government has the wherewithal to drag Europe along with it, like they did with Huawei's 5G equipment.

So far, a general tiktok ban (as opposed to a tiktok ban on things like government phones) has only been in effect in the USA. I highly doubt Europe would play ball at any attempt at banning imports of DeepSeek.

Besides, it's kinda too late for this. The model is freely accessible, so any attempt at banning it would be _completely_ moot. If DeepSeek keeps releasing their future models for free, I don't see how a ban could ever be effective at all. Worse case scenario, big tech can't use those models... but then individuals (and startups willing to go fast and break laws) will be able to use them and instantly get a leg up on the competition.

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#656

Post by lenerdenator »

Sophira wrote: post #617
riantogo wrote: post #461

Earlier quoted context omitted.

Why would it cast any doubt? If you can use o1 output to build a better R1. Then use R1 output to build a better X1... then a better X2.. XN, that just shows a method to create better systems for a fraction of the cost from where we stand. If it was that obvious OpenAI should have themselves done. But the disruptors did it. It hindsight it might sound obvious, but that is true for all innovations. It is all good stuf…

Honestly, it's kind of silly that this technology is in the hands of companies whose only aim is to make money, IMO.

Well, originally, OpenAI wasn't supposed to be that kind of organization.

But if you leave someone in the tech industry of SV/SF long enough, they'll start to get high on their own supply and think they're entitled to insane amounts of value, so...

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#657

Post by beepbooptheory »

runako wrote: post #499
bangaladore wrote: post #435

Earlier quoted context omitted.

> So they'll go for getting DeepSeek banned like TikTok was now that a precedent has been set ? Can't really ban what can be downloaded for free and hosted by anyone. There are many providers hosting the ~700B parameter version that aren't CCP aligned.

I'm old enough to remember when the US government did something very similar. For years (decades?), we banned any implementation of public-key cryptography under the guise of the technology being akin to munitions. People made shirts with printouts of the code to RSA under the heading "this shirt is a munition." Apparently such shirts are still for sale, even though they are not classified as munitions anymore. [1] -…

I am not that old, but I did a deep dive on this in the past because it was just so extremely fascinating, especially reading the archives of Cypherpunk. There is a very solid, if rather bendy, line connecting all that to "crypto culture" today.

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#658

Post by rahimnathwani »

SubiculumCode wrote: post #628

If you have a set of weights A, can you derive another set of weights B that function (near) identically as A AND a) not appear to be the same weights as A when inspected superficially b) appear uncorrelated when inspecting the weight matrices?

Do you mean for a given model structure, can two sets of weights give substantially the same outputs?

Even if that were possible, it would be suspicious if you were to release an open model whose model architecture is identical to that of a closed one from a competitor.

If that is what happened, we'd know about it by now.

Re: OpenAI says it has evidence DeepSeek used its model to train competitor

#659

Post by hugoromano »

OpenAI initially scraped the web and later formed partnerships to train on licensed data. Now, they claim that DeepSeek was trained on their models. However, DeepSeek couldn't use these models for free and had to pay API fees to OpenAI. From a legal standpoint, this could be seen as a violation of the terms and conditions. While I may be mistaken, it's unclear how DeepSeek could have trained their models without compensating OpenAI. Basically, OpenAI is saying machines can't learn from their outputs as humans do.
Post reply on HN