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

881–890 of 1001 posts

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

#881

Post by jjallen »

blast wrote: post #374

Everyone is responding to the intellectual property issue, but isn't that the less interesting point? If Deepseek trained off OpenAI, then it wasn't trained from scratch for "pennies on the dollar" and isn't the Sputnik-like technical breakthrough that we've been hearing so much about. That's the news here. Or rather, the potential news, since we don't know if it's true yet.

That may be true. But an even more interesting point may be that you don’t have to train a huge model ever again? Or at least not to train a new slightly improved model because now we have open weights of an excellent large model and a way to train smaller ones.

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

#882

Post by jjallen »

tensor wrote: post #389
blast wrote: post #374

Everyone is responding to the intellectual property issue, but isn't that the less interesting point? If Deepseek trained off OpenAI, then it wasn't trained from scratch for "pennies on the dollar" and isn't the Sputnik-like technical breakthrough that we've been hearing so much about. That's the news here. Or rather, the potential news, since we don't know if it's true yet.

That's not correct. First of all, training off of data generated by another AI is generally a bad idea because you'll end up with a strictly less accurate model (usually). But secondly, and more to your point, even if you were to use training data from another model, YOU STILL NEED TO DO ALL THE TRAINING. Using data from another model won't save you any training time.

The DS R1 Model is slightly better though. So how does your statement square with that?

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

#886

Post by tw1984 »

emsign wrote: post #851
bilekas wrote: post #95

> “It’s also extremely hard to rally a big talented research team to charge a new hill in the fog together,” he added. “This is the key to driving progress forward.” Well I think DeepSeek releasing it open source and on an MIT license will rally the big talent. The open sourcing of a new technology has always driven progress in the past. The last paragraph too is where OpenAi seems to be focusing their efforts.. > we…

500 billion for a few US companies yet the Chinese will probably still be better for way less money. This might turn out to be a historical mistake of the new administration.

the biggest mistake was made 20 years by allowing China to join the WTO.

everything is already too late.

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

#887

Post by tw1984 »

cscurmudgeon wrote: post #857
bilekas wrote: post #95

> “It’s also extremely hard to rally a big talented research team to charge a new hill in the fog together,” he added. “This is the key to driving progress forward.” Well I think DeepSeek releasing it open source and on an MIT license will rally the big talent. The open sourcing of a new technology has always driven progress in the past. The last paragraph too is where OpenAi seems to be focusing their efforts.. > we…

The US doesn't need to ban DeepSeek from US The US should only ban DeepSeek (and other Chinese companies) from accessing US frontier models.

> The US should only ban DeepSeek (and other Chinese companies) from accessing US frontier models.

The US should only ban DeepSeek (and other Chinese companies) from accessing US frontier models designed and trained by Chinese Americans.

fixed for you.

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

#888

Post by Logiar »

razster wrote: post #772
buyucu wrote: post #454

Earlier quoted context omitted.

I'm willing to bet ''ban DeepSeek'' voices will start soon. Why compete, when you can just ban?

The fact it is out and improving day by day. Unsloth.ai is on a roll with their advancements. If DeepSeek is banned hundreds more will popup and change the data ever so slightly to skirt the ban. Pandora's box exploded on this one.

I'd imagine a ban would be on their service, not the model itself.

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

#889

Post by pknerd »

The reason OpenAI is whining:

> OpenAI’s o1 costs $60 per million output tokens; DeepSeek R1 costs $2.19. This nearly 30x difference brought the trend of falling prices to the attention of many people.

From Andrew Ng's recent DeeplearningAI newsletter

Post reply on HN