DeepSeek may have used Google's Gemini to train its latest model

Re:DeepSeek may have used Google's Gemini to train its lates... by philipwhiuk on Wednesday June 04, 2025 @08:38AM (#44180099)

> Distillation isn’t an uncommon practice, but OpenAI’s terms of service prohibit customers from using the company’s model outputs to build competing AI.

I have the absolute tiniest of violins for this given OpenAI's behaviour vs everyone else's terms of service.

[ Parent | on HN ]

Re:DeepSeek may have used Google's Gemini to train its lates... by sovietmudkipz on Wednesday June 04, 2025 @09:44AM (#44180679)

“Copyright must evolve into the 21century (…so that AI can legally steal everything produced by people”

And also “Don’t steal our AI!”

[ Parent | on HN ]

Re:DeepSeek may have used Google's Gemini to train its lates... by jsheard on Wednesday June 04, 2025 @10:44AM (#44181232)

The world is not prepared for the mental gymnastics that OpenAI/Google/etc will employ to defend their copyright if their big models ever get leaked.

[ Parent | on HN ]

Re:DeepSeek may have used Google's Gemini to train its lates... by bitpush on Wednesday June 04, 2025 @11:40AM (#44181959)

I see no evidence that Google is doing this. Any sources?

[ Parent | on HN ]

Re:DeepSeek may have used Google's Gemini to train its lates... by Zetaphor on Wednesday June 04, 2025 @10:17AM (#44180984)

I'm still unclear how they are able to claim this considering their raw thinking traces were never exposed to the end user, only summaries.

[ Parent | on HN ]

Re:DeepSeek may have used Google's Gemini to train its lates... by dev_hugepages on Thursday June 05, 2025 @10:00AM (#44191796)

The raw thinking traces were shown previously in the API.

[ Parent | on HN ]