News by Xiaomi Miui Hellas
Home » All the news » News » Technology » OpenAi: Accuses DeepSeek of copyright infringement through use of ChatGPT
Technology

OpenAi: Accuses DeepSeek of copyright infringement through use of ChatGPT

OpenAI-logo

Η OpenAI accuses Chinese artificial intelligence start-up DeepSeek for copyright infringement, claiming that it used its own models to train its AI chatbot


Since the Chinese artificial intelligence (AI) start-up DeepSeek shook her Silicon Valley and Wall Street with its cost-effective models, the company has been accused of data theft through a practice that is common across the industry.

Η OpenAI stated that he has evidence that the DeepSeek used the method of "distillation" of the GPT models OpenAI, to train the open source V3 and R1 models in DeepSeek, at a fraction of the cost that Western tech giants spend on their own models, as reported by Financial Times on Wednesday.

Although it is distillation It is common practice in the industry to OpenAI argues that its use to create a competitive model violates its terms of use.

Distillation is a means of training smaller models to mimic the behavior of larger, more sophisticated models. This practice is common among many companies that want to reduce the size of their models while still offering similar performance to larger models to their users.

All of the above, combined with the fact that the training of all known AI models is often based on a lot of data of dubious origin, has led some experts to question OpenAI's sincerity in its accusations of intellectual property infringement by the DeepSeek.

It is noted that DeepSeek surprised the industry with its release model R1, which achieved performance comparable to leading American models despite its low development cost. Concerns about a reduced need to purchase expensive AI hardware contributed to the Nvidia stock drops 17%, although he partially recovered the next day.

As he said Lutz Finger, a senior lecturer at Cornell University who has worked on artificial intelligence at technology companies such as Google and LinkedIn,

With the same justification, we can say that the education of Chat GPT in its content Forbes or of New York Times It also violated OpenAi's terms of service. Knowledge is free and difficult to protect.

But apart from all of the above, things seem to be getting even more complicated when it comes to DeepSeek, as according to what the @Dorialexander, The DeepSeek R1 LLM may have been initially trained in Nvidia H100 hardware, which is one of the most powerful processors for training AI models.

However, the model execution is carried out using the Huawei Ascend 910C, because the Huawei's Ascend chip they are not capable of taking on the burden of training AI models.

The reason for using the Ascend 910C is that the processing power requirements for running the model are lower than for training. However, this type of chip is more efficient and more economical to meet these needs.


Mi TeamDo not forget to follow it Xiaomi-miui.gr on Google News to be informed immediately about all our new articles! You can also if you use RSS reader, add our page to your list by simply following this link >> https://xiaomi-miui.gr/feed/gn

 

Follow us on Telegram so you can be the first to hear about our news! (English version HERE)

Read also

Leave a comment

* By using this form you agree to the storage and distribution of your messages on our page.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Leave a Review

Xiaomi Miui Hellas
The official community of Xiaomi and MIUI in Greece.
Read also
You can now get Heroes Legend – Epic Fantasy for free from Google…