OpenAI, the owner of ChatGpt, has accused its Chinese rival, DeepSeek, of using its learning models to develop its own artificial intelligence. Bloomberg reports that Microsoft is investigating whether OpenAI’s data was used without authorization.
David Sacks, the White House AI czar, has raised the possibility of possible intellectual property theft by DeepSeek from OpenAI.
Agenzia ANSAGarante chiede informazioni a DeepSeek, dati a rischio – Ultima ora – Ansa.it
Il Garante per la protezione dei dati personali ha inviato una richiesta di informazioni a Hangzhou DeepSeek Artificial Intelligence e a Beijing DeepSeek Artificial Intelligence, le società che forniscono il servizio di chatbot DeepSeek, sia su piattaforma… (ANSA)
“There is a technique in artificial intelligence called distillation… when a model learns from another model, it sort of sucks knowledge from the main model,” Sacks told Fox News. “There is substantial evidence that what DeepSeek did here was distill knowledge from OpenAI models, and I don’t think OpenAI is very happy about it,” Sacks added, although he did not provide evidence.
White House spokesperson Karoline Leavitt announced that U.S. officials are assessing the national security implications of DeepSeek. “I spoke with (the National Security Council, ed.) this morning, they are examining what the implications for national security could be,” said Leavitt.
According to CNBC, the U.S. Navy has sent an email to its personnel instructing them not to use the DeepSeek app due to “potential security and ethical issues.”
In China, hashtags praising DeepSeek’s supremacy are proliferating
All across China, citizens are praising the success of the tech startup DeepSeek and its founder, after the company’s new AI model disrupted Silicon Valley and Wall Street.
Agenzia ANSAOpenAI lancia ChatGPT Gov per il governo americano – Nord America – Ansa.it
OpenAI lancia ChatGPT Gov, costruita in modo specifico per l’uso da parte del governo americano. La nuova piattaforma consente alle agenzie federali di inserire “informazioni non pubbliche” nei modelli OpenAI lavorando in sicurezza. (ANSA)
“DeepSeek overturns U.S. actions overnight,” declared a trending hashtag with tens of millions of views on the Chinese social platform Weibo. “DeepSeek is driving Meta crazy,” claims another, referring to the U.S. tech giant that has heavily invested in developing its own AI models.
At the beginning of this week, more than a dozen hashtags related to cutting-edge technology were trending on Weibo. DeepSeek’s founder, Liang Wenfeng, has been hailed as a technology visionary who could help China usher in a culture of innovation to rival that of Silicon Valley. Wenfeng is known for hiring only national talents and keeping his AI models open-source, allowing other companies or users to test and develop the model.
‘Huawei chips also used for DeepSeek’
It’s been revealed that Huawei chips, not just Nvidia’s, have been used for DeepSeek’s artificial intelligence. An informant told the specialized site Gsm Arena that behind the work done by the Chinese AI, there is a chip known as Ascend 910C from Huawei. This chip is dedicated to the model’s inference activity, generating responses to users’ questions.
Training, on the other hand, as per official information, takes place on an Nvidia H800 processing unit, a limited version for China of the more powerful H100. In both cases, it is not the latest generation hardware, but it allows DeepSeek to achieve the performance of ChatGpt and other more renowned models that rely on more powerful hardware.
If confirmed, the news of Huawei’s involvement would be a significant proof of the Shenzhen-based group’s commitment to delve even deeper into the field of artificial intelligence. During a summit last April, Huawei outlined its long-term AI strategy, which also includes developing a new intelligent assistant for mobile devices.
According to Gsm Arena, Huawei is also close to launching the 920C chips aimed at competing with Blackwell B200, Nvidia’s main hardware for AI operations.
Reproduction reserved © Copyright ANSA
