Mr Viral
Google AI chief also believes that DeepSeek has relied on some Western models to learn from
Feb 12, 20255:06 PM EST
0 comments
Google’s DeepMind boss, Demis Hassabis, indicated that China’s DeepSeek AI did not spend under US$6 million (roughly C$8.5 million) to develop the AI system and says the claim is “exaggerated and a little bit misleading.”
Hassabis, who runs Google’s AI unit, told Bloomberg that DeepSeek’s under $6 million claim is misleading and that it “seems to have only reported the cost of the final training round, which is a fraction of the total cost.”
Reportedly, DeepSeek indicates that it spent about US$5.6 million (roughly C$8 million) on computing costs when training its model to use older Nvidia chips. However, U.S. authorities are probing to see if DeepSeek has circumvented the chip ban by buying new chips through Singapore.
Further, OpenAI and Microsoft are also investigating whether a group tied to DeepSeek distills from OpenAI. Distillation uses a large language model to transfer knowledge to a smaller one, which Hassabis agrees with. DeepSeek seems “to have relied on some Western models to distill from,” he told Bloomberg, without providing more details.
Alphabet, on the other hand, has reportedly planned for US$75 billion (roughly C$107 billion) capital that will go towards its cloud-computing division to improve services like Google’s AI model Gemini, as well as Search and other products.
Source: BNN Bloomberg
MobileSyrup may earn a commission from purchases made via our links, which helps fund the journalism we provide free on our website. These links do not influence our editorial content. Support us here.
GIPHY App Key not set. Please check settings