Nevertheless Rattling Markets and Silicon Valley, Google (googl-0.41%) Artificial Intelligence Heads are not worried about Chinese AI startup, DeepSeek.
Demis Hassabis, CEO of Google Deepmind, told employees of the Tech giant that when they looked at the “deepseek’s AI model, they found that the AI startup’s claims are “exaggerated.”CMCSA-0.11%)) It has been reportedQuotes audio from Wednesday’s All Hand Meeting.
Hassavis was reportedly asked “what lessons and implications” that could be learned from Deepseek’s success in an AI summary of employee questions. He reported to employees that the low-cost, competitive AI model of Deepseek’s training is probably “a small portion” of what it spent building AI systems, and AI startups say He said he likely used more hardware than he said. He also reportedly told employees that Deepseek probably relies on sophisticated models from AI companies in the West.
“In fact, we have a more efficient and performant model than DeepSeek,” Hassabis reportedly told employees. “So we’re very calm and confident in our strategy and we have all the elements that will help us maintain our leadership this year.”
Neither Google nor DeepMind responded to requests for comment immediately.
Earlier this week, Hassavis said that the AI model for the Hangzhou-based startup was “probably the best job” in China, and “impressive” during a Google event at the AI Action Summit held in Paris on CNBC. He said that. It has been reported. Hassabis said Deepseek has demonstrated “very good engineering” and that its AI model has a deep geopolitical implication.
However, he also said Deepseek has not shown “real new scientific advances” and is “using known techniques” in the AI industry, according to CNBC.
Last month, Deepseek released it result The latest open source inference model, Deepseek-R1, has been relatively implemented on Openai inference models O1-Mini and O1 in several industry benchmarks. In December, the startup launched the DeepSeek-V3 model. Train and develop nvidia’s (NVDA+2.72%) H800 chip – A version that reduces the capabilities of Nvidia’s H100 chips used by US companies.
Deepseek’s cheap competitive model raises doubts Big Tech’s big spending How effective is AI infrastructure and US chip export control?
