Google DeepMind CEO Confident in AI Strategy Amid DeepSeek's Rise

Google DeepMind CEO Demis Hassabis 'Calm and Confident' About AI Strategy | PYMNTS.com

Asked during an all-hands meeting about DeepSeek, which claimed last month that it built an artificial intelligence model at a fraction of the cost of other models, Hassabis said Google has what it takes to stay ahead, CNBC reported Friday (Feb. 14). Hassabis said some of DeepSeek's claims are "exaggerated," adding that the company probably reported only a fraction of the total cost of developing its systems, used more hardware than it said and relied on Western AI models, according to the report. At the same time, Hassabis said DeepSeek's accomplishments are impressive and something to be taken seriously, per the report. Hassabis said of Google: "We actually have more efficient, more performant models than DeepSeek. So, we're very calm and confident in our strategy, and we have all the ingredients to maintain our leadership into this year."

Google AI Chief tells employees company has 'all the ingredients' to hold AI lead over China's DeepSeek

Google DeepMind co-founder and Chief Executive Officer Demis Hassabis gives a conference during the Mobile World Congress (MWC), the telecom industry's biggest annual gathering, in Barcelona on February 26, 2024. Google's AI chief told employees that he's not worried about China's DeepSeek and said the search giant has superior artificial intelligence technology, according to audio of an all-hands meeting in Paris on Wednesday. At the meeting, Alphabet CEO Sundar Pichai read aloud a question about DeepSeek, the Chinese start-up lab that roiled U.S. markets recently, when its app shot to the top of the Apple's App Store, supplanting ChatGPT. DeepSeek released a research paper last month claiming its AI model was trained at a fraction of the cost of other leading models. The question, which was an AI summary of submissions from employees, asked "what lessons and implications" Google can glean from DeepSeek's success as the company trains future models. Google DeepMind CEO Demis Hassabis was called on to provide the answer. "When you look into the details," Hassabis said, some of DeepSeek's claims are "exaggerated." Hassabis added that DeepSeek's reported cost of its AI training was likely "only a tiny fraction" of the total cost of developing its systems. He said DeepSeek probably used a lot more hardware than it let on, and relied on western AI models. "We actually have more efficient, more performant models than DeepSeek," Hassabis said. "So we're very calm and confident in our strategy and we have all the ingredients to maintain our leadership into this year." But he admitted that DeepSeek's accomplishments are impressive. "It's definitely also the best team I think I've seen come out of China so something to be taken seriously," Hassabis said, noting that there are "security" and "geopolitical" implications. Several U.S. agencies have barred staffers from using DeepSeek, citing security concerns. Google declined to comment. DeepSeek didn't respond to a request for comment.

Google AI chief tells employees company has 'all the ingredients' to hold AI lead over China's DeepSeek

Google DeepMind Founder and CEO, Demis Hassabis, in 2024.Dan Kitwood / Getty Images Google's AI chief told employees that he's not worried about China's DeepSeek and said the search giant has superior artificial intelligence technology, according to audio of an all-hands meeting in Paris on Wednesday. At the meeting, Alphabet CEO Sundar Pichai read aloud a question about DeepSeek, the Chinese start-up lab that roiled U.S. markets recently, when its app shot to the top of the Apple's App Store, supplanting ChatGPT. DeepSeek released a research paper last month claiming its AI model was trained at a fraction of the cost of other leading models. The question, which was an AI summary of submissions from employees, asked "what lessons and implications" Google can glean from DeepSeek's success as the company trains future models. Google DeepMind CEO Demis Hassabis was called on to provide the answer. "When you look into the details," Hassabis said, some of DeepSeek's claims are "exaggerated." Hassabis added that DeepSeek's reported cost of its AI training was likely "only a tiny fraction" of the total cost of developing its systems. He said DeepSeek probably used a lot more hardware than it let on, and relied on western AI models. "We actually have more efficient, more performant models than DeepSeek," Hassabis said. "So we're very calm and confident in our strategy and we have all the ingredients to maintain our leadership into this year." But he admitted that DeepSeek's accomplishments are impressive. "It's definitely also the best team I think I've seen come out of China so something to be taken seriously," Hassabis said, noting that there are "security" and "geopolitical" implications. Several U.S. agencies have barred staffers from using DeepSeek, citing security concerns. Google declined to comment. DeepSeek didn't respond to a request for comment. Google executives also received a number of employee questions about the company's recent decision to change its "AI Principles" to no longer include a pledge against using AI for weapons or surveillance. Pichai read aloud an AI-summarized version of the questions, ending with "Why did we remove this section?" Pichai directed the question to Kent Walker, Google's president of global affairs, who said he had worked with Hassabis, James Manyika, a senior vice president a the company, and others on an effort that "shifted our approach," starting last year. Google established its AI principles in 2018 after declining to renew a government contract called Project Maven, which helped to analyze and interpret drone videos using AI. "Some of the strict prohibitions that were in v1 of the AI principles don't jibe well with the more nuanced conversations that we're having now," Walker said, referring to the rules from 2018. Walker said "an awful lot has changed in those seven years," and that the technology has advanced to the point where "it's used in lots of very nuanced scenarios."

Google's AI head wants employees to chill out over DeepSeek

Elon Musk denies knowledge of a $400 million 'Armored' Tesla deal with the Trump Administration Google DeepMind chief executive Demis Hassabis told the tech giant's employees that when looking "into the details" of DeepSeek's AI models, the AI startup's claims are "exaggerated," CNBC (CMCSA+0.31%) reported, citing audio of an all-hands meeting on Wednesday. Hassabis was reportedly asked "what lessons and implications" the company could learn from DeepSeek's success in an AI summary of employees' questions. He told employees that DeepSeek's reported low cost of training competitive AI models is possibly "only a tiny fraction" of what it spent to build its AI systems, and that the AI startup likely used more hardware than it said it did. He also reportedly told employees that DeepSeek probably depended on advanced models from AI companies in the west. "We actually have more efficient, more performant models than DeepSeek," Hassabis reportedly told employees. "So we're very calm and confident in our strategy and we have all the ingredients to maintain our leadership into this year." Neither Google nor DeepMind immediately responded to a request for comment. Earlier this week, Hassabis said the Hangzhou-based startup's AI model "is probably the best work" from China, and is "impressive," during a Google event at the AI Action Summit in Paris, CNBC reported. Hassabis said DeepSeek has demonstrated "extremely good engineering," and that its AI models have deeper geopolitical implications. However, he also said DeepSeek doesn't show "actual new scientific advance" and is "using known techniques" in the AI industry, according to CNBC. Last month, DeepSeek released results for its latest open-source reasoning models, DeepSeek-R1, which performed comparably to OpenAI's reasoning models, o1-mini and o1, on several industry benchmarks. In December, the startup launched its DeepSeek-V3 models which it said cost just $5.6 million to train and develop on Nvidia's (NVDA+1.30%) H800 chips -- the reduced-capability version of Nvidia's H100 chips used by U.S. firms. DeepSeek's cheaper-yet-competitive models have raised questions over Big Tech's big spending on AI infrastructure, as well as how effective U.S. chip export controls are.

Google's CEO Praised AI Rival DeepSeek This Week for Its 'Very Good Work.' Here's Why.

Google CEO Sundar Pichai praised Chinese AI startup DeepSeek on Wednesday, which created a competitive AI model for under $6 million compared to the industry standard of $100 million. "I think the DeepSeek team has done very, very good work," Pichai said at the World Governments Summit in Dubai. Pichai said Google learned some lessons from DeepSeek's rise to the top of the Apple App Store and Google Play charts last month, including showing the world "how global" AI competition really is. "I think the DeepSeek innovation reinforces that point," he said. Related: 'Pride of His Hometown': Who Is DeepSeek Founder Liang Wenfeng? Pichai also praised DeepSeek's AI for being highly efficient and releasing accurate results while using minimal computing power. According to DeepSeek's technical report, the startup only used 2,000 AI chips to train its AI model, compared to Meta, which used more than 16,000 chips for its latest AI. More chips require more money. Pierre Ferragu, a New Street Research analyst, told the New York Times last month that Google spent $2 billion to $3 billion last year creating a million custom AI chips, while Amazon spent $200 million building 100,000 AI chips. DeepSeek's AI is also open-source, meaning that developers around the world can freely access its work and build more advanced AI on top of it. Related: DeepSeek AI Cost Less Than $6 Million to Develop. Here's Why Meta and Microsoft Are Justifying Spending Billions. "I think what caught people's attention with DeepSeek was that you could have an efficient model and open-source and something everyone can immediately access, and I think that creates a lot of excitement," Pichai said. He stated that Google was focused on efficiency, too. Google released its Gemini 2.0 Flash-Lite AI model last week, calling it the company's "most cost-efficient model yet" without going into specifics about how much it cost to make.

Google's CEO Pichai says China's DeepSeek has done very good work

DUBAI, Feb 12 (Reuters) - Alphabet (GOOGL.O), opens new tab CEO Sundar Pichai said Chinese startup DeepSeek has done very good work that showed how global artificial intelligence is. Pichai expected Alphabet to be a leading player in the AI space but said others would be there in that space too, he said at the World Governments Summit held in Dubai. Reporting by Yoused Saba; Writing by Nayera Abdallah; Editing by Alison Williams Our Standards: The Thomson Reuters Trust Principles., opens new tab Suggested Topics:Artificial Intelligence