DeepMind CEO Calls DeepSeek’s Success “Overstated”

The buzz around DeepSeek’s new budget-friendly AI has certainly put some big players in the industry on alert, not least of which is Google. At a recent company-wide meeting in Paris, DeepMind’s CEO, Demis Hassabis, assured employees not to be overly concerned about the rapid rise of this Chinese AI upstart. According to a report by CNBC, Hassabis emphasized Google’s strong position in the AI arena, downplaying the threat posed by DeepSeek.

During this meeting, Alphabet’s CEO, Sundar Pichai, fielded a curious question from an employee about what lessons Google could glean from DeepSeek’s seemingly “overnight success.” For those unfamiliar, DeepSeek’s R1 V3-powered open-source model has shaken up the market drastically, with NVIDIA seeing a massive $600 billion dip in their market value in just a day. Allegedly, this model outperforms OpenAI’s advanced o1 reasoning model across diverse domains like science, math, and coding—and does all this at a significantly lower cost than conventional proprietary models.

Yet, Hassabis was quick to call DeepSeek’s reported success into question, suggesting it might be exaggerated. After examining the details, he stated the cost efficiency claimed by DeepSeek could be “only a tiny fraction” of the actual expenses involved. He pointed to their heavy reliance on hardware as a sign that costs are perhaps higher than reported. A separate insight reveals that DeepSeek invested $1.6 billion and acquired 50,000 NVIDIA GPUs to develop its AI.

Hassabis confidently asserted that Google’s AI models are both more efficient and perform better than what DeepSeek offers. “We’re very calm and confident in our strategy,” he said, pointing out that Google possesses all the necessary components to hold its leadership position in the industry.

Even with these reassurances, the executive acknowledged the standout talent at DeepSeek, referring to it as “the best team” to emerge from China, though he noted they face significant security and geopolitical challenges which could impede their growth potential.

The intrigue around who will command the future of generative AI continues to deepen. AI safety researcher Roman Yampolskiy weighs in, suggesting that the resourceful can indeed achieve advanced AI today. Meanwhile, Microsoft’s CEO, Satya Nadella, has expressed skepticism that Google is poised to dominate this space, despite their considerable capabilities. Nadella once pointed out that Google possesses all the essential resources—from data and computing to distribution—necessary for success but hasn’t capitalized fully.

In a rather candid retort, Pichai made a bold comparison to Microsoft’s side, insinuating that they are leveraging third-party models via their partnership with OpenAI. This partnership involves Microsoft providing financial and computational support while benefiting from OpenAI’s cutting-edge technologies.

As the AI race intensifies, all eyes are on these industry titans to see who will come out on top. Microsoft’s CEO has highlighted DeepSeek’s AI as exceptionally impressive, urging the tech world to pay close attention to advancements emerging from China.