DeepMind CEO Calls DeepSeek’s Success “Overstated”

DeepSeek’s budget-friendly AI model is stirring up a fair share of anxiety among tech giants, particularly Google. At a recent meeting in Paris, Demis Hassabis, the CEO of DeepMind, reassured his team that they needn’t be alarmed by the rise of this Chinese AI startup. He defended Google’s edge in the industry, as reported by CNBC.

During the meeting, Alphabet’s CEO Sundar Pichai addressed an employee’s question about what insights Google could garner from DeepSeek’s rapid ascent. To give some perspective, DeepSeek unveiled an open-source model powered by their R1 V3, which has sent NVIDIA’s market value plummeting by a staggering $600 billion in just one day. Research papers suggest this model outperforms OpenAI’s proprietary o1 reasoning model in areas like science, math, and coding. However, the real game-changer seems to be that DeepSeek achieved these feats with significantly lower development costs compared to traditional proprietary AI models.

Despite the buzz around DeepSeek, Hassabis quickly downplayed their success, labeling it as somewhat “exaggerated.” He suggested that the cost of training mentioned by DeepSeek likely represents just a small slice of the actual expenses incurred, noting that the startup may have relied on additional hardware to achieve their results.

Adding to the intrigue, a recent report disclosed that DeepSeek poured $1.6 billion into acquiring 50,000 NVIDIA GPUs for their AI model’s development.

Hassabis confidently stated, “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.”

He subtly hinted that although Google stays a step ahead of DeepSeek, the Chinese contender deserves recognition as a formidable team from the region. However, Hassabis expressed concerns over security and geopolitical issues that could potentially hinder DeepSeek’s growth.

Curious about whether Google might have missed its chance with AI? Let’s delve into it.

AI development is advancing at such a pace that it’s tricky to predict who will lead the charge. As AI safety researcher Roman Yampolskiy mused, “If you have enough money to buy enough compute, you could build AGI today.”

While Yampolskiy’s theory sparks debate, Microsoft CEO Satya Nadella seems skeptical about Google’s chances of leading in AI. Last year, he noted that despite having the potential and resources, Google failed to secure the top spot in AI innovation. Nadella commented, “Google’s a very competent company, with talent and compute, integrated vertically in data to silicon, products, and distribution.”

This prompted Sundar Pichai to fire back at Microsoft, implying, “I would love to do a side-by-side comparison of Microsoft’s own models and our models any day, any time. They’re using someone else’s models.”

Pichai’s remark pointed to Microsoft’s partnership with OpenAI, a collaboration that combines Microsoft’s access to advanced AI models with OpenAI’s computing power and financial support.

Acknowledging DeepSeek’s influence, Nadella lauded their AI as “super impressive” and emphasized, “We should take the developments out of China very, very seriously.”