The emergence of DeepSeek's ultra-cost-effective AI continues to raise concern among major AI firms, including Google. Recently, Demis Hassabis, DeepMind CEO, urged employees not to worry about the Chinese AI startup, citing Google's superiority in the AI landscape (via CNBC).
Hassabis shared the sentiments during an all-hands meeting in Paris where Alphabet CEO Sundar Pichai read aloud a question from a company employee. The question heavily leaned on lessons and implications Google could learn from DeepSeek's seemingly "overnight success."
For context, DeepSeek launched its R1 V3-powered open-source model, prompting NVIDIA's market share to plummet by a whopping $600 billion in a single day. A research paper claims the model surpasses OpenAI's proprietary o1 reasoning model across a wide range of capabilities, including science, math, and coding. However, perhaps more concerning is that the Chinese AI startup reportedly unlocked these impressive feats at a fraction of the cost of developing proprietary AI models.
However, Google's DeepMind CEO Hassabis quickly watered down DeepSeek's success as "exaggerated," especially after keenly examining the details. The executive claimed the model training cost reported by DeepSeek is probably "only a tiny fraction" of the total cost incurred. He further indicated that the Chinese startup used more hardware.
This news comes after a separate report suggested that DeepSeek spent $1.6 billion and bought 50,000 NVIDIA GPUs to facilitate the development of its AI model.
According to Google DeepMind CEO Demis Hassabis:
“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.”
The executive subtly indicated that Google holds a significant lead over DeepSeek in the AI landscape but admitted that it's "the best team" to come out of China, which should be taken seriously. However, he cited that DeepSeek is riddled with security with security and geopolitical concerns, which may stunt its progress, hindering the startup from realizing its full potential.
Read more: DeepSeek hit by large-scale cyberattack
Did Google miss its opportunity with AI?
As generative AI scales greater heights, it's increasingly difficult to tell who will win the race. Well, according to AI safety researcher Roman Yampolskiy:
"If you have enough money to buy enough compute, you could build AGI today."
While Yampolskiy's sentiments pose an interesting theory, Microsoft CEO Satya Nadella doesn't think Google will be in the running to become the leader in AI. Last year, the executive indicated that Google had all the potential and resources to become the default leader in AI but failed. According to the executive:
"Google's a very competent company and obviously they have both the talent and the compute. They're the vertically integrated player in this. They have everything from data to silicon to models to products and distribution."
This prompted Alphabet CEO Sundar Pichai to throw a lethal jab at Microsoft's AI efforts, indicating:
"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."
The CEO was referring to Microsoft and OpenAI's multi-billion partnership, which grants the Redmond giant access to next-gen AI models while the latter gains computing power and funding.
Microsoft's CEO touted DeepSeek's AI as "super impressive". "We should take the developments out of China very, very seriously," Nadella added.
