DeepSeek, a Chinese start-up lab, recently made waves by catapulting its app to the top of Apple's App Store, overtaking ChatGPT. However, this success has not gone unnoticed or unchallenged. Multiple U.S. agencies have barred their employees from using DeepSeek, citing concerns over security. Despite the hype surrounding the start-up, Demis Hassabis, Google's head of artificial intelligence, remains unperturbed about DeepSeek's rapid ascent and its implications for Google's AI dominance.
In a January research paper, DeepSeek claimed that its AI model was trained at a fraction of the cost of other leading models. However, Hassabis questions the validity of these claims, stating that the reported cost is likely "only a tiny fraction" of what it truly takes to develop such sophisticated systems. He believes that DeepSeek's assertions are "exaggerated" and suggests that the start-up probably relied heavily on Western AI models and used more hardware than disclosed.
Hassabis expressed confidence in Google's superior AI technology, asserting that "We actually have more efficient, more performant models than DeepSeek." He acknowledged the talent behind DeepSeek, describing it as "the best team I think I've seen come out of China," but emphasized that Google's technological advancements and strategic approach keep it ahead in the AI race.
Google's AI principles, established in 2018 after declining to renew a contentious government contract, have undergone modifications. The company eliminated a pledge against using AI for weapons or surveillance, leading to internal queries from employees about "what lessons and implications" such changes entail. Nevertheless, Hassabis reassures that Google remains steadfast in its strategy and possesses "all the ingredients" necessary to maintain its leadership position in AI development.