Google DeepMind CEO Demis Hassabis has acknowledged the DeepSeek AI model of China stating that it is an impressive achievement. However, he downplayed its significance in terms of technological advancement while speaking at a Google event in Paris and ahead of the AI Action Summit. He described DeepSeek as probably the best work he has seen coming out of China and recognized its strong engineering as well as geopolitical influence.

Google CEO Sundar Pichai commended the team of DeepSeek for their excellent work during the Q4 2024 earnings call of Alphabet. Hassabis emphasized that the new AI model is not based on any groundbreaking scientific developments but basically relies on existing AI techniques. He argued that the Gemini 2.0 Flash models of Google are more efficient and suggested that claims of innovation of the Chinese startup may not hold up under scrutiny.

One major debating point is the claim of DeepSeek that it trained its AI model at a fraction of the cost compared to other AI chat tools and using less advanced Nvidia chips. The announcement resulted with a stock sell-off as concerns were raised about excessive AI infrastructure spending.

Hassabis also expressed confidence that Artificial General Intelligence (AGI) could be just five years away. He defined it as a system that is equipped with cognitive capabilities similar to humans.

The conversation was mainly around cost, efficiency and real innovation as these are critical in the development of AI further. The AI model of DeepSeek model may be a notable achievement for China and the comments of Hassabis suggest that the true breakthroughs in AI are yet to come.