Google Deepmind CEO claims Deepseek hype is exaggerated
- Demis Hassabis praised Deepseek's AI model as a significant achievement from China during a recent event.
- Despite the praise, he stated that the model did not introduce any new scientific advances and used established AI techniques.
- Hassabis concluded by emphasizing that the AI industry is potentially five years away from achieving artificial general intelligence.
On February 9, 2025, in Paris, during a Google-hosted event, Demis Hassabis, the CEO of Google DeepMind, acknowledged that Deepseek's AI model was likely the finest work from China. This statement came in the wake of Deepseek's recent research publication, which claimed an impressive achievement in training their AI model at a significantly lower cost and on less advanced technology compared to leading AI firms. The announcement caused substantial reactions in global financial markets, prompting discussions about whether major tech companies are overspending on AI infrastructures. However, Hassabis emphasized that while Deepseek's work was commendable, it did not represent any groundbreaking scientific innovations, reiterating that the model utilized established AI techniques, indicating a lack of true advancement in technology despite the surrounding hype. He also pointed out that Google’s newly released Gemini 2.0 Flash models surpass Deepseek in efficiency, further questioning the authenticity of Deepseek's claims regarding their model's cost-effectiveness and technological sophistication. The AI community remains cautiously optimistic about the future, with debates surrounding the potential arrival of artificial general intelligence (AGI). Hassabis speculated that we might be about five years away from creating a system with human-like cognitive abilities, urging society to prepare for the implications of AGI while also addressing associated risks. These remarks coincide with perspectives shared by other industry leaders, who also believe AGI is drawing closer to reality but caution that significant risks must be mitigated to prevent loss of control over AI systems, a sentiment echoed recently by leading AI experts. This ongoing discourse highlights the complexities in the trajectory toward realizing AGI alongside the technological developments by companies like Deepseek and Google.