In recent comments, Demis Hassabis, the CEO of Google DeepMind, acknowledged that DeepSeek’s AI model is likely the finest output from China in the field of artificial intelligence.
However, he noted that the company has not introduced any groundbreaking scientific advancements.
DeepSeek’s AI Model: A Game-Changer or Just Hype?
DeepSeek, a Chinese AI company, recently released a research paper that made waves in global markets. The company claimed its AI model was trained at a fraction of the cost compared to leading AI companies, using less-advanced Nvidia chips. This announcement ignited a sharp sell-off in stock markets and sparked a broader conversation about the high expenses associated with AI infrastructure in major tech firms.
Hassabis, speaking at a Google-hosted event in Paris ahead of the AI Action Summit, described DeepSeek’s model as an “impressive piece of work” and one of the best to emerge from China. He commended the company for its “extremely good engineering” and acknowledged the geopolitical significance of the model’s release.
However, Hassabis was quick to point out that the model, despite the fanfare, does not represent any new scientific breakthrough. “Despite the hype, there’s no actual new scientific advance… it’s using known techniques in AI,” he stated, emphasizing that much of the excitement surrounding DeepSeek’s model had been “exaggerated a little bit.”
Gemini 2.0: Google’s Answer to DeepSeek
While DeepSeek’s model may have garnered attention, Hassabis highlighted that Google’s Gemini 2.0 Flash models, which were released globally this week, are more efficient than DeepSeek’s offering.
Experts have also raised doubts about DeepSeek’s claims regarding the low cost of its models and the chips it uses, suggesting that the development costs may actually be higher than the company suggests.
Artificial General Intelligence: Just a Few Years Away?
The arrival of Artificial General Intelligence (AGI)—AI that surpasses human intelligence—has long been a topic of debate within the tech world. Hassabis weighed in on this ongoing conversation, stating that the AI industry is on the verge of creating AGI, which he defines as a system capable of exhibiting all cognitive capabilities that humans possess.
“We’re close now, maybe only about five years away from a system like that, which would be pretty extraordinary,” Hassabis remarked. He also pointed out the societal implications of AGI and the need for global readiness to harness its benefits while mitigating the associated risks.
Hassabis’ comments align with those made by OpenAI CEO Sam Altman, who earlier this year expressed confidence in the industry’s ability to build AGI as it has traditionally been understood. However, many experts, including Max Tegmark and Yoshua Bengio, have voiced concerns about the risks of AGI, particularly regarding the potential loss of human control over such systems.
Key Points at a Glance:
| Topic | Key Details |
|---|---|
| DeepSeek AI Model | Praised for engineering but lacks new scientific breakthroughs. |
| Costs & Claims | Experts question the affordability and costs associated with DeepSeek’s AI development. |
| Gemini 2.0 Models | Google’s new AI models are more efficient than DeepSeek’s according to Demis Hassabis. |
| AGI Timeline | Hassabis suggests AGI could be just 5 years away, emphasizing societal readiness and risk management. |
| Industry Concerns | Risks of AGI losing human control highlighted by experts. |
While DeepSeek’s AI model has generated significant attention and praise, particularly for its engineering, it has not revolutionized the field in terms of scientific breakthroughs.
As the race towards Artificial General Intelligence continues, industry leaders like Demis Hassabis are optimistic about the future but remain cautious about the potential risks involved.
With AGI possibly just five years away, the focus will need to shift toward societal readiness and responsible management of these powerful systems.
FAQs
What did Demis Hassabis say about DeepSeek’s AI model?
Hassabis praised DeepSeek’s AI as impressive but clarified that it doesn’t present any new scientific advancements. He also noted that the excitement around the model has been exaggerated.
What’s the difference between DeepSeek’s model and Google’s Gemini 2.0 models?
While DeepSeek’s model has drawn attention for its cost and efficiency, Hassabis stated that Google’s Gemini 2.0 Flash models are more efficient than DeepSeek’s offerings.
What is Artificial General Intelligence (AGI)?
AGI refers to AI systems that have the ability to perform any intellectual task that a human being can. This is considered a significant leap from current AI, which is specialized in narrow tasks.