Although Google’s Hassabis acknowledged Deepseek’s performance, he explained that the reported cost-effectiveness figures are “exaggerated and a bit misleading.”
read more
Demis Hassabis, CEO of Google Deepmind, has questioned the recent claims made by Chinese AI startup Deepseek regarding the cost-effectiveness of its models. Speaking at the Artificial Intelligence Action Summit in Paris, Hassavis acknowledged Deepshek’s achievements, but explained that the reported cost-effectiveness figures are “exaggerated and a bit misleading.”
Deepseek recently attracted global attention with its R1 model. This claims it can compete with Openai’s GPT-4-level O1 model despite being trained at just $5.6 million, at just $5.6 million, which Openai reportedly spent on GPT-4. It’s there. However, Hassavis argues that $5.6 million is likely to reflect only the final training run, omitting substantial costs associated with data collection, infrastructure and developments that include multiple training. It’s there.
Hassavis also suggested that Deepseek may have relied heavily on knowledge gained from existing Western AI models and relied on its own knowledge. Openai has previously raised similar concerns, claiming that companies from China often distill or replicate the work of major US AI companies. Openai said after Deepseek’s launch Bloomberg Companies based in China and other countries frequently attempt to extract and refine US AI models for their own use.
Google doesn’t consider Deepseek as a game changer
Despite the topics surrounding Deepseek, Hassabis downplayed the concept of representing a major leap in AI efficiency. He argued that Google’s Gemini model would improve performance cost-effectiveness, even if it wasn’t promoted with the same level of marketing fanfare. “It’s impressive, but it’s not a new outlier in the efficiency curve,” Hassavis said.
Deepseek’s claim has sparked a lively debate across the technology world, with experts splitting on whether startups have truly redefine the costs of AI development or have found a clever way to bring it to the market itself .
Practical testing: Innovation and long-term impact
Deepseek’s reported efficiency figures are impressive on paper, but industry leaders like Hassabis have found that true success in AI is not a bold claim, but to long-term innovation and reliability It suggests that. As competition for stronger and more cost-effective AI models heats up, both established players like Google and emerging startups like DeepSeek will face a growing scrutiny about performance and transparency.
Ultimately, the ongoing competition for AI domination may depend on companies being able to deliver consistent, groundbreaking technology without cutting corners.
