## China’s AI Capabilities May Be Just Months Behind the West, Says Google DeepMind CEO
**Demis Hassabis, CEO of Google DeepMind, believes China’s artificial intelligence models are closer to Western capabilities than previously thought, potentially just “a matter of months” behind.** This assessment challenges prevailing views that have suggested a wider gap between China and the U.S. in the AI race.
Speaking on CNBC’s “The Tech Download” podcast, Hassabis, a pivotal figure behind Google’s Gemini AI assistant, indicated that Chinese AI models have made significant strides. “Maybe they’re only a matter of months behind at this point,” he stated, a notable shift from perceptions held even a year or two ago.
This evolution in China’s AI landscape is not without precedent. Roughly a year ago, the Chinese AI lab DeepSeek released a model that surprised the market with its strong performance, achieved using less advanced and more cost-effective chips compared to U.S. alternatives. While the initial shock has subsided with subsequent releases from DeepSeek and advancements from other Chinese tech giants like Alibaba and startups such as Moonshot AI and Zhipu, these entities have consistently introduced highly capable AI models.
However, Hassabis stressed that while China has demonstrated a remarkable ability to catch up, the crucial question remains whether its companies can achieve genuine breakthroughs and innovate beyond the current frontier of AI. He elaborated, “The question is, can they innovate something new beyond the frontier? So I think they’ve shown they can catch up … and be very close to the frontier … But can they actually innovate something new, like a new transformer … that gets beyond the frontier? I don’t think that’s been shown yet.” The transformer architecture, a groundbreaking development by Google researchers in 2017, remains the foundational technology for many of today’s leading large language models, including those powering OpenAI’s ChatGPT and Google’s Gemini.
This perspective echoes sentiments from other industry leaders. Earlier this year, Nvidia CEO Jensen Huang remarked that the U.S. is “not far ahead” in the AI race, highlighting the complex dynamics at play. Huang noted, “China is well ahead of us on energy. We are way ahead on chips. They’re right there on infrastructure. They’re right there on AI models.”
### Navigating Chip Restrictions and the Path to Innovation
Despite rapid progress, China’s technology sector faces significant hurdles, particularly concerning access to cutting-edge semiconductors. U.S. export controls restrict the sale of advanced Nvidia chips essential for training the most sophisticated AI models. While the U.S. government has approved the sale of Nvidia’s H200 chip to China, it represents a step down from their most powerful offerings.
Chinese firms like Huawei have been actively working to bridge this gap with domestically developed chips, yet their performance still trails behind Nvidia’s top-tier products. Some industry analysts suggest that this continued lack of access to premium U.S. chips could eventually lead to a widening divergence in AI capabilities between the U.S. and China.
Richard Clode, portfolio manager at Janus Henderson, articulated this view, stating, “I do suspect, though that we will start seeing a divergence as that superior U.S. AI infrastructure starts iterating those models and starts making those models more capable over time in years to come. So I would expect from here we’re probably at peak relative Chinese AI capability versus the U.S.”
Even within China, there is an acknowledgment of these challenges. Lin Junyang, technical lead of Alibaba’s Qwen team, recently suggested at an AI conference in Beijing that there was a less than 20% probability of a Chinese firm surpassing U.S. tech giants in AI development within the next three to five years. He reportedly pointed to China’s computing infrastructure as being “one to two orders of magnitude larger” than that of the U.S., underscoring the scale of the disparity.
Hassabis, however, attributes the lag in frontier innovation not solely to technological restrictions but also to a difference in “mentality.”
### The DeepMind Approach: Fostering Exploratory Innovation
The DeepMind CEO drew a parallel between his company and “a modern day Bell Labs,” emphasizing an environment that cultivates “exploratory innovation” rather than merely scaling existing technologies. Bell Labs, a historic powerhouse of scientific discovery, was responsible for numerous Nobel Prize-winning breakthroughs.
“And of course, that’s already very difficult, because you need world-class engineering already to be able to do that. And China definitely has that,” Hassabis commented. He added, “The scientific innovation part that’s a lot harder. To invent something is about 100 times harder than it is to copy it. … That’s the next frontier really, and I haven’t seen evidence of that yet, but it’s very difficult.”
Hassabis, a leading figure in artificial intelligence, founded DeepMind over a decade ago. Acquired by Google in 2014, DeepMind has been instrumental in the development of Google’s advanced AI products, including the Gemini series. The recent introduction of Gemini 3 by Google aimed to reassure the market and users of its competitive standing against rivals like OpenAI, solidifying its position in the rapidly evolving AI landscape.
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/15803.html
