A former ByteDance engineer says China's AI industry is 'far behind' the US — and the gap is getting worse

In a striking critique of China's artificial intelligence advancements, Zhang Chi, a former engineer at ByteDance and current research scientist at Peking University, argues that the nation's AI industry is actually falling further behind the United States. During a recent episode of the 'Into Asia' podcast, Zhang expressed skepticism about claims that Chinese AI models are closing the gap with their American counterparts. Zhang, who spent a year at ByteDance developing AI models before returning to academia, highlighted that the apparent progress of Chinese startups may not translate to real-world performance. He stated, 'On paper, every big tech company in China has a good model, but I don't think they're good enough.' This sentiment reflects a broader concern that many teams are prioritizing 'benchmaxxing'—focusing on optimizing for test scores rather than actual usability. Despite high-profile AI initiatives from companies like ByteDance and Alibaba, which include projects ranging from video generation tools to open-source AI systems, Zhang emphasized the challenges these entities face. Issues such as deepfake controversies and copyright disputes raise questions about the reliability of these models in practical applications. Speed of development presents another significant hurdle, according to Zhang. He pointed out that leading US firms like Google can iterate their models within three months, whereas ByteDance may only accomplish one iteration in six months. Structural disadvantages, including limited access to advanced chips and inferior training data, also plague Chinese companies. 'There's a huge difference between the infrastructure at Google and ByteDance,' he noted, stressing that the quality of data being utilized in China is insufficient. Zhang further critiqued the feedback mechanisms in place, stating that US AI products, such as ChatGPT and Claude, benefit from continuous user interaction that refines their models over time. In contrast, he warned that Chinese models risk stagnation due to their lower adoption rates. This perspective starkly contrasts with the views of other tech leaders, such as Nvidia CEO Jensen Huang and Elon Musk, who assert that China is rapidly advancing and could soon surpass the US. Alibaba chairman Joe Tsai has suggested that the outcome of the AI race will hinge more on deployment speed than model strength. Nevertheless, Zhang's insights underscore a growing concern within China's AI sector, suggesting that the divide with the US may only be widening in the foreseeable future. 'To be fair, I don't think any Chinese company can catch up with them soon,' he concluded.