China's artificial intelligence sector is not only lagging behind the United States but is falling further behind, according to Zhang Chi, a former engineer at ByteDance, the parent company of TikTok. Zhang, who now serves as a research scientist and assistant professor at Peking University, stated that despite impressive benchmark scores, Chinese AI models fail to deliver real-world performance comparable to their US counterparts.
Structural Disadvantages
Zhang pointed to several structural issues hindering China's AI progress. 'Google can train or perform a full round of LLM training, both pre-training and post-training, in three months,' he said. 'But ByteDance — probably we can only do one iteration in half a year.' This slower pace of innovation is compounded by limited access to advanced chips, weaker infrastructure, and lower-quality training data.
'There's a huge difference between the infrastructure at Google and ByteDance,' Zhang said. 'I don't think we're getting high-quality data.'
Focus on Benchmarks vs. Practical Use
Many Chinese tech companies, including ByteDance and Alibaba, prioritize optimizing AI models for benchmark performance rather than practical applications. Zhang referred to this practice as 'benchmaxxing,' noting that it creates a disconnect between test scores and real-world utility. Despite rolling out high-profile AI models, these companies face criticism over deepfakes, copyright disputes, and unreliable outputs.
US Advantage in User Feedback
Zhang highlighted the importance of user feedback loops, which US firms like OpenAI leverage to refine their models. Products such as ChatGPT, Claude, and Gemini improve through constant interaction with users. In contrast, Chinese models risk entering a negative cycle: 'Chinese models started not as good, so no one really uses them for really important things,' Zhang explained. 'And the models continue to be not that good.'
Contrasting Views
While Zhang's assessment paints a bleak picture, other tech leaders, including Nvidia CEO Jensen Huang and Tesla's Elon Musk, believe China is rapidly closing the gap. However, Zhang remains skeptical: 'To be fair, I don't think any Chinese company can catch up with them soon.' His perspective underscores the growing divide between China and the US in the race for AI supremacy.