China's AI Ambitions Hit Data Wall as Chinese-Language Training Data Runs Out

China's pursuit of next-generation AI models is encountering a significant hurdle: a critical lack of high-quality training data in Chinese. Although the US export restrictions on advanced computing chips have garnered much attention, Chinese AI specialists are increasingly cautioning that the shortage of quality data may pose an even greater existential risk, one that cannot be easily remedied by hardware alternatives. This issue is not exclusive to China; it affects AI leaders on both sides of the Pacific, with some US firms adopting aggressive strategies to maintain their competitive edge. Epoch AI, a US research organization, predicts that the global supply of quality, publicly accessible human-generated text could be depleted within six years. OpenAI co-founder Andrej Karpathy has also highlighted the potential emergence of a 'data wall' by decade's end, which could hinder model advancements without new, reliable data. Leading American laboratories are investing heavily in extracting offline human knowledge, sparking intense ethical discussions. This scenario emphasizes the increasing acknowledgment that data, alongside computing power, is vital for AI advancement, and that the lack of high-quality data could restrict AI model development globally.