China behind in LLM race but it can still win in AI, ex-Tencent AI lead says (www.scmp.com)

Former Tencent AI lead, Liu, expressed concerns about China's position in the global large language model (LLM) race, emphasizing that the country's AI industry lacks transformative "paradigm" innovations, or fanshi. Speaking to the South China Morning Post following his departure from Tencent, Liu criticized Chinese companies for primarily emulating U.S. advancements rather than developing groundbreaking technologies of their own. He pointed to the early success of DeepSeek as a notable moment but highlighted that the company has struggled to replicate that success, falling behind U.S. leaders like OpenAI and Anthropic, who continue to innovate with models like Anthropic's newly released Mythos. This situation poses significant implications for the AI/ML community, as innovators like Liu warn that merely catching up with benchmarks isn't enough. The disparity in real-world applications between Chinese and U.S. LLMs suggests a deeper issue of creativity and originality in AI advancements, which could hinder China's ability to become a leader in AI technologies. As the landscape evolves, Liu’s insights underscore the urgency for Chinese firms to foster unique innovations that redefine AI capabilities, rather than relying on existing frameworks established by U.S. counterparts.