Leading Chinese computer scientist Tang Jie stated at an event in Shanghai on Saturday that recursive self-improvement is the primary research challenge for achieving artificial general intelligence (AGI).
Tang, who serves as a computer science professor at Tsinghua University, argued that using artificial intelligence to construct smarter systems will make AGI more attainable if performed safely.
He delivered a keynote speech at the World Laureates Forum, where he described recursive self-improvement as a “new paradigm” of research and development that would speed up AI progress beyond current rates.
“Probably … the world should organise something together so that we can guarantee the safety for not just one country but for all the world,” Tang Jie said. Tang is also the co-founder of Z.ai, a Beijing-based company that developed the GLM family of models and was spun out of his Tsinghua University laboratory in 2019.
Last month, Z.ai reported that its GLM-5.3 model assisted in optimising the company's inference infrastructure for Chinese chips. Other industry figures, such as Moonshot AI CEO Yang Zhilin, are involved in the sector's rapid evolution.
Data from Anthropic in September showed that AI was responsible for 26 per cent of the company's internal research and development in August, a significant rise from approximately 1 per cent in March. Despite these advancements, some experts have cautioned that recursive self-improvement might cause humans to lose control over AI if the development pace exceeds the ability to understand or intervene.
