What is AGI? How China’s AI firms are catching up as OpenAI Astra sparks debate

Artificial general intelligence (AGI) is a term used to describe systems that match or exceed human-level intelligence. First defined in 1997 by US physicist Mark Gubrud, the concept gained traction in the 2000s through the work of computer scientists Ben Goertzel and Shane Legg. Recent advancements in large language models (LLMs), such as OpenAI's Astra, have reignited discussions about whether AGI is now a reality.

However, experts remain divided on the matter, with some arguing that current AI models have already displayed flexible, general competence akin to human thought, while others view claims of AGI as premature.

China is actively pursuing AGI, with top executives at leading AI firms declaring their ambitions for the technology. DeepSeek's Liang Wenfeng prioritized AGI in a private investor meeting, while Z.ai's Tang Jie vowed to make AGI the company's top priority over near-term commercialisation. MiniMax founder Yan Junjie even pledged to forgo his salary until the company achieves AGI. Alibaba, the owner of the South China Morning Post, has also expressed belief in an increasingly clear trajectory toward AGI.

Despite this enthusiasm, China's government maintains a cautious stance towards AGI. The country's 15th five-year plan mentioned "exploring development paths for general artificial intelligence," but experts argue that Beijing's terminology refers to general-purpose AI rather than human-surpassing autonomous intelligence. President Xi Jinping reiterated at the World AI Conference in Shanghai that AI must remain safe, secure, and under human control.