Artificial superintelligence may be available by 2030, but its capabilities are only likely to surpass humans in some aspects, the head of one of China’s hottest AI startups said on Tuesday.
The AI industry is abuzz with speculation about when the technology will overtake human intelligence. OpenAI CEO Sam Altman predicted last week that ASI could arrive by the end of the decade, while SoftBank Group CEO Masayoshi Son speculated last year it could exist by 2035
Zhang Peng, the CEO of Zhipu AI, said on Tuesday the concept was too vague to pin down to a specific timeline.
Supass in part, fall short in many
“People reach different conclusions when discussing this issue,” he said, as the company released its latest large language model, GLM-4.6.
“I think achieving or exceeding human intelligence levels by 2030 might mean surpassing humans in one or several aspects, but likely still falling far short in many areas.”
Founded in 2019 as a Tsinghua University spinoff, Zhipu AI has emerged as a frontrunner in China’s AI race and filed documents in April indicating plans to list on mainland Chinese markets.
In June, it was singled out by OpenAI as a fast-rising rival and characterised as an extension of Beijing’s efforts to push Chinese-developed AI abroad. Zhang said the company was “flattered” but its overseas expansion was “normal business”.
Zhang said overseas revenue had begun gaining traction, but acknowledged the company would not compete directly with US models in consumer subscriptions yet.
However, he said Zhipu AI was competing with companies like OpenAI in serving enterprise clients. The company recently launched a coding subscription plan targeting developers as part of efforts to expand direct-to-consumer revenue streams.
While many AI firms have found it difficult to confuse Chinese consumers to pay for AI services, unlike in the United States, Zhang said he believed that could change in a few years, due to growing acceptance of the value of AI and falling prices.
He said the GLM-4.6 model was an upgraded version of GLM-4.5 released in July and had enhanced capabilities in coding, reasoning, writing and agent applications.
