Zhipu AI CEO: AI may surpass human intelligence in some areas by 2030 but will likely fall short in many others

BEIJING: “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,” said Zhipu AI CEO Zhang Peng, noting that people reach different conclusions on the issue.

According to Reuters, the Chinese startup has just released its latest large language model (LLM), GLM-4.6, an upgrade to its GLM-4.5 model in July, with enhanced coding, reasoning, writing, and agent applications.

Speculation about artificial superintelligence overtaking human intelligence has intensified across the industry. Last week, OpenAI’s Sam Altman said it could be here before the decade ends, while SoftBank’s Masayoshi Son said last year that it could exist by 2035, but Mr Zhang described the concept as too vague to tie to a specific timeline.

Founded in 2019 as a spinoff from Tsinghua University, Zhipu AI indicated listing on mainland Chinese markets in April. In June, OpenAI pointed to Zhipu as a rising competitor and linked its expansion to Beijing’s push to promote Chinese-developed AI abroad.

Mr Zhang said the comparison was flattering but described the company’s overseas moves as simply part of normal business.

While the company’s revenue overseas had started gaining traction, he noted that they would not yet go head-to-head with US firms in consumer subscriptions. For now, the company is focusing on enterprise clients, where it competes more directly with firms like OpenAI.

Zhipu recently launched a coding subscription aimed at developers to expand revenue. Although Chinese consumers have been slower to pay for AI services than their US counterparts, Mr Zhang believes this will change in the coming years as prices fall and awareness of AI’s value grows. /TISG

Read also: Singapore ranks second among global financial centres set for significant growth