Business

China pushes toward task-oriented AI

Firms leveraging industrial heft to develop dynamic applications, system integration

By LI JING | China Daily

Updated: 2026-11-16 07:38

Share - WeChat
A new generation of intelligent bionic hand from BrainCo is showcased during the 2026 WIC Wuzhen Summit on Nov 2. CHINA DAILY

The economic logic is straightforward on the factory floor, he noted, pointing to automotive assembly. A paintwork blemish caught and buffed out on a domestic assembly line costs a fraction of what it would take to rectify once the vehicle has been shipped overseas to an export market such as Australia.

However, moving from identifying problems to performing physical tasks brings additional engineering challenges.

Unitree Robotics founder Wang Xingxing told a summit forum that robots can achieve success rates approaching 100 percent on certain fixed tasks after sufficient training, although their overall efficiency still needs to improve.

More complex operations, such as assembling mobile phones and threading needles, remain challenging. Wang said 2027 could be a turning point for the robotics industry.

Zhang, however, cautioned that developing general-purpose robots would take considerably longer time.

Wang also pointed to opportunities in overseas markets, where Chinese robotics companies can supply hardware platforms that international customers adapt for local applications.

He said overseas customers had accounted for a substantial share of Unitree's business in recent years, with international partners adapting its robots for applications including construction and commercial services.

For businesses considering wider AI adoption, however, the investment calculation extends beyond the initial cost of acquiring a model or machine.

Siroky said companies moving AI projects beyond the pilot stage sometimes found that their existing networks or electricity supplies were insufficient to support the additional computing demand. Smaller models running on a company's own equipment could be more suitable for certain applications and help avoid recurring cloud-computing charges for every request.

"You don't always need the largest, most expensive system," Siroky said. "Sometimes you only require the model that can do the job reliably, without compromising on the value or business outcome, at a lower cost."

Reliable deployment also requires cooperation on standards, testing and safeguards. John Higgins, chair of the International AI Governance Association, said applications of AI in healthcare and transportation require trust supported by evidence. He called for greater international cooperation in testing, evaluation and research.

At a Wuzhen forum, standards organizations and companies also discussed agent identities, communication protocols and safety requirements, with the aim of enabling AI agents developed by different providers to work together.

These requirements are creating business opportunities in cybersecurity and AI governance as well. Fan said risks including malicious instructions, data leakage and AI agents exceeding their authorized permissions are generating demand for model evaluation and protection services.

"The more open source thrives, the more important security becomes," he said.

The economic benefits of AI will depend not just on the growth of model developers, computing infrastructure providers and specialist service companies, but also on whether businesses using the technology achieve lasting improvements in productivity and develop products and services that generate demand.

Zhu said the key indicators to watch include lasting improvements in productivity, the development of new products, stronger domestic suppliers and job creation.

"If only a few leading companies benefit, AI's impact on overall domestic demand and employment will be greatly reduced," he said.

|<< Prev 1 2 3   
Copyright 1994 -

Registration Number: 130349

Mobile