Two significant trends in artificial intelligence merit joint consideration. Computing resources, once dispersed across regions and institutions, are now increasingly interconnected and coordinated. At the same time, industries such as energy, manufacturing, transportation and telecommunications are opening their doors to real-world applications, enabling innovators to test new technologies beyond laboratories.
One trend focuses on the origins of computing power, while the other determines where technology goes. Together, they signal a fundamental shift: AI innovation is moving beyond isolated technical breakthroughs toward an integrated system in which computing power, data, models, real-world applications, talent, capital and rules reinforce each other.
For China, this raises a larger question: How can the strengths of the Party leadership be better harnessed to move from supporting individual innovations to shaping a robust ecosystem that enables innovation to flourish?
Supporting a project can get an idea off the ground, but building an ecosystem creates conditions for enduring innovation — allowing research to reach the market and new ideas to emerge through competition and collaboration. Algorithms need computing power and data; models need real-world testing.
Even a functioning technology must earn users' trust, backed by clear rules on intellectual property and safety. A bottleneck at any stage can keep a promising idea from becoming a practical solution.
The first step in establishing such an ecosystem is to set priorities without dictating technological choices. Basic research, critical technologies and major scientific infrastructure require sustained investment. Their costs are high, risks substantial and returns often too distant for short-term market incentives to justify the necessary commitment.
Yet AI is evolving so rapidly that no administrator can reliably predict which technical approach will succeed or which product the market will favor. Therefore, the focus should be on establishing strategic direction, strengthening fundamental capabilities and creating conditions for experimentation — not selecting technological pathways for researchers or products for companies.
Strategic priorities require consistency; technological exploration requires flexibility. Concentrating resources on key national objectives need not mean narrowing the range of approaches pursued. Innovation thrives when competing ideas have room to develop, fail and improve.
The second priority is to build a strong public foundation for innovation. Computing infrastructure, high-quality datasets, open-source models and shared technology platforms may seem far removed from individual products, but they determine who gets to innovate.
Large companies can afford their own computing power, data and research teams. But startups and young researchers may find these costs prohibitive.
Public investment should focus on what individual companies cannot provide efficiently, have little incentive to fund or would otherwise build redundantly. Better-connected computing networks, lawful data sharing, common technology platforms and open-source communities can lower the cost of innovation for everyone.
But building infrastructure is only half the job. Computing resources that cannot be easily shared, data platforms that create new silos and facilities that fail to meet industry needs will do little to advance innovation. The real measure is not what gets built, but what people can actually access, afford and use.
The third task is to open up real-world applications. AI developers often face a familiar challenge: a technology works in the lab but doesn't have a customer. Once deployed in a factory or other complex setting, new problems emerge.
Real-world applications are not where innovation ends but where it is tested and improved. State-owned enterprises have access to major infrastructure and industrial settings.
Private companies tend to be more agile and responsive to market demand, while universities and research institutions bring expertise in basic and frontier research. Merely bringing them together on paper is not enough. Industrial problems must reach the lab, new technologies must be tested in production, and lessons from the field must feed back into research.
It is equally important that rules evolve alongside innovation. Businesses are not necessarily deterred by regulation itself, but by the uncertainty about its limits. How can training data be used lawfully? Who owns AI-generated content? Who is responsible when a model causes harm? Which applications can be tested before wider deployment?
Good governance is not about slamming the brakes on technological progress. Like traffic rules, it should set clear boundaries while showing innovators which paths are open to them. Clear rules give businesses and researchers greater confidence to invest, experiment and learn from failure.
Research evaluation matters, too. AI increasingly relies on teamwork, engineering and open collaboration, yet papers alone cannot capture every contribution.
Code, datasets, open-source models, technical tools, standards and real-world testing can be just as valuable. What gets rewarded shapes how researchers work and what they choose to pursue.
More government intervention does not automatically lead to more innovation. The challenge is to know where public authority can make a difference — and where it should leave room for others to act.
Strategic priorities must be established, shared infrastructure developed, industrial applications opened, institutional barriers removed and essential safety safeguards maintained. However, the choice of technical approach, the development of individual products and the eventual winners in the marketplace should largely be left to researchers, businesses and market competition.
As AI advances, competition is increasingly about not just individual technologies but the ecosystems that produce them.
Their strength lies not in the number of flagship models or funded projects, but in whether small research teams can access computing power, companies can test ideas in real-world settings, and promising technologies can advance under clear, predictable rules.
A breakthrough can come from a focused research effort. An era of innovation cannot be built project by project. A thriving ecosystem does more than support today's innovators; it gives rise to new ones, makes room for experimentation, keeps failure affordable and helps successful technologies reach the wider economy.
The deeper challenge for China in the age of AI is to move beyond nurturing individual projects and cultivate an innovation forest — one capable of continually producing new technologies, businesses and applications.
Yin Yunqiang is a professor at the School of Economics and Management, University of Electronic Science and Technology of China, and Wang Dujuan is a professor at the Business School, Sichuan University.
The views don't necessarily reflect those of China Daily.
If you have a specific expertise, or would like to share your thought about our stories, then send us your writings at opinion@chinadaily.com.cn, and comment@chinadaily.com.cn.
