Seizing the Golden Window for AI Development

Wei Kai (Leader of the China Artificial Intelligence Industry Alliance and Director of the AI Research Institute of the China Academy of Information and Communications Technology)

Currently, artificial intelligence (AI) is deeply embedding itself into various sectors of the economy and society at an unprecedented speed, leading to profound changes in industrial structure, production methods, and employment forms. To effectively build a support system, prevent strategic risks, and seize development opportunities in the fierce competition of technology, it is necessary to correctly view this trend.

Data, computing power, and algorithms are the three pillars of AI development. The capabilities of AI stem from learning. Data serves as the sample for learning, and the abilities acquired by the system come from the data provided; the training process of AI consumes computing power; and how to learn effectively from data and use computing power efficiently relies on algorithms. All three elements are indispensable for the development of AI.

Currently, the country is promoting the construction of a national integrated computing power system. The 14th Five-Year Plan clearly states the need to “build a multi-level computing power facility system and a national integrated computing power network.” In recent years, the government has been advancing the “East Data West Computing” project, which mainly allocates computing power for model training to areas rich in energy and with suitable climate conditions to reduce training costs. For scenarios that are sensitive to latency or have high security requirements, small computing power deployments at the edge need to be considered as a supplement.

Behind computing power is electricity. Both training and inference of AI consume a large amount of computing power, which essentially translates to electricity consumption. This year, the government work report first introduced the concept of “computing-electricity coordination,” which aims to promote a more organic alignment between AI computing power consumption and future renewable energy layouts to maximize overall efficiency. Currently, the specific implementation of this concept is still being explored, whether at the micro level of technical coordination or the macro level of national productivity layout coordination, further solutions are needed. It is clear that this is a highly certain trend.

Currently, there is a surge of enthusiasm for investment and financing in the domestic AI sector. Recently, the company Moonlight Kimi completed a new round of financing of about $2 billion, with a post-investment valuation exceeding $20 billion. The development of large models and intelligent agents requires substantial financial support and has high demands for talent density, thus necessitating adequate funding to incentivize researchers’ innovation. From an overall scale perspective, investment and financing in China’s AI sector still lag significantly behind that of the United States. Observing core indicators, the overall computing power consumption is currently in a state of supply not meeting demand, indicating a healthy momentum for industrial development. It can be said that the AI industry is currently in a phase of “both supply and demand flourishing,” with further room for acceleration in development. Meanwhile, expectations for applications are generally high, with a pressing desire to fully unleash the productivity of AI.

It is undeniable that while unleashing productivity, AI also impacts traditional employment. On one hand, it triggers a crisis of job replacement; on the other hand, it creates the possibility of “one-person companies.” In the wave of technology, digital literacy is becoming a core competency for future workers.

Today’s AI resembles a “digital employee” or “digital partner” and is not yet capable of being an independent actor that executes tasks autonomously and stably throughout. Just as human hands should not leave the steering wheel before autonomous driving technology matures, current AI applications are also in the “human-machine co-driving” stage. This presents challenges for workers but also brings opportunities for advancement.

With the empowerment of technology, in the future, everyone will have the opportunity to “lead” multiple “digital employees” and may become managers of digital teams. Skills such as decision-making, vision, taste, and judgment will become increasingly important. If one continues to cling to traditional skills and cognition, workers will face the risk of being eliminated. Therefore, individuals, especially the new generation of young people, urgently need to develop “digital leadership.”

As the AI industry rapidly develops, various security issues have also emerged. AI security issues are broad, and establishing a “safety boundary” for AI development requires addressing both supply and demand sides.

On the supply side, technology providers need to conduct internal safety research on models, ensuring that released models undergo rigorous safety testing and establish protections and barriers against content safety, output safety, and malicious attacks.

On the demand side, users, enterprises, and government departments need to deeply understand AI security issues and cannot blindly rely on AI systems. For example, in the recent “lobster craze,” some companies and individuals hastily installed intelligent agents to take over local data without sufficient understanding, leading to some risks. It is evident that if one does not understand the permission boundaries of AI, it can potentially harm information systems and lead to sensitive information leakage. Individual users, in particular, need to remain vigilant and not blindly trust the answers provided by the system.

Promoting the healthy development of AI requires society as a whole to enhance AI literacy, including how to correctly use AI and understand its safety boundaries. In recent years, the national level has been continuously strengthening AI safety work, promoting ethical reviews and safety testing, and all parties should clarify red lines to develop and use AI within legal and compliant frameworks.

Overall, AI development is currently entering a critical stage where technology and applications are deeply intertwined, and opportunities and challenges are amplifying simultaneously. In the next five years, technology will accelerate iteration and upgrade, and the applications of AI rooted in the real economy will become increasingly numerous. Seizing this golden window period and effectively integrating AI with the real economy relies on the joint efforts of all parties.