A few weeks ago, China’s State Council released the AI+ initiative. This is a fairly significant document.
Matt Sheehan covers this in his newly launched substack and has this to say -
I’d rank it below the 2017 national AI plan and the 2023 generative AI regulation, but above a raft of other regulations and standards that have come out over the past ~8 years
This initiative is focused on the diffusion of AI technology across China’s economy and society. The document specifies six priority sectors for this -
AI+Science and Technology
AI+Industrial Development
AI+Consumption Upgrading
AI+People's Livelihood and Well-being
AI+Governance Capacity
AI+Global Cooperation
A few points to note here.
Firstly, as Jeffrey Ding points out in his testimony before the US-China economic Economic and Security Review commission, in his assessment, China faces a diffusion deficit. He argues that general-purpose technology diffusion, and not leading sector innovation, determines national power and he indicates that there is a serious gap between the two in the case of China.
Among other indicators, he points to weak connective tissue and linkages between Chinese academia and industry as highlighted by metrics such as joint research publications between the two, and number of university-originated patents that were commercialised. Importantly, he highlights that while high-speed railways and Chinese consumer apps have found mass scale and adoption, the same is not true for ICT adoption by businesses.
The deep integration of AI that this policy document envisages could help bridge this diffusion deficit for China in the long run, if implemented. But it’s still unclear if this focus as laid out in the document deprioritises AGI, or more broadly in Ding’s articulation, competing with the US in cutting-edge innovations in the technology. Or maybe that document is still under the hood.
The document also sets some targets -
By 2027, China aims to achieve significant progress in the deep integration of AI in six key sectors, with the penetration rate of new-generation intelligent terminals and AI agents expected to surpass 70 percent, the guideline said. The core industries of the intelligent economy will see rapid growth and the role of AI in public governance will be significantly enhanced.
The guideline also stated that by 2030, AI will empower China's high-quality development across all fronts, with the penetration rate of next-generation smart terminals and AI agents exceeding 90 percent, highlighting that the intelligent economy will become a significant growth driver for China's economy.
By 2035, China will enter a new stage of intelligent economy and intelligent society comprehensively, providing strong support for realizing socialist modernization, according to the guideline.
While highly ambitious, these targets are also extremely vague. For example, what constitutes a 70% adoption rate and how would that be measured?
In any case, Ding himself points out that the set targets are unachievable -
In the section titled, “The Decisive Years in the US-China AI Competition,” here’s what I wrote, “Thus, regardless of which arrival date is used, if AI, like previous GPTs, requires a prolonged period of gestation, substantial productivity payoffs should not materialize until the 2040s and 2050s.”
If AI follows the path of electricity and the computer, we will not even get close to 70 percent adoption across the economy — let alone 90 percent! — until after 2040.
Additionally, this initiative has some parallels to the Internet+ initiative launched in 2015. As this article explains,
Ten years later, with the rise of generative AI, China is shifting from “Internet Plus” to “AI Plus.” According to some experts, “Internet Plus” was about “connection” (连接)—linking information to redesign processes and improve efficiency. “AI Plus” builds on that by adding cognitive ability, moving from “information connection and diffusion” to “knowledge application and creation.” Its essence is “empower”(赋能). It promises to reorganize production factors, upgrade value creation models, transform organizational structures, and reshape governance.
However, it is crucial to note that integration of AI across multiple sectors is very different from transitioning offline activity online. AI integration requires skill training, redesigning workflows, restructuring organisations and upgrading processes, alongside squaring these with ethics, security and hallucination concerns that may arise from the deployment of the technology. Not as simple a plug-and-play that Internet integration was.
Of the six priority areas, the sections on Science, Industrial Policy and Global Cooperation are interesting.
Accelerate scientific discovery. Accelerate the exploration of new AI-driven research paradigms and accelerate the process of major scientific discoveries from "0 to 1." Accelerate the development and application of large scientific models, promote the intelligent upgrade of basic research platforms and major scientific and technological infrastructure, create open and shared high-quality scientific datasets, and enhance the processing of complex cross-modal scientific data. Strengthen the interdisciplinary driving role of AI and promote the integrated development of multiple disciplines.
2. Drive innovation in technology R&D models and improve efficiency. Promote the integrated and coordinated development of AI-driven technology R&D, engineering implementation, and product launch, accelerate the implementation and iterative breakthroughs of "from 1 to N" technologies, and promote the efficient transformation of innovative achievements. Support the promotion and application of intelligent R&D tools and platforms, strengthen technological collaborative innovation in areas such as AI and biomanufacturing, quantum technology, and sixth-generation mobile communications (6G). Use new scientific research results to support the implementation of scenario applications, and use new application needs to drive technological innovation breakthroughs.
The Global cooperation section particularly mentions this -
Treat AI as an international public good that benefits humanity, and foster an open ecosystem for AI capacity building that fosters equality, mutual trust, diversity, and win-win outcomes. Deepen high-level openness in the field of AI, promote open-source access to AI technology, strengthen international cooperation in areas such as computing power, data, and talent, assist countries in the Global South in strengthening their AI capacity building, empower all countries to participate equally in the intelligent development process, and bridge the global intelligence divide.
Matt Sheehan goes on to excellently analyse the document, its potential impact and hinges on a bearish perspective. Local government fiscal stress, a slowing economy, challenges faced by the Chinese private sector amidst geopolitical tensions and capital flight, paint a difficult picture for the future imagined by this policy document.
#8 of this newsletter exemplified China’s top-down, command and control model using the 2017 national plan as a case study, especially the local governments’ accumulated targets far outpacing the already ambitious central directives. I had quoted this Merics report then -
The current top-down approach in China's AI industry is thus in line with the country's overall industrial policy, in that it mobilises massive amounts of capital and labour towards a specific target even at the risk of creating inefficiencies and wasting resources.
As Matt points out, the propensity and intent for this would be far less in 2025 China than it was prior to 2020.
Arguably, China is better placed for AI diffusion than the US, where AGI dreams distort product development. In contrast, China with its electric stack might be able to embed AI in all kinds of hardware. Whether that’s a good thing or a bad thing remains to be seen.