Artificial Intelligence: A National Strategic Initiative

Tencent Research Institute occupies an unusual perch in Chinese technology policy: it is the in-house think tank of a company that both operates one of the country’s largest consumer internet platforms and sits on the small roster of firms the State Council designated to lead national AI development. The volume Springer published in English in 2020 grew out of that role. It gathers analysis produced in the years immediately following the July 2017 release of the New Generation Artificial Intelligence Development Plan, and it is written by researchers who work close to the ministries, provincial governments and industrial champions the plan named. It is less a work of independent scholarship than a piece of applied policy exposition, addressed in its Chinese original to executives and officials who needed to know what the plan meant for their sector, and translated for a foreign audience that in 2020 was becoming acutely interested in how Beijing thinks about machine intelligence.

The book arrived at a specific moment in the conversation about AI and geopolitics. Kai-Fu Lee’s AI Superpowers had reached Western readers in 2018 with the argument that China’s data-rich consumer environment gave it a structural edge in applied deep learning. American policy debates were catching up: the National Security Commission on Artificial Intelligence chaired by Eric Schmidt was drafting its interim reports, the Pentagon had established the Joint Artificial Intelligence Center, and the first US export controls aimed at Chinese AI chip supply chains were being drawn up. Into that debate the Tencent volume introduced something that had been largely absent — a Chinese account, written from inside the system, of what China’s AI strategy was actually meant to be, in the vocabulary its architects used with each other.

The central argument is that AI in the 2020s is a systemic national capability, not a product line or a research area, and that leadership therefore belongs to the country that can most tightly align the layers on which such a capability rests. The authors describe these layers as infrastructure, data, algorithms, computing hardware, talent and standards, and they treat each in turn as a place where policy can either open or close the road ahead. Around this argument sits a distinctive theory of the state. Government is the natural convener that sets targets, allocates land and capital, opens datasets, funds foundational research, and enforces the standards that let separate industrial efforts interoperate. Private companies are cast as the executors that build products at scale, and universities as the pipeline that supplies engineers. The relationship is not framed as a tension to be managed but as a division of labour that Western liberal economies, in the authors’ reading, cannot easily replicate.

The chapters follow the logic of the 2017 plan itself. An early section reconstructs the plan in detail — its three horizons of 2020, 2025 and 2030, its language about becoming a “world-leading” AI power by the end of the decade, and the way responsibility for its execution is distributed across the Ministry of Science and Technology, the Ministry of Industry and Information Technology, the National Development and Reform Commission and the Cyberspace Administration. From there the book turns to the industrial architecture the plan created. The AI Open Innovation Platforms are described in some detail: Baidu given custody of autonomous driving, Alibaba of smart cities under the “City Brain” banner, Tencent of medical imaging, iFlytek of intelligent voice, and Sensetime added later for computer vision. The book treats these designations not as prizes but as assignments, each carrying an obligation to build shared infrastructure that other firms can plug into.

A further set of chapters descends from the national plan to the regional plans that translate it into local capital projects. Beijing’s Zhongguancun district is examined as the country’s densest concentration of AI research labs; Shanghai’s Zhangjiang as a chip and industrial-AI hub; Shenzhen as the manufacturing complement that turns algorithms into cameras, drones and robots; Hangzhou as the home of Alibaba and its cloud and city-management platforms. The authors treat this regional layer as more than atmospheric — it is where land, tax, procurement and hukou policy are actually held, and therefore where the plan either lands or does not.

The heart of the book, and the section a reader learns most from, is the tour through application domains. Finance is presented through the lens of risk scoring, anti-fraud and robo-advisory work at Ant Group and Ping An. Manufacturing runs through the automation of Foxconn’s assembly lines, machine-vision quality inspection in electronics factories, and predictive maintenance across the industrial base. Retail is anchored in the Alibaba group’s “New Retail” experiments and the cashless convenience formats that reached scale in Chinese cities several years before their Western equivalents. Agriculture is illustrated with drone-based spraying and crop monitoring, and by the animal-recognition systems that a handful of Chinese cloud providers deployed to track pig herds on large farms. Healthcare gets substantial space, both because it is Tencent’s own assigned platform area and because the deployment of medical imaging AI in Chinese tier-two and tier-three hospitals was, by 2020, further advanced than in most Western health systems. Transport chapters cover autonomous driving pilots in Beijing’s Yizhuang district and in Wuhan, alongside the intelligent traffic management the City Brain rolled out in Hangzhou. Education and public services close out the survey, with reference to adaptive tutoring platforms, facial recognition in schools and the integration of AI into municipal service portals.

Separate chapters take up the harder-to-see layers of the strategy. On hardware the book examines China’s push to reduce dependence on Western chips: the AI accelerators from HiSilicon, the neural-network processor designs of Cambricon, the specialised inference silicon a series of start-ups were shipping, and the long-standing gap between design and fabrication that Chinese planners understood they had not yet closed. On software the discussion turns to home-grown deep-learning frameworks — Baidu’s PaddlePaddle receives the most attention as the domestic answer to TensorFlow and PyTorch, and the broader open-source strategy is presented as an attempt to build alternatives that are not hostage to American licensing decisions. On talent the authors describe the expansion of undergraduate AI degree programmes across dozens of Chinese universities after the Ministry of Education added AI as a formal specialisation, and the return of overseas-trained researchers under recruitment programmes such as the Thousand Talents Plan. On standards the book maps the working groups inside SAC/TC28 and the industrial standards bodies where Chinese firms are attempting to write the technical norms that AI systems will eventually be certified against.

Governance and ethics receive a chapter of their own, and it is one of the more revealing passages in the book. The authors describe the Beijing AI Principles of 2019, the growing academic literature on machine-learning fairness in Chinese journals, and internal industry codes of conduct. The framing is that ethics is an engineering discipline to be integrated into the development process rather than an external constraint imposed by civil society. Data protection is discussed in the vocabulary of the Personal Information Security Specification and the drafts that would eventually become the Personal Information Protection Law, with candid acknowledgment that the balance between data availability and personal rights had to be recalibrated as consumer expectations shifted.

Comparative material runs throughout. American initiatives — the 2016 Obama-era National AI R&D Strategic Plan, the DARPA AI Next campaign, the Trump administration’s American AI Initiative executive order — are described as fragmented next to Beijing’s coordination. The European Union’s High-Level Expert Group on AI and the 2019 Ethics Guidelines are noted for their regulatory ambition and their thinner industrial base. Japan’s Society 5.0 and South Korea’s national AI plans are covered more briefly, presented as adjacent East Asian efforts working under different constraints. The comparisons are used less to argue superiority than to show that a global race is underway on several tracks at once.

The reception of the book abroad, and of the wider genre of Chinese-policy-in-translation to which it belongs, has been mixed and instructive. Analysts at the Center for Security and Emerging Technology at Georgetown, at RAND and at the Center for a New American Security have treated this kind of source as unusually valuable primary material — translations by CSET’s Ryan Fedasiuk, Jeffrey Ding’s ChinAI newsletter, and the standing analytical work of MERICS in Berlin have all leaned on documents of this type to reconstruct Chinese planning intent. At the same time, careful readers have flagged what such a book cannot do. It presents plans as if their execution were already underway; it understates the gaps between provincial press releases and industrial reality; it says little about the frictions between the AI champions and the regulators, or about the crackdowns on Ant, Didi and the tutoring sector that would follow within a year of publication and would demonstrate how the political weight above the industry could shift without warning. Critics have also noted that the volume’s treatment of military AI is thin, in a way that reflects the sensitivity of the topic in mainland Chinese publishing rather than an absence of Chinese thinking on it — the fusion between civilian AI and the People’s Liberation Army under the “military-civil fusion” doctrine is present in the background of every industrial chapter but is nowhere the explicit subject.

For a reader assembling a shelf on AI in warfare, the book plays a specific role. It is not itself a book about autonomous weapons or algorithmic targeting. It is the reference volume that explains the industrial and policy substrate from which China’s military AI programmes draw their engineers, their chips, their datasets and their frameworks. Read alongside Elsa Kania’s work at CNAS on the PLA and intelligentised warfare, or alongside Gregory Allen’s writing on Chinese AI strategy at CSIS, it fills in the civilian half of a picture whose military half is discussed elsewhere. It also pairs naturally with the annual AI Index reports from Stanford, which quantify from the outside many of the industrial trends the Tencent volume describes from the inside, and with the National Security Commission on Artificial Intelligence’s final report, which reads much of the same material as a strategic threat.

What is likely to age well in the book are the descriptions of the institutional architecture — the ministries, the platforms, the standards bodies, the recruitment programmes — because these structures persist even as their inhabitants change and their policies are recalibrated. The framing of AI as a systemic capability requiring alignment across layers has, if anything, become more widely accepted, and Western policy discourse now uses vocabulary that is closer to the book’s than it was in 2020. What has aged less well is the confident forward-projection of the 2017 plan’s milestones. The US export controls of October 2022 on advanced AI chips, the sanctions that followed on Chinese fabrication equipment, and the regulatory turn against consumer platforms inside China itself have all reshaped the environment the book was describing. The volume remains, for that reason, a period document as much as a strategy document — a careful and coherent statement of how China’s AI planners saw the road ahead at the moment before that road was diverted, and a baseline against which everything that has happened since can be measured.