China Economic Bulletin | No. 25 (11 April 2025) 

This Title Wasn’t AI Generated – But (Here’s Why) It Could Have Been “Made in China”

Authors: Anke Klara Helbig, Dennis Trescher, Sarah Lucht

Heroes, hype, and the rise of China’s AI narrative

China’s society and economy need a new national ‘hero’ and it seems recently that China’s Artificial Intelligence (AI) industry could be just the one that China’s been waiting for to wear the cape. In the first few weeks after the revolutionary update of DeepSeek, which shook global stock markets, the company saw itself fall victim to a massive onslaught of hacker attacks, forcing the system to momentarily restrict new user registration (Field 2025). While what actually happened is still akin to speculation, the story took on quite a few turns and arguably an entirely new dimension on Chinese social media. Versions of the story may vary, yet the majority seem to involve a Chinese hacker alliance formed by patriotic companies and individuals, an epic battle on the grounds formed by zeroes and ones, a secret weapon in the shape of self-developed cutting-edge new generation computing equipment and a recording of the Great Compassion Mantra, preaching that the evil-doers repent and turn away from evil (Sing & Stars满天星光和一路歌唱 2025).

A depiction on par with the epicness of the Avengers series, comedy of Adam Sandler movies, and yet at the end of the day, still somewhat far from the truth. Around the start of February, different allegedly affiliated factions had attested to never having participated in such a defense procedure, amongst others the Hongke alliance (a government affiliated party on cybersecurity and hacking-matters, allegedly forming the core of the DeepSeek defense operation) had denied any partaking in the action and hinted at the rumor originating from a form of marketing initiative (SINA新浪科技 2025).

At the end of the day the entire narrative can, with extremely high certainty, be labeled off as fictional and could also be recognized as such by most with only little need for critical thinking. Despite that, the story spread like wildfire on WeChat shorts, Chinese online forums and could also be found on youtube (Sing & Stars满天星光和一路歌唱 2025). Reasons for this could vary, such as sparking excitement in times of boredom, inciting patriotism in times of lingering economic uncertainty, or simply the beholding of a new glimmer of economic upturn in times of a post-pandemic depression. Nevertheless, through all misinformation and clouded facts, the questions remain: Why cite fake strength when actual potential lies bare to see and what does this potential specifically entail?

History of AI and ML

History sees the start of the field of AI in the 20th century. Despite there being mentions of humanoid robots dating back as early as the Zhou Dynasty (1046 BC–771 BC) in Chinese literature, factually recorded introduction of AI as a field of scientific interest in China proceeded only in the 1950s at a great laborious effort. Scientific pioneers, such as the scientist Xuesen Qian, the first to translate a piece of literature, “regarded as the forerunner of research in the field of AI” (Zhou 2023), into the Chinese language, had to struggle with the concurrent rejective stance of China towards AI. Greater progress did not take place until shortly before the 1980s, during which the Chinese scientific community was revitalized with support from the sides of the government. In the following years, a number of initiatives were launched, amongst others in the fields of intelligent information processing in 1986 and intelligent automation in 1993. Similarly, research institutions, such as Tsinghua University, kickstarted their own research into the field of AI. Chinese researchers, such as Wenjun Wu or the research partners Chien, Zhang and Zhang, received international recognition for their AI-pioneering research (Zhou 2023). 

The shift to the 21st century even further accelerated the research into AI. Emergence of aspects such as big data, neural networks and ever-improving hardware, have opened a vast playing field for researchers, containing amongst others the domains of “image classification, speech recognition, automated knowledge Q&A […] and automated driving” (Zhou 2023). As of 2015, China formally recognized the pivotal role of AI for the future development of Chinese Information Technology, firing off numerous governmental papers on the aspired course of AI progression (Zhou 2023). 

Despite these endeavors two years prior, China still experienced its “Sputnik Moment” (Lee, 2019) in 2017. In May 2017, the world champion Ke Jie found himself locked in an intense battle against one of the world’s most advanced AI systems – AlphaGo. This artificial intelligence, developed by Google, represented the pinnacle of Machine Learning (ML) and strategic computation. The battlefield was a 19-line board, scattered with small black and white stones – the deceptively simple yet profoundly complex game of Go.The deep, estimated 2,500 year old history of Go is mirrored by the game’s immense complexity. Its vast decision tree had made surpassing the world champion a monumental challenge for the AI community. Yet, on this day, AlphaGo was not merely winning against the Go world champion – it was methodically dismantling him, leaving no opportunities, gradually tightening its grip until there was no escape (Lee 2018, 1 f.). What might have seemed to the US as merely another victory of western technology, for China this symbolized a moment of shock, frustration and motivation. In the same year as Ke Jie’s defeat against AlphaGo, China had begun shifting gears with calls towards “greater funding, policy support, and national coordination for AI development” (Lee 2019). By the end of 2017, China had for the first time surpassed the USA in terms of AI venture funding sums, with startups and students being swept along with the AI hype, causing an influx of enrollments in the fields related to AI (Lee 2019).

Today, the topics of AI and ML in China are evolving at an unprecedented pace, often causing discussions to become outdated within only a short time. This leads to a situation wherein newer developments overtake each other, data, themes and numbers partially lose relevance and the creational timeline starts to blur. The constant shift makes it challenging to keep up with the latest trends, making it even more difficult to maintain relevance and enduring value of insights in the field of AI. In this field of technology, progress and development are increasing at an unrivalled pace, leading to a flood of information in the scientific and business community, with no clear answers as to wherein the future lies.

Why (and why not) open source AI 

An observable, increasingly constant development in Chinese AI lately, has shown that a prominent portion of Chinese firms such as DeepSeek and, most recently, Alibaba, have chosen to go the way of open-sourcing their models over the closed models of ChatGPT or Gemini (Butts 2025). This, while having garnered great affirmation from social media and the scientific community, has also resulted in the further deepening of an already growing controversy over the subject as to what “Open Source” specifically means for AI. 

Currently, the heated debate on open source AI revolves around whether the definition of open source is limited to defining the ability to use, study, modify and share the underlying system, or whether the definition of the open source term requires the inclusion of all pre-training structures and training weights used to teach the system. DeepSeek for example only fits the first definition, while only qualifying as an ‘open model’ according to the second definition and not a full on open source (Nest 2025). Colloquially speaking, the more frequently used definition of open source seems to be the first definition and may also, according to Meta chief artificial intelligence scientist Yann LeCun, be the reason why it seems as if China was overtaking the US in certain technologies (Mak 2025). According to Yann, the image that China was outclassing the US was not entirely correct, but had to be corrected into “open source models surpassing proprietary ones” (Mak 2025, citing Yann LeCun), with Chinese companies simply being more aligned with the conventional definitions. China’s further backing of open source is also ingrained in its inability to access more powerful Nvidia chips due to American tech restrictions, as sharing technologies eases the pressure on China’s goal of technological independence and increases the likelihood of invaluable breakthroughs in the fields of computing and AI (Mak 2025). 

Nevertheless, open source bears not only benefits, with the negative side being a possible reason for China to loosen its embrace on the open source hype. Open sourcing systems by definition means that there will be no direct monetary gains from the research and development invested. While this may only be a lower hurdle for tech-giants, it may well form an entrance barrier for future startups hoping to introduce new technology. More importantly, political grasp on AI development remains tight in China, with regulations in place regarding adherence to “core socialist values [and content which] endangers national security” (Mak 2025). Furthermore, China already has a reputation for safeguarding information from sectors where it has a position of global leadership, such as electric vehicle batteries and green energies, with a more recent example being the ban on the export of certain rare earth processing technologies in 2023 (Wischer and Trytten 2024). It is entirely possible that the ongoing sentiment of promoting AI open sourcing in China may not be long lasting, especially as this may be a new edge in scientific research, economic development and national security globally (Mak 2025).

Effects on various industries

One of the leading experts on the subject of AI is Prof. Claudia Bünte. In her book Die Chinesische KI-Revolution, Bünte (2020) predicts that the development of AI is expected to significantly impact future practices, with routine jobs increasingly performed by machines. This is to be understood as an extension of automation, following historical technological advances such as water mills, steam engines, and robots in automotive manufacturing. Currently, AI is largely specialized – “narrow AI” (Bünte 2020, 198) – and supports tasks in sectors such as healthcare, marketing, and law enforcement, where machines outperform humans in speed and cost. On the flip side however, this shift naturally will lead to great amounts of job displacements, especially in roles reliant on repetitive tasks.

As AI and ML continue to evolve, their role in law enforcement, journalism, and even customer service will become increasingly sophisticated, automating complex tasks while augmenting human capabilities. AI and ML are already visible across various sectors: in healthcare, AI-driven diagnostics and personalized treatment plans are revolutionizing patient care; in marketing, intelligent algorithms optimize campaigns and consumer interactions; in ecological research, ML aids in biodiversity conservation and climate modeling. 

In his paper Ecology with artificial intelligence and machine learning in Asia: A historical perspective and emerging trends, Ryo (2023) focuses on past and contemporary trends of AI and ML, with a central point of interest specifically set on ecological research across Asia. Southeast Asia is one of the most biodiverse regions in the world, home to a rich variety of flora, fauna and ecosystems, including rainforests and unique aquatic environments. As a result of modern day development, many of these ecosystems are under significant threat from human-induced activities, including deforestation, hydropower, dam construction, hunting and mining, all of which are major drivers of diversity loss in the region. Studying these ecosystems is crucial for both enhancing ecological comprehension and halting biodiversity decline (Ryo 2023, 9 f.). Generally, Ryo (2023) mentions that China has been the leader in AI and ML applications in Asia, with Korea, Japan, India, and Iran following behind (Ryo 2023, 8). AI and ML have become increasingly powerful tools in ecological research, allowing scientists to process and analyze large data sets, uncover complex ecological patterns and make accurate predictions on species distributions, habitat suitability, etc. Asia in particular can widely benefit from the advancement of these technologies, as ecological patterns and anthropogenic pressures there partially differ significantly from other continents. As Asian countries face unique challenges due to rapid economic growth and land development, AI and ML have been adapted to study and address these matters, pushing the boundaries of ecological research and contributing valuable insights to global conservation efforts (Ryo 2023,10 f.). Ryo highlights how AI and ML are applied in Asian ecological studies to address the mentioned challenges, demonstrating their role in advancing biodiversity conservation and reinforcing the importance of ecosystems worldwide (Ryo 2023, 10 f.). 

Similarly, in healthcare, AI aids in diagnostics but still requires human empathy in patient interactions (Bünte 2020, 198). In marketing, AI assists with image selection, freeing up time for strategy and creativity. Voice assistants like Amazon’s Alexa or Google’s Duplex already mimic human conversation (Bünte 2020, 199 f.). Advancements suggest AI might soon recognize emotions and carry out more complex tasks such as customer service and secretary work. Additionally, AI is streamlining tasks journalism, in which AI generated content in some cases cannot be distinguished from reports written by humans (Bünte 2020, 200 f.).

Although AI systems are still far from flawless, their increasing capabilities signal that many administrative roles, such as call center and translation services, could well be at risk. On the flip side, new tasks that require different skills and demand continuous learning are constantly emerging. Marketing managers in Germany, Austria and Switzerland acknowledge AI’s importance in reducing repetitive tasks and making marketing more efficient, with a noted 92% expecting its growing significance (Bünte 2020, 203).

Returning to the subject of AI in healthcare, the rise of AI was expected to enhance medical practice by supporting doctors, but research suggests AI may well outperform them despite being used as a tool. A study co-authored by Dr. Rajpurkar found radiologists often ignored accurate AI input, leading to lower diagnostic accuracy. AI alone achieved 92% accuracy, while doctors with AI assistance scored 76%, and just 74% without it. This shows that giving doctors AI tools does not guarantee better outcomes. Many physicians remain skeptical, which shows that a clear division of labor is needed. In one model, doctors gather clinical data, and AI analyzes it – a Harvard-Stanford study showed AI's accuracy dropped when it tried to collect patient information directly. Another model has AI suggest diagnoses and treatments, which doctors refine. A third approach lets AI handle routine cases, leaving complex ones to physicians. The last method especially seems to suggest strong results. A Danish study found that AI could flag normal chest X-rays, freeing radiologists for complex ones. A Swedish trial with 80,000 mammograms saw AI-assisted screening detect 20% more cancers and cut radiologist workload in half. Challenges like regulation, liability, and training persist. Still, AI offers the promise of better outcomes, shorter waits, and more efficient – even, according to some, more human – healthcare (Rajpurkar and Topol 2025).

Shifting human roles in the age of AI

In summary, while routine jobs are being replaced by AI, the evolving nature of work suggests that human roles could potentially shift towards tasks (and topics) that are more demanding in creativity, strategy and emotional intelligence. Supporting this is an ongoing trend that can be found throughout a variety of industries, namely the continuous automation of processes and the expansion of AI-related task management into human-occupied positions. 

The future of AI and ML will be driven by increasingly rapid advancements, fueled by breakthroughs like the model DeepSeek, intensifying global competition, and ever-expanding data processing capabilities. As AI models grow more sophisticated and access to vast datasets improves, innovation cycles will shorten. This leads to faster technological progress across different involved industries. Further, an acceleration will not only enhance automation and decision-making processes, but also push the boundaries of AI’s and ML’s potential in areas such as scientific research, healthcare, and creative fields. 

However, with rapid progress comes the challenge of keeping up with ethical considerations, regulatory frameworks, and societal impacts. As AI systems become more autonomous and deeply integrated into daily life, ensuring transparency, security, and fairness is and should be crucial. The global AI race, particularly between China and Western countries (USA and Europe), will continue to shape policies, investments, and technological priorities. In this fast-moving landscape, those who can harness AI’s potential while addressing its risks may very well define the next era of innovation and societal transformation.

References  

• Butts, Dylan (2025): Alibaba launches new open-source AI model for ‘cost-effective AI agents’, CNBC, 27 March (online at https://www.cnbc.com/2025/03/27/alibaba-launches-open-source-ai-model-for-cost-effective-ai-agents.html, last visited 10 April 2025).
• Bünte, Claudia (2020): Die chinesische KI-Revolution: Konsumverhalten, Marketing und Handel: Wie China mit Künstlicher Intelligenz die Wirtschaftswelt verändert. Wiesbaden: Springer.
• Field, Hayden (2025): DeepSeek hit with large-scale cyberattack, says it’s limiting registrations, CNBC, 27 January (online at https://www.cnbc.com/2025/01/27/deepseek-hit-with-large-scale-cyberattack-says-its-limiting-registrations.html, last visited 10 April 2025). 
• Lee, Kai-Fu (2018): AI Super-Powers – Chinas, Silicon Valley, and the New World Order. Boston, New York: Mariner Books.
• Lee, Kai-Fu (2019): China’s Sputnik Moment and the Sino American Battle for AI Supremacy. Asia Society, 3 December (online at https://asiasociety.org/magazine/article/chinas-sputnik-moment-and-sino-american-battle-ai-supremacy, last visited 10 April 2025). 
• Mak, Robyn (2025): China’s love of open-source AI may shut down fast, Reuters, 2 April (online at https://www.reuters.com/breakingviews/chinas-love-open-source-ai-may-shut-down-fast-2025-04-02/, last visited 10 April 2025).
• Nest, Daniel (2025): Just How Open Is “Open-Source” AI?, Why Try AI, 6 February (online at https://www.whytryai.com/p/open-source-vs-closed-source-ai, last visited 10 April 2025).
• Rajpurkar, Pranav and Topol, Eric J. (2025): The robot docter will see you now, The New York Times, 2 February (online at https://www.nytimes.com/2025/02/02/opinion/ai-doctors-medicine.html, last visited 10 April 2025).
• Ryo, Masahiro (2023): Ecology with artificial intelligence and machine learning in Asia: A historical perspective and emerging trends. Ecological Research 39(1), pp. 5. 
• SINA新浪科技 (2025): 中国红客联盟就 DeepSeek事件再发声：系营销牟利、切勿上当受骗,  SINA新浪科技, 3 February (online at https://finance.sina.com.cn/tech/discovery/2025-02-03/doc-ineifhrt7111408.shtml, last visited 10 April 2025).
• Sing & Stars满天星光和一路歌唱 (2025): 因 Deepseek 红客大战黑客 然后响起了大悲咒, YouTube, 1 February (online at https://www.youtube.com/watch?v=8YSlufb6kvE, last visited 10 April 2025).
• Wischer, Gregory and Trytten, Lyle (2024): China’s Threat to Ban Critical Mineral Export is Not a Bluff. IndustryWeek, 3 December, (online at https://www.industryweek.com/the-economy/competitiveness/article/55128029/chinas-threat-to-ban-critical-mineral-exports-is-not-a-bluff, last visited 10 April 2025).
• Zhou, Longjun (2023): A Historical Overview of Artificial Intelligence in China. Science Insights 42(6), pp. 969 ff.