The Digital Order and AI in China
Winning the AI race means winning everything. So let's not lose this one.
Could China Win the AI Race?
A remarkable comment in the otherwise optimistic Marc Andreesson essay “Why AI Will save the World,” which we discussed in our article “The Case for AI Optimism” was this declaration:
The single greatest risk of AI is that China wins global AI dominance and we – the United States and the West – do not.
We’d like to raise these questions: Where does China stand right now in AI? Is it a risk that China wins global AI dominance?
Chinese AI Progress
First, when it comes to AI research, it’s clear China is a strong player that comes in a close second behind the US. Stanford HAI’s 2023 AI Index report notes:
China continues to lead in total AI journal, conference, and repository publications. The United States is still ahead in terms of AI conference and repository citations, but those leads are slowly eroding. Still, the majority of the world’s large language and multimodal models (54% in 2022) are produced by American institutions
LLM releases: Chinese organisations launched 79 AI large language models since 2020, according to a report by China’s own Ministry of Science and Technology released in May. The top three Chinese AI chabots released this year are:
Ernie Bot by Baidu, which just had a 3.5 release that Baidu claims beats ChatGPT.
Tongyi Qianwen by Alibaba Group, released in April.
HunyuanAide is under development by Tencent Holdings, basing it on the Hunyian AI model that has been scored as the best AI model for the Chinese language.
Divided AI ecosystems: China bars some US technologies in China while supporting home-grown technologies companies, which has led to parallel technology ecosystems. The divisions are getting deeper to the point where American companies are skipping China’s biggest AI conference while US-based AI is denied to Chinese users:
With domestic regulators flagging risks about generative AI, ChatGPT and similar services such as Google’s Bard are unlikely to be allowed into the Chinese market, making AI another closed garden inside the Great Firewall that could benefit local tech giants.
It’s hard to evaluate exactly where the Chinese competitors are these days, but it is safe to say no Chinese company is matching where OpenAI and Google are in the AI race. With a divided ecosystems, that may continue. It should also be noted that the top-end NVidia A100 and H100 GPUs are barred from sale to China, hampering China’s development of the best AI.
Public attitudes: As shared by Stanford’s AI Index report, Chinese attitudes towards AI are much more positive than in the US: “In a 2022 IPSOS survey, 78% of Chinese respondents agreed with the statement that products and services using AI have more benefits than drawbacks. … Only 35% of sampled Americans (among the lowest of surveyed countries) agreed.”
AI technology for Government Control
The most disturbing issue with Chinese development of AI has been its use as an instrument of state control. In particular, leaked Chinese government documents have exposed how technology is used to escalate the persecution of Uighurs. The Government in China used “the power of technology to help drive industrial-scale human rights abuses” creating a “state of terror” for targeted groups in China.
The ICIJ reports that the Integrated Joint Operations Platform (IJOP), a policing platform, is used by the police and other authorities to collate personal data, along with data from facial-recognition cameras and other surveillance tools, and then uses artificial intelligence to identify categories of Xinjiang residents for detention.
… documents published by the New York Times and ICIJ show how deeply entrenched the Chinese government’s surveillance technology has become in the daily life of Xinjiang residents and underscores how imperative it is for the world to pay attention to the atrocities being carried out against minority groups there.
Chinese authorities are using AI facial recognition technology as part of this control, and not coincidentally Chinese companies are leading the world in this technology.
Regarding AI regulations, China is drafting regulations to cover generative AI. These draft regulations were described as a “mixture of sensible restrictions on AI risks and a continuation of China’s strong government tradition of aggressive intervention in the tech industry.” They both seek to protect intellectual property and privacy, but also impose a level of censorship and control, with an overall dictate that AI software should “reflect the core values of socialism.”
The Digital Order and the World Order
This recent Ted Talk by Ian Bremmer, “The Next Global Superpower Isn’t Who you think” asked the question ‘Who runs the world?’ The three global orders coming up:
The global security order: This is defined by military and diplomatic strength, and it is American-led and assured by American military might and our strategic alliances around the globe.
The global economic order: This is defined by economic power. It multi-polar, and while the US leads, China is a strong contender, the EU is the largest total combined market, and other nations are players on the field with their own strengths.
The digital order: Leading global technology companies define the digital order. While Governments ultimately have a say based on laws, the scale, scope and reach of technology companies makes them powerful as well.
When put in this context, it is clear why winning the AI race is so important: The rise of AI makes this a new world in the same sense the internet age and the atomic age create a new world. Winning the AI race is about leading in the digital order. Maintaining leadership in advanced AI technology yields economic as well as military strength as well, so AI gives you the trifecta of technology, economic and security supremacy.
AI changes everything, and winning the AI race means winning everything.
China is serious about winning the AI race, and they have made explicit plans to do so. We should therefore take them seriously:
“AI has become a new focus of international competition. AI is a strategic technology that will lead in the future; the world’s major developed countries are taking the development of AI as a major strategy to enhance national competitiveness and protect national security; intensifying the introduction of plans and strategies for this core technology, top talent, standards and regulations, etc.; and trying to seize the initiative in the new round of international science and technology competition.” - China’s ‘New Generation Artificial Intelligence Development Plan’ (2017)
The Rise of China and The American Imperative
With China determined to win in AI and some in the US determined to pause AI progress, it’s plausible that China might become dominant in AI. It would be very disturbing if China wins the AI race, and humanity’s experience of AI is as a censored instrument of social control by a central Government.
If the American perspective is to do everything we can to avoid that outcome, then the American imperative must be to win the AI race, take a pro-innovation pro-freedom approach to AI development, encourage open source.
I am cautiously optimistic that we will make AI technology that is safe, humane and aligned with values of liberty and personal autonomy. In doing so, we will beat any nation - China or others - who fail to grasp the essential element of freedom in innovation. We won’t let it become an instrument of repression or tyranny. As Jack Kemp put it, “I’ve read the final chapter, we win.”
"Our problems are manmade, therefore, they can be solved by man. And man can be as big as he wants. No problem of human destiny is beyond human beings. Man's reason and spirit have often solved the seemingly unsolvable--and we believe they can do it again.” - President John F Kennedy
Postscript - Living in a Zero Moat World
What if there is a way to have a digital order of maximally dispersed power?
In the industrial era, large corporations gained benefits of size due to economies of scale. In the technology era, powerful tech companies won the ‘gorilla game’ due to unique technology advantages (for example Qualcomm’s innovations in CDMA). Then with social media, a powerful network effect encouraged a ‘winner-take-most’ result for companies like Facebook.
In the digital world, data is the new oil, and the owners and gatekeepers of data are the ones who wield the most power. Google, with its near-monopoly hold on search engine market share due to its superior algorithms and deep reach of data, is most strong.
In the AI era, there may effects that undue the power of social media networks and data monopolization. The continued churn of innovation and the power of open source is creating breaks in the ‘moats’ that Big Tech players enjoy:
Open source data: Trillions of words of open source data can replicate proprietary data sources.
Open source model architectures: AI has advanced with open AI research and open source models. All the critical pieces of technology that make LLMs have been openly shared AI research.
Open source software: Most of the frameworks and tools to build AI models are open source software.
Low barriers to AI model development: Developers can quickly replicate proprietary models, the only bottleneck is GPU compute cost and availability. That bottleneck has been overcome with fine-tuning using LoRA.
Low switching costs for AI model users: It’s a one-line switch to change an AI model API call for AI applications. Users can switch from one model to another as easily as opening another browser tab or app on a phone.
All this adds up to an potentially open, diverse and ‘flat’ AI ecosystem where few players will have any real ‘moat’ or monopoly. It will be great if it works out that way, because power and economic benefits will be more dispersed and we all win.