Is the future of AI Chinese?

Digital China

Central to the Chinese government’s plan for the 21st century is to become the global leader in tech innovation. While substantial progress has been made towards this goal, the country has long been considered a copycat on the world stage, choosing to adapt western inventions rather than truly innovating. However, the tide is changing, with China pulling ahead on a number of emerging technologies. For AI, perhaps the most important emerging technology of the 21st century, this begs the question: is the future with China? Or will the country continue to sit firmly on the coattails of the USA?

Since the end of the cold war, China has undergone a boom in living standards unprecedented in history. 700 million people have been lifted out of absolute poverty, with a 75% rate of rural poverty transformed to near 0. A Chinese child born as the Berlin wall fell will have seen a 30x increase in GDP per capita in their lifetime. The second biggest jump, another post-cold war success story - Poland, increased less than 10x.

One of the hallmarks of modern China is the rapid development and comprehensive adoption of digital technology. Again the pace of this change is unparalleled. In 2005, 70% of households in the USA had internet access; in China 10%. Today, China is arguably the world’s most deeply digital society, with vendors barbecuing on street corners preferring payment via mobile phones and QR codes. State infrastructure projects have built the world’s largest fibre-optic network and the number of 5G terminal connections outstrips any other region on earth. This staggering change is the result of a tightly choreographed dance between tech innovators and the CCP. The state intensely encourages private companies to build world-class products but will never yield an inch of overarching control.

Do innovation and Chinese politics mix?

Chinese tech companies operate as independent private companies but are ultimately answerable to the CCP. Chinese law demands that “any organisation or citizen shall support, assist and cooperate with the state intelligence work in accordance with the law”. Most Chinese people do not access the internet beyond the ‘great firewall’ - a vast and sophisticated filter which prevents access to many western apps, such as Google and Facebook. Hosting a server behind the firewall requires a licence from the government. Firms of more than 50 people are also required to hire a party representative to sit on their boards and the CCP has shown it is not afraid to flex its muscles with big tech. The outspoken founder of marketplace giant Alibaba, Jack Ma, recently disappeared suspiciously from the public eye; this coincided with a regulator crackdown on his business operations.

To those familiar with the laissez-faire conditions that Western tech giants tend to thrive in, this sort of state intervention would seem antithetical to innovation. However, China is home to 8 out of the 10 fastest companies to reach unicorn status and the second most unicorn companies worldwide. Typifying the strength of the Chinese tech sector is its poster boy Tencent, developer of the omnipresent WeChat, which rolls social media, ride-hailing and banking into a single sleek app. Although tencent is comically aligned to the CCP, releasing an app in 2017 which allowed users to tap their phone screens to clap for President Xi’s party conference speech, they offer a user experience rivalling or even outstripping western competitors. The result is a product foreign to the west: high-quality, state-aligned software.

Despite the meteoric rise of tech and tech companies in modern China, a key sticking point remains that must be unpicked before China can overtake the USA as the world’s leading technological force. Can the Chinese system foster truly revolutionary products? While the “made in China” tag is no longer a flag of low quality, the might of the Chinese tech market has tended to yield evolution rather than revolution. Just creating a product that mirrors an original western idea for the Chinese market can be hugely lucrative, with the leading search engine Baidu generating over a billion dollars in yearly revenue only from a USP that does not challenge the CCP in the way Google would. This opportunity to create wildly successful companies from pre-existing technology strips the incentive to be truly creative and sit on the bleeding edge.

China and AI

While China is not currently challenging the US hegemony on conceptually innovative technology, change is beginning to take shape. A domestic consumer market saturated with affordable high-quality technology and a newfound sense of China as a global leader has led companies to push the boundaries and look outwards. TikTok, launched by Beijing-headquartered ByteDance in 2018, has changed the face of media consumption for generation Z and become the first Chinese software company to achieve truly global penetration.

Tiktok’s success lies in AI, its powerful recommendation engine matching content to users and keeping them glued to the platform. While these recommendation algorithms are not conceptually new, it demonstrates how Chinese companies can comfortably compete with their American counterparts.

The nature of AI research has meant that despite decades of field leadership, American outfits have been caught by China. Advances in AI don’t require a physical manufacturing process to implement and tend to be published openly, not guarded as trade secrets. As a result, a small team of machine learning engineers can replicate cutting-edge technology developed by competitors. This is inconceivable in areas such as chip manufacture and design which are closely guarded secrets and in which China’s domestic market lags severely behind and buys more than 90% of its chips from foreign companies.

But by far the sharpest tool in the Chinese AI arsenal is data. Data is the key driving force in AI, with most cutting-edge models requiring vast amounts of high-quality data to train. Often AI projects are infeasible due to a lack of data, rather than conceptual shortcomings. The current state of the art generative language model, GTP-3, uses technology that is several years old; its sheer size and the volume of training data giving rise to its staggering performance.

Official CCP policy makes it clear this is no issue in China. Data privacy laws are few and far between and it is an officially stated goal of the government is to drive economic development through vast quantities of rapidly accessible data. This is nefariously demonstrated by Hikvision, the global leader in CCTV cameras, whose AI is able to recognise and track individuals, as well as monitor their facial expressions and clothing. The company has faced accusations of complicity in the mass surveillance and opression of China’s muslim minority. They’ve also been accused of their technology being used to mark potentially dangerous individuals for re-education programmes.

Can China dominate AI?

But despite China’s data advantage and vast engineering workforce, becoming the dominant force in AI requires creativity as well as brute force. 2022’s most remarkable AI developments have come in generative image models, capable of creating images from text prompts. Although training these models requires raw computing power, their design was predicated on creative use of AI theory. Despite prioritising technological innovation, the CCP may scare away the best minds in AI, if they think their creativity will put them on a collision course with the state.

Perhaps most indicative of the difference between AI research in China and the USA is their respective publication records. Since 2016, China has published almost twice as many AI related papers than any other country. But at the two most prestigious and high profile conferences, ICML and NeurIPS, representation of US research is almost 5 times that of China’s.

With politics predicated on lack of state intervention in private business, the USA has long been the destination of choice for tech’s brightest minds. But vast markets and rich data access could be the springboard for China to brush the privacy-focussed US aside and dominate AI. However, to break the mould of Chinese tech and fulfil the CCP’s goal of becoming the global AI powerhouse, China needs more than brute-force and data, it needs creativity and invention. If the country can find that, it will be the force in AI for decades to come.