Why AI is next flashpoint in US-China tech rivalry

The technological rivalry between the US and China is heating up – and nowhere more fiercely than in artificial intelligence. That's hardly a surprise given its potential to boost both nations' productivity and competitiveness, and to tip the geopolitical and military balance. Prominent Chinese technology executives have conceded that the US has an edge in generative AI systems capable of completing complex cognitive tasks at lightning speed in response to simple commands. Washington's effort to curtail Chinese access to the powerful chips needed to train these new AI models – nipping Beijing's ambitions in the bud – shows how far the US will go to defend its advantage. 

1. What's the state of the competition?

It depends on what you're measuring. The US has mostly led the way in generative AI and hosted its biggest public breakthrough in November when San Francisco-based OpenAI launched the ChatGPT chatbot. This has triggered a wave of Silicon Valley investment as Alphabet Inc's Google and other rivals race OpenAI and its backer Microsoft Corp. to develop commercial applications using ever-more sophisticated AI tools. China is seen as further ahead in fields such as image recognition, at least when it comes to practical applications. Most demos of Chinese generative AI tools so far suggest they have a long way to go. But there's been an explosion in startup creation in the past few months. Virtually every single major Chinese technology player, from Baidu Inc and ByteDance Ltd to Tencent Holdings Ltd and Alibaba Group Holding Ltd, have unveiled ChatGPT contenders or announced plans to do so. In late June, Baidu said its ChatGPT-style service outperformed OpenAI's seminal product on several measures. 

2. What do the numbers show?

The US leads the way in AI investment, drawing $26.6 billion in the year to mid-June versus $4 billion for China, according to consultancy Preqin. The gap is gradually narrowing, at least in terms of deal flow. There were about 447 AI venture deals in the US in the same period, while China's deal count was equivalent to more than two-thirds of the US total, up from about half over the previous two years. From a research perspective, China has been the leader in AI journal citations since 2010. China's share of citations in AI publications stood at almost 30% in 2021, almost double that of the US, according to the latest Artificial Intelligence Index Report from Stanford University. In terms of turning ideas into real-world entrepreneurship, China lags behind historically. From 2013 to 2022, the US led all regions with the largest number of newly funded AI companies at over 4,600, 3.5 times the amount in China in the same period, according to the report.

3. Why does China's AI progress worry US officials? 

China's surveillance apparatus has employed AI for years to police crime and social unrest, and allegedly to help the country's security agencies monitor Uyghur minorities in the Xinjiang region. The US has imposed sanctions on Chinese companies including SenseTime Group Inc and Hangzhou Hikvision Digital Technology Co. as a result. China's use of AI for military purposes is another source of concern. In October, the Biden administration rolled out a series of restrictions controlling US chip exports to China stemming from concerns that China could use advanced semiconductors and AI to improve military capabilities. The US Department of Defense released a report on military and security developments in China in November that detailed Beijing's vision of "intelligentised warfare," defined by the expanded use of AI and other technologies at every level of combat. 

4. What is the US doing about it? 

Washington's approach has shifted gradually from targeted sanctions aimed at curbing Chinese development of products that could threaten US national security to a broader effort to protect US technological supremacy. Longer term, the US is weighing whether to impose curbs on investment in Chinese tech firms. This would have a significant impact as China has long relied on American money to fuel its most promising startups. Chinese technology has been a money-spinner for US investors and is seen as more focused on practical applications, or applying AI to existing business models. That contrasts with the approach of companies like Google, which have researched AI deeply for years, open-sourced some of their discoveries and remain focused on pushing barriers to innovation. 

5. What are China's limitations in AI?

Firstly, there's the problem of accessing US hardware. Powerful chipsets from Nvidia Corp and Advanced Micro Devices Inc are crucial for training large AI models, but Washington bars the most capable of these from China. Google Research scientist Zhou Yanqi said at an AI industry conference in June that the US chip ban is "a big disadvantage for Chinese startups." Chinese tech executives have argued that their generative AI models can still perform well if they tack together larger numbers of inferior chipsets to make up for lesser performance. China has plenty of hardware to marshal for its AI efforts: Billionaire Baidu founder Robin Li has said the US and China both account for roughly a third of the world's computing power. Skeptics of China's technological ambitions say true innovation requires the free-wheeling exploration and experimentation that the US cultivates but is restrained in China. Pervasive censorship means the datasets that China's aspirants are using are inherently flawed and artificially constrained, they argue. However, past experience suggests China's relative isolation is no barrier to technological success. Tencent's WeChat benefited from the absence of Facebook and Twitter to become a social media giant and Baidu became China's leader in domestic online search after Google pulled out.

6. What about regulation?

One wildcard in the race is each country's approach to regulation, given the risk of AI platforms being overwhelmed by disinformation or becoming tools for the mass violations of privacy and content rights. China has moved first to regulate its AI industry – hardly a surprise given Beijing's preoccupation with controlling online content. It plans to require a security review of AI services before they're allowed to operate. The draft regulations signal the onus for training AI algorithms and implementing censorship will fall on platform providers – a fairly significant burden. The US has begun to take steps toward regulation. But gridlock in Congress makes it hard to enact any meaningful AI policy. 

Disclaimer: This article first appeared on Bloomberg, and is published by special syndication arrangement.