If tariffs and blacklistings are the weapons in the China-US trade war, wait until the battle really begins over artificial intelligence, a contest that could force New Zealand to pick sides.
But visit Beijing or Shenzhen and you enter a different world, one where a burgeoning internet industry has sprung up in less than a decade and is now applying AI across industries in China – in many ways, well ahead of the West.
The district of Zhongguancun in northwest Beijing used to be a graveyard for eunuchs. Today, it is China’s Silicon Valley, a tightly packed collection of innovation parks, university labs and corporate offices for tech companies such as Lenovo, e-commerce giant JD.com and Baidu, China’s largest search engine provider.
There are now about 200 Chinese AI “unicorns” – companies valued at over US$1 billion – and many of them are based in Zhongguancun.
The Chinese Government has become an enthusiastic proponent of AI. In 2017, it announced the New Generation AI Development Plan, which laid out a strategy to lead the world in AI by 2030. Since then, it has poured billions into AI companies and university research and co-opted AI technology to apply it in education, infrastructure and military uses and, most controversially, for mass surveillance of its citizens.
Talk of an “arms race” over AI between the US and China may be overblown – for now. Certainly, AI is being developed by the US military and China’s People’s Liberation Army, which are both building weapons with varying levels of autonomy. But collaboration on fundamental AI research is still strong. China is heavily engaged with US universities and companies on AI research. Google and Microsoft both have research operations on the fringes of Zhongguancun.
The use of AI for economic and social development to boost GDP in both countries, however, will be more significant than any military application of the technology.
With President Donald Trump’s administration applying the thumbscrews to Chinese tech giant Huawei because of its dominance in 5G next-generation mobile technology, the worry for many businesspeople in China is that AI will similarly become a tool used by the two global powers to gain an edge over each other.
The US feels its dominance is under threat. Earlier this month, Trump’s chief technology adviser, Michael Kratsios, said the government and private sector needed to work together to keep the US on top.
“Our goal is very clear: the unique American ecosystem must do everything in its collective power to keep America’s lead in the AI race and build on our success.”
New Zealand’s dilemma
Every country is scrambling to build capability in AI as the technology is seen as critical to improving productivity – a measure in which New Zealand lags behind other developed nations – and industrial competitiveness.
The question New Zealand may increasingly face is where to buy its AI technology: if it chooses China, what will that mean for our relationship with the US? The stand-off over Huawei and 5G technology could play out on a larger scale given the power of AI technologies to transform entire industries and address social and environmental issues.
New Zealand investment adviser David Mahon has been doing business in China since 1985, when he opened his office in Beijing, becoming one of the first foreign private-equity managers on the mainland.
“The reason countries are moving towards artificial intelligence and robotics is that there are needs that humanity is trying to meet,” says Mahon. “China is just the same. The US is deciding to contain China’s growth because it doesn’t want the competition. It has chosen technology. It is what you do when you want to stop a rising power.”
Mahon has watched American internet companies try to establish themselves in China and largely fail, in part because of their unwillingness to accommodate the Government’s censorship and data- access demands, but also as a consequence of the strong state protectionism that has allowed the Chinese internet industry to flourish, with the likes of Alibaba, Tencent and Baidu catering to 800 million Chinese customers.
“Google is blocked in China, partly because it gathers information on users,” Mahon says. “But it is also to protect Baidu.
“There’s a price to pay for that. Ask any Chinese person and they’ll tell you Baidu sucks. China is paying a price for its attempts to manage the internet and censor it.”
Increasingly prohibited from buying American semiconductors and other technology because of US Government fears that doing so may threaten national security, Huawei and other Chinese companies are accelerating their efforts to plug holes in their technical capabilities, just as they did when China rejected the US internet giants in favour of its own.
“If China wasn’t facing the resistance that it is meeting, then the concept of self-reliance wouldn’t be as deep,” says Mahon.
“It isn’t looking to dominate, it is looking not to be dominated.”
China’s Sputnik moment
The race is now on in labs all over the world to develop so-called artificial general intelligence (see sidebar). China is a relatively new contender, fuelled by the billions being invested in its AI industry and research sector and the backing of the state, which views AI as national infrastructure as important as highways and power plants. But China is scrambling to catch up on the fundamental research that will drive the future of AI.
“The story of the birth of deep learning took place almost entirely in the US, Canada and the UK,” argues Kai-Fu Lee, the Taiwanese-born venture capitalist and former president of Google China, in his book AI Superpowers: China, Silicon Valley, and the New World Order.
“The great majority of China’s technology community didn’t properly wake up to the deep-learning revolution until its Sputnik moment in 2016.”
That was the year that Ke Jie, China’s top player of the ancient board game Go, was defeated by AlphaGo, an AI-powered computer program developed by Google-owned AI company DeepMind, which was co-founded by New Zealander Shane Legg.
AI players had long since conquered humans at chess, but Go is a much more complex game. It is like four games of chess taking place on the same board. It was played by Chinese emperors thousands of years ago and holds a special place in the culture. Lee maintains it was a wake-up call for the Chinese, similar to how the US was shocked into action when the Russians launched the Sputnik satellite into orbit in 1959.
If Silicon Valley companies, supported by computer labs at leading universities such as Stanford, MIT and Cambridge, have been responsible for most of the fundamental innovation in AI to date, Lee sees China as increasingly leading in the implementation of AI. A new generation of “ruthless” Chinese entrepreneurs has literally copied American technology and business models and created their own versions of Google, Facebook and eBay, says Lee. Now they are turning their attention to AI.
As a developing country, China has also leapfrogged the West in its use of technology, such as digital payments. It is rare in Chinese cities to see people using cash or credit cards anymore. Instead, they produce their mobile phone, waving it over a terminal to pay using WeChat or Alipay.
These companies are gathering unprecedented amounts of information from their customers. The availability of large mountains of data, whether it be financial transactions on Alipay or video recorded by companies running China’s surveillance cameras, is the key to harnessing machine learning and improving AI.
Many of these companies are largely unknown as brands outside China but they are now leveraging their scale in AI to compete for international business. Based in Hefei, eastern China, iFlytek is the country’s largest AI speech-recognition company. Founded 20 years ago by a group of university graduates, its intelligent assistants are used daily by hundreds of millions of Chinese when they call contact centres in banks, insurance firms and phone companies.
Its senior vice-president and chief financial officer, Dawei Duan, is a regular visitor to New Zealand and has signed a memorandum of understanding with Auckland business incubator The Icehouse. Duan sees three factors giving China an edge in AI: its massive population constituting the world’s biggest single market; the rising middle class, which is creating demand for AI-powered services; and the strong backing of the Government.
“Almost every industry, every company and even every person is thinking about how to use AI to change their business model or their personal life,” says Duan, who compares China’s acceleration in AI capability to the rapid development of its vast rail network. Commuting from his family home in Shanghai to Hefei each week is a 470km journey that takes less than three hours on a high-speed train. “Ten or 20 years ago, it took 12 hours.”
In 2017, iFlytek’s health AI computer became the first in the world to pass a medical examination, completing the entrance test that all doctors in China must sit. It scored in the top 12% of candidates. Now, iFlytek’s AI tools are used in hospitals to help with patient assessments, diagnosis and follow-up care. Duan says it is helping tackle doctor shortages, which are particularly acute in parts of rural China. “We have a lot of people but a lack of high-quality doctors, who may need 10 years to qualify. Our society can’t wait that long. When we introduce robotics to the fundamental healthcare systems, it helps a lot.”
He says the same underlying technology could be applied in New Zealand to ease pressure on our stretched health sector, though work would have to be done to adapt it to local requirements and the Kiwi accent.
Listed on the Shenzhen Stock Exchange with a $16.5 billion valuation, iFlytek has business interests spread through the private sector as well as the Government, where its algorithms are used in courts by judges to assess the quality of evidence. As such, it embodies President Xi Jinping’s 2030 vision for AI.
Partnerships with MIT in the US, Surrey University in the UK and, to a minor extent, Massey University in New Zealand have been key to the growth of iFlytek, which invests about 25% of its revenue in research and development.
What could disturb its growth trajectory is the current US-China trade war escalating to the point where research collaborations are severed or iFlytek, like mobile-equipment maker Huawei, finds itself on the US Commerce Department’s Entity List. It’s a list of foreign companies and individuals that US companies are prohibited from exporting to unless they are granted a licence by the US Government. Hundreds of Chinese companies are listed, because sending them technology has been deemed a risk to national security.
“Not only are Chinese companies against the trade war,” says Duan, “our US partners don’t like the trade war. It’s still a very young industry so we need global co-operation.”
The Chinese Communist Party’s absolute rule and a general sense of inevitability in Chinese society that the state is monitoring everything have allowed China to avoid the roadblocks hindering the encroachment of AI on our privacy in the West.
San Francisco recently banned government use of facial recognition and politicians are working on a bipartisan bill to limit its use throughout the US.
In China, the cameras are already part of the fabric of city infrastructure, many of them bearing the label of facial-recognition company SenseTime, which, after raising capital last year, became the world’s most valuable start-up, valued this month at US$7.5 billion. But SenseTime also wants to sell its AI technology internationally, to help steer driverless cars and detect cancers in medical images.
There’s also a growing realisation within China that to build trust in artificial intelligence for use in increasingly intimate aspects of its citizens’ lives, and for the AI industry to become a major exporter, ground rules are needed.
Yi Zeng is one of China’s leading AI researchers and ethicists. He says China urgently needs to apply a stronger ethical framework to the use of AI by both the Government and the private sector.
As a member of the National Governance Committee for New Generation Artificial Intelligence, he helped draft the Beijing AI Principles. Released in May, the 15 tenets call for the development of artificial intelligence that, among other things, respects “human privacy, dignity, freedom, autonomy and rights”. AI experts were surprised that the principles mirrored much of what’s in AI ethics guidelines drafted by Western governments, universities and tech companies.
If the Western perception is that the Chinese state has used AI in unethical ways to keep tabs on its citizens, the reality, as with most things in China, is a lot more complicated, says Zeng, deputy director of the Beijing-based Research Center for Brain-inspired Intelligence, part of the Chinese Academy of Sciences. “People raise questions about the Chinese Government and its use of AI. I can see it from both perspectives,” Zeng says. “The first principle for using it is to keep everyone safe. It reduces the crime rate, particularly in big cities.”
Zeng sees a public good being served in some of the AI-powered surveillance in operation across the country but has his reservations about the naming-and-shaming aspect of China’s emerging social-credit system. “When you go across the road because you didn’t see the red light, then your identification is shown on a screen. I don’t really like that kind of thing,” he says.
“Everyone makes mistakes. It depends on what kind of mistake you are making.”
Zeng is similarly wary of some of the ways AI is being applied to China’s vast education system, such as the use of facial-recognition technology in some schools to monitor students’ behaviour.
“People say they’ll get better grades. But in applying that type of technology, you could also have a generation that basically hates AI.
“I may not make eye contact with the teacher, I could be thinking about what the teacher in the last class said. It doesn’t mean I’m not interested in this class.”
Zeng is heading a new AI ethics and safety research centre set up about the same time the principles were published. One of its potential roles, he says, could be to create a secure “sandbox” where AI companies can send their software systems to be evaluated for safety and bias.
“You may have a recommendation system for job applicants, but only 10% of the data is related to females. That’s problematic from a bias point of view. We will tell them what the risky parts are.”
But will giants such as Tencent, a social-media provider that rivals Facebook in market value, and Alibaba, which is in a similar league to Amazon, subject themselves to such scrutiny? “It’s them coming to me because they want to understand the technical grounding of the AI principles,” he says. “They want it.”
Convincing the Government to change, he admits, may be harder. Although the 15 principles were developed with the backing of the Chinese Ministry of Science and Technology, there’s no obvious mechanism to enforce them. Will companies ever be able to resist the Government’s efforts to co-opt their technology to serve its national security agenda? “It will take many steps to realise every principle,” says Zeng. “It will take pressure from the bottom as well as from the top.”
Centre of the storm
The 9sq km Dongguan campus of Huawei, an hour’s drive north of the company’s Shenzhen headquarters, is an unlikely symbol of China’s growing technological might. The US$1.5 billion facilities are made up of replicas of 12 European cities including Paris, Granada and Verona. The finishing touches are being put on Oxford and Bruges, after which 25,000 Huawei staff will occupy the campus. An orange tram, imported from Switzerland, shuttles employees and visitors from one faux slice of Europe to another, the cloying humidity often the only clue that you are on a different continent.
It’s an architectural statement as grand as Apple Park, the massive, circular Norman Foster-designed headquarters of Apple in Cupertino, California. It is also a testament to the wealth that has flowed to Huawei since it became the world’s leading supplier of mobile network equipment and, in the space of five years, the second-largest smartphone maker behind Samsung.
The last time I was here, in 2007, Huawei wasn’t yet in the mobile-phone business and had revenue of US$12.8 billion. Last year, its revenue surpassed US$100 billion for the first time.
In a bid to convince the world it has no intention of using its new 5G mobile technology to spy on Western countries and that it is not a puppet of the Chinese Communist Party, Huawei has opened itself up to a level of public scrutiny no other tech company has faced.
The transformation started in 2013 in New Zealand, when visiting founder and chief executive Ren Zhengfei, then aged 68, held his first ever press conference in the company’s 25-year history. He used it to vehemently deny allegations that Huawei posed a risk to US national security, a claim asserted the previous year in a report by the US House Intelligence Committee.
The US antipathy towards Huawei has been “incessant” since, says Joe Kelly, Huawei’s senior vice-president of communications. The Irishman describes himself in his Twitter profile as a “business/tech hack turned in-house comms guy, allergic to BS”.
“I think there’s a realisation that when you start detaining executives, the game has changed,” says Kelly, referring to Huawei’s chief financial officer, Meng Wanzhou, the founder’s daughter, who was arrested in December in Vancouver at the behest of American authorities.
She remains under house arrest as she fights extradition to the US, which alleges that she tried to sidestep the country’s sanctions in an attempt to allow Huawei to do business with Iran.
Huawei had long been banned from contributing to mobile infrastructure in the US. But its bid to become a significant player in the US smartphone industry also failed in late 2017, when mobile carriers AT&T and Verizon abandoned plans to offer Huawei phones to customers, allegedly because of US Government demands.
The US finds itself trailing China in 5G thanks to a series of missteps in the late 1990s by Lucent, its biggest telecoms equipment maker. Lucent developed the CDMA system for mobile communications but lost the technology race to European rivals united behind the GSM standard. By the time it moved to embrace the more dominant standard in the early 2000s, it was in a financially precarious state, and its renowned Bell Labs couldn’t afford the investments in R&D to catch up.
The path to self-reliance
Huawei reported strong financial results for the six months to June 30 of US$53.8 billion, up 23% on the same period last year. But, last month, it launched its own mobile-phone operating system, HarmonyOS, as it appears increasingly likely that key aspects of Google’s Android operating system may be off-limits to it in future.
Its next flagship phone, the Mate 30, was scheduled for launch this week but, as the Listener went to press, it was reported that key Google apps would be missing from it or require a workaround to install. Huawei will be able to feature a standard, publicly available version of Android but will struggle to attract people to its phones without the full set of Google features. It’s a big blow for the company, which sold 188 million phones in the first half of the year, more than half of them in China.
“We cannot contribute to the trade discussions because it is not our battle,” Kelly tells the Listener in Huawei’s towering Shenzhen headquarters. “We don’t sell anything much in the US, so we are not making lots of products that are subject to tariffs,” he adds.
In May, Huawei was added to the Entity List and hundreds of affiliates of the company have since joined it there.
“Any vendor that wants to sell any technology that falls within the restrictions will have to formally apply for a licence,” says Kelly. “We have 1200 suppliers in the US. I don’t know how big the office is that will review these licence applications.”
This month, in an attempt to defuse tensions with the US – or to call its bluff – Huawei made an extraordinary offer: to license its entire 5G technology platform to US companies so they can manufacture the equipment, install and operate it, completely independently of Huawei.
The other key area of exposure for Huawei, and indeed Chinese tech companies in general, is in computer chipsets, which the US has expertise in producing. Huawei has its own chip division, HiSilicon, but analysts suggest Chinese industry is still several years away from being able to fulfil its own chip requirements.
Huawei has, as a result, fast-tracked what Kelly calls “plan B”, its efforts to become as self-sufficient as possible.
“We are not being held back by a lack of components. There is no issue with supporting existing networks. If you’ve got a Huawei phone today, it will continue working with Google services, regardless of the outcome,” he says.
“We are moving forward with plan B, but not because we want to. We want to stay with the global supply chain. If you are going to enforce us to become more self-reliant, you are forcing it to happen.”
Technology correspondent Peter Griffin visited Shenzhen as a guest of Huawei and self-funded the rest of his China trip.
Artificial intelligence in general describes the development of computer systems that can perform tasks normally requiring human intelligence.
Narrow AI – AI that is focused on one narrow task, such as voice translation or facial recognition. All currently existing AI systems are narrow AI.
Machine learning – a subset of artificial intelligence that often uses statistical techniques to give computers the ability to “learn” with data – creating models to process, interpret, and respond – without being explicitly programmed with a predefined set of rules.
Artificial general intelligence – the as-yet hypothetical existence of intelligent machines that can successfully perform any intellectual task that a biological human can. It’s still in the realm of science fiction.
Source – New Zealand AI Forum
Scientists need to understand emotions to take AI to the next level.
But these tools, although useful, constitute what researchers call “narrow” AI. These are subsets of cognitive functions. They perform tasks better than humans can, such as translating Chinese into English or spotting a tumour on a mammogram image.
Zeng is deputy director of the Research Center for Brain-inspired Intelligence, a Chinese Government agency. “[Brain-inspired intelligence] is different from deep learning and from symbolic AI,” Zeng says. “You are simulating different types of neurons and their spike in activity and patterns, and [trying] to simulate different brain areas and see how they are co-ordinating with each other to realise more complex cognitive functions.”
The goal of many AI scientists is to achieve artificial general intelligence. That’s when robots can do many things at once in complicated and unpredictable real-world scenarios, making their own decisions, learning and adapting just like humans.
The machine-learning method that has advanced AI the most is known as deep learning, which is inspired by neural networks in the brain. By presenting an artificial neural network with lots of photos of cats, for instance, the network becomes adept at quickly identifying cats in millions of images online. But, Zeng says, we need to gain a much deeper understanding of how the plastic and dynamic human brain works – including what underpins our emotions, self-awareness and how we view the world – if we are to get beyond narrow AI.
More importantly, understanding the human brain and its hundreds of cognitive functions is essential if we want to avoid AI turning on its human masters. “This is why we prefer brain-inspired AI, rather than a brain simulation. With simulation, you’ll get every aspect of a human. We need to build an evolutionary network model that maximises the altruistic behaviour and avoids the negative parts,” he says. “If super-intelligence is really super, it should be super in an altruistic way. If not, it hasn’t been well developed. It should be kept as narrow AI.”
This article was first published in the September 28, 2019 issue of the New Zealand Listener.