Inside the AI revolution that’s reshaping Chinese society
Artificial intelligence, once a novelty, is now being applied in everyday life. From academia to business, government and the military, ambitious China is betting big on AI, raising US suspicions yet offering opportunities for collaboration
Seven-year-old Chen Jiahao has a problem sum he can’t solve and he can’t wait to get home from school to pose the question to his all-knowing maths tutor.
His tutor is amazing, the boy says. Just snap a photograph of the question and the tutor will provide every possible approach to solve the problem, step by step – all in a split second.
Jiahao’s tutor is inside his mother’s smartphone. It is, in fact, an app that draws on artificial intelligence (AI) technology to solve challenging maths problems for primary school children.
And it’s just one of many AI-enabled apps Jiahao uses daily on his mother’s phone. When the boy started primary school in Beijing, his teachers recommended that his parents install the apps on their phones. The software give out school assignments, grade pupils’ work and even generate unique sets of exercises for each child based on their areas of weakness.
“Jiahao likes his AI teachers,” said his mother, Yu Ting, adding that her son spends at least two hours on the AI apps every day. “He greets my phone as eagerly as he greets me.”
Jiahao’s story shows how AI is shaping modern Chinese society as the technology shifts slowly but surely from the realm of mere novelty towards common, everyday application.
On Chinese social media, video-sharing platforms and shopping sites, AI technology is already widely used to cater specifically to individual tastes and preferences.
For example, online news aggregator Toutiao provides a selection of articles tailored for its users based on information such as their age, gender and location. Video-streaming website iQiyi recommends programmes based on users’ search and viewing history.
Ali Xiaomi, the AI-powered customer service chatbot of tech giant Alibaba, which owns the South China Morning Post, can reply to a million text queries and takes thousands of phone calls from online shoppers every day. The use of AI has cut e-shops’ customer service costs by 90 per cent, according to Alibaba.
That’s not all. An AI traffic controller introduced on trial in Hangzhou in Zhejiang province last year eased vehicle flow on roads, allowing cars to pass at speeds of up to 11 per cent faster than usual, state-owned broadcaster China Radio International reported.
In April, search giant Baidu’s AI system reunited a couple in Chongqing with their long-lost son. The machine analysed a photograph of the six-year-old boy, who went missing 27 years ago, and matched it to the face of a 33-year-old man in Fujian province, the Beijing Evening News reported. DNA tests confirmed the match.
What, exactly, is AI?
Popular culture, especially in the West, often either romanticises the notion of artificial intelligence – such as in the 2013 Hollywood film Her, in which a lonely man falls for his “female” AI operating system – or portrays it negatively, as in the hit US television series Westworld, where oppressed androids in an AI theme park turn murderous against their abusive human guests.
In reality, AI technology – at least in its current stage – is both less romantic and frightening, but its possibilities may be every bit as boundless as imagined in the movies.
AI refers to a computer software that mimics intelligent human behaviour. Creating such intelligent systems requires teaching machines to learn for themselves – an application of AI known as machine learning – rather than manually teaching them everything there is to know.
Machine learning involves feeding computer systems with large volumes of data and programming the systems to interpret the information for themselves through pattern recognition. The machines hence “learn” by calculating probabilities and drawing conclusions from patterns found in the data at its disposal.
A powerful form of machine learning is deep learning, which categorises information according to hierarchical layers of concepts. The arrangement allows systems to interpret complex data with greater flexibility, speed and accuracy.
“AI is like a child,” said Professor Feng Jufu, a machine learning scientist at Peking University’s school of electronics engineering and computer science. “The more people use it, the faster it learns. The more it learns, the faster it improves.”
United States’ rising rival
China, whose population of 1.38 billion people makes it the world’s biggest user base and data pool, is a “paradise” for machine learning technology, Feng said.
And the nation – from its computer scientists, tech businesses, the government and military – is exercising its competitive advantage.
For decades, AI initiatives have been launched and developed in the United States and the field dominated by American experts. But now, the balance appears to be tipping as China’s AI technology comes up neck and neck with that of the US.
There was no clearer demonstration of this shift than what occurred at the annual meeting of the world’s biggest AI research community this year.
The Association for the Advancement of Artificial Intelligence rescheduled its yearly event in New Orleans, originally set in January, to the following month after finding out that the dates conflicted with China’s Lunar New Year holiday, The Atlantic magazine reported.
“Our organisation had to almost turn on a dime and change the conference venue to hold it a week later,” the association’s president Subbarao Kambhampati was quoted as saying.
The clash might not have mattered in the past, but with Chinese scientists now producing more research papers on deep learning than Americans, the meeting would have been pointless if the Chinese could not attend, according to the association.
“The race is tight,” said Li Xiaowei, executive director of the Chinese Academy of Sciences’ State Key Laboratory of Computer Architecture in Beijing. China has only one main competitor – the US – and its goal was to beat its rival on the other side of the Pacific, he added.
Li said he and colleagues were developing computer chips, built specifically for machine learning, that would significantly boost the speed of an AI system, running “as fast as a car against a bicycle” compared with existing AI machines on traditional CPUs.
Chinese researchers have already developed AI chips with faster performance on specific tasks such as image recognition and natural language processing, but they still consumed more energy and tended to be less reliable than American-made chips, Li admitted.
While US experts are still making most of the fundamental AI breakthroughs, this may soon change as Chinese tech giants like Baidu, Alibaba and Tencent, with access to the vast amounts of data needed for AI training through their millions of customers, inject massive investment into the technology, setting up their own AI research laboratories to create new products at a speed and scale never before seen in the West.
China was the world’s second biggest investor in AI enterprises last year, injecting US$2.6 billion into the sector, according to Chinese think tank the Wuzhen Institute. The US topped the list with US$17.9 billion.
Smaller firms aren’t passing up the chance to make a foray into AI either. This month, AI robots owned by two Chinese education-tech companies sat the maths section of the nation’s notoriously difficult college examinations.
One, which took the test in an isolated room at a technology park in Chengdu in Sichuan province, scored 70 per cent, spending about 20 minutes completing the questions. The other, which sat the test in Beijing and was connected to the internet, scored 90 per cent in less than 10 minutes. The second bot’s score was good enough to gain admission into China’s top universities.
“Artificial intelligence has undergone several waves of hype over the past decades, but this time it’s different. This time, it may really come alive,” said Feng, the Peking University academic.
Over in the public sector, the Chinese government has pinpointed AI as a key area for advancement in its latest five-year plan. Top technology official Wan Gang said in March a national development plan was being drafted that would see AI technology adopted in areas including “the economy, social welfare, environmental protection and national security”.
Last year, the Chinese government said it would create an AI market worth more than US$15 billion by 2018. Beijing has already sunk millions into studying AI in universities and research institutes around the nation. It is also already applying the technology across the full spectrum of governance.
Traffic authorities in the city of Jiaxing in Zhejiang province this month began using an AI coach in a driving school. The system monitors students’ driving behaviour and detects mistakes they make, instructing them through a speaker and rating their performance, the Jiaxing Daily reported. The passing rate of students who learned with the AI coach was 20 per cent higher than those who had human coaches.
Over in China’s most innovative city of Shenzhen in Guangdong province, the use of a tiny chip in public surveillance cameras has helped police crack hundreds of cases and find several lost children. The intelligent chip whittles down the speed of human facial recognition to just a few seconds.
And in Jiangsu province’s city of Nantong, an AI judge will be put into use later this year to organise and analyse legal documents and material as well as perform paper work to lighten the workload for human judges. The system is expected to speed up the handling of legal cases by 30 per cent, the Nantong Daily reported.
US suspicions (and collaboration)
China has also ventured into AI on the military front. The nation is developing cruise missiles with “a very high level of artificial intelligence and automation”, the China Daily quoted a senior missile designer as saying last August.
As the country’s AI capabilities grow, so have US suspicions. The Pentagon had concerns about Beijing’s access to US-developed AI technology, the Reuters news agency reported this month.
Citing a leaked document, the report said the US defence department recommended blocking Chinese organisations from investing in some American start-ups working on cutting-edge technologies. The report suggested Washington fears that its advanced algorithms might be re-purposed for Chinese military use.
Individual Chinese AI researchers might also have become a concern for the US government, according to Zhang Lijun, an associate professor of computer science with Nanjing University’s learning and mining from data group in Jiangsu province.
“Each time we go to the US for an academic conference, we encounter extensive background checks by the US embassy,” Zhang said. But if the US stopped issuing visas to Chinese AI scientists, the move would do as much damage to America as it would to China, he added.
Despite Washington’s concerns, American companies are still flocking to join hands with their Chinese counterparts in AI research given the sheer amount of funds injected into the industry. And the collaborations have seen considerable progress in the field.
In May, Google’s AlphaGo AI programme – developed to play the Chinese board game Go – defeated world champion Ke Jie in a series of three matches, all of which the machine won.
The strategy game, played on a 19x19 grid board with more permutations than the estimated number of atoms in the observable universe, was previously believed too sophisticated for the machine to handle. Scientists had predicted AI would take at least a decade to decisively conquer the game; the final match took less than four hours.
The same month, Microsoft’s Chinese-language chatbot Xiaoice published the world’s first collection of AI-authored poems in a book titled The Sunlight That Lost The Glass Window. The book caused a stir among China’s literary circle, with some poets hoping the technology would revive appreciation of the art. Pirated copies have already appeared on Chinese websites, reflecting interest in the book.
“The US is good at coming up with new ideas in fundamental research while China is good at implementing these ideas in applications. International collaboration has played a key role in the rapid development of AI technology in recent years,” Zhang at Nanjing University said.
All these advancements are just the beginning of an AI revolution, according to the Peking University academic Feng.
“The only limit is your imagination,” he said, adding that AI technology could have even broader and deeper applications in people’s lives.
The Chinese Academy of Sciences’ expert Li concurred. The AI user experience of the future would be vastly different from today, as with software and hardware upgrades newer AI machines would become far faster and more human-like.
For instance, Feng said, today’s exam taking robot could be developed into an exam-setting machine. Like AlphaGo considered permutations never conceived by human players in the past, the AI system could pose students challenging questions that would push them to achieve results beyond what they thought possible.
“If you can answer maths questions designed by machines, you should then be able to easily tackle questions designed by humans,” he said.
But Professor Li Qingan, an educational psychologist at Beijing Normal University, cautioned against the unregulated use of AI in schools.
“Artificial intelligence may create super students, but it can also turn them into cold-blooded creatures with little care for how others think and feel,” Li said. “Thirty years from now, we may regret giving our children over to AI.”
There is also a limit to AI systems, according to professor Huang Biqing, a robotics scientist with Tsinghua University.
“If human-generated data can no longer improve an AI system’s performance, the machine will treat it as noise,” Huang said, adding that this meant the system would regard human input as no longer necessary and could evolve based on its own machine-generated data.
Chen Yi, the father of Jiahao the primary schoolboy who loves his AI tutors, recalls his childhood addiction to Nintendo’s Game Boy as he observes his son interacting with the apps on his wife’s smartphone.
“This is different from my childhood addiction,” Chen said, referring to Jiahao’s attachment to the AI-enabled programmes. “Jiahao’s condition is more like, I don’t know, a kind of dependence?”