China’s AI Dreams Aren’t for Everyone
The Chinese government has big plans for artificial intelligence. Can it make them a reality in its education system?
One morning, Ming Ming wakes up for school to the soothing, mechanical voice of his artificially intelligent robot housekeeper. “Ming Ming! It’s March 29, 2028, and a new day has begun!”
This is not a work of science fiction. It is the opening chapter of a new Chinese high school textbook, put out by SenseTime, the world’s largest artificial intelligence start-up, with a valuation of over $4.5 billion, in partnership with a research center at East China Normal University and with middle and high school teachers in Shanghai. Published in April 2018, the textbook is part of the government’s recent push to prepare Chinese youth to help the nation become an AI superpower. According to SenseTime, it is currently being taught in pilot programs in more than 100 schools throughout the country in Shanghai, Beijing, Shanxi, Shandong, Guangdong, Jiangsu, and Heilongjiang. SenseTime is training over 900 teachers to implement its curricula.
In July 2017, the government issued an ambitious master plan to lead the world in AI research and deployment by 2030. The road map outlined the steps by which AI will be deployed in areas such as military readiness and city planning, and the government announced that AI courses would be included in all primary and secondary schools. In response, the Chinese Ministry of Education has drafted its own “AI Innovation Action Plan for Colleges and Universities,” calling for 50 world-class AI textbooks, 50 national-level online AI courses, and 50 AI research centers to be established by 2020.
But it’s a long way from buzzwords to reality. The details of implementation—how these courses will be taught, what these textbooks contain, and how new curricula will be rolled out nationwide—remain blurrily defined. From the elite boarding schools of cosmopolitan Shanghai to the migrant classrooms of China’s rural outskirts, integrating the government’s top-down mandate into the existing realities of the Chinese education system may be unworkable. The advent of AI in education may not address the imminent threat of job displacement—but it could exacerbate a deepening digital divide and corresponding economic inequality.
One of the most glaring challenges driving China’s education push is job displacement. The venture capitalist and AI expert Kai-Fu Lee believes that 40 percent of jobs globally will be lost to automation in the next 15 to 25 years. In China, according to a report by PwC, the largest net job losses will most likely be in agriculture—although it’s difficult to distinguish those disappearing jobs from the shift away from agriculture that a developing economy hopes for. On the other hand, the demand in China for AI professionals, such as machine learning engineers and data scientists, may surge to 5 million in the next few years, according to a 2017 report from the Tencent Research Institute. China may not be able to solve its talent shortage in time. “Just as it was impossible for us to predict that an Uber driver or a Didi driver would be a job 10 years ago, to predict what jobs exist in eight to 10 years is nearly impossible,” said Lee.
Faced with the uncertainty of the job market, the question that educators must consider, said Yong Zhao, a professor specializing in Chinese education at the University of Kansas, is how to educate young people for jobs that don’t exist. Lee said that schools must prioritize teaching 21st-century learning skills commonly known among global educators as the “Three C’s”: creativity, collaboration, and critical thinking.
However, China’s current education system—rigidly standardized, outcome-driven, and heavily metrics-based—is poorly equipped to teach these skills. Chinese high school education, for example, revolves almost entirely around preparing students for the gaokao, its national university entrance exam, which has been criticized for its excessive emphasis on test scores and rote memorization. If educational training for the AI job market should create generalists—capable of thinking critically across disciplines and working collaboratively in teams—many of China’s schools are effectively preparing students to do the exact opposite, developing the skills to carry out specialized, routine, and repetitive tasks.
“In China, math education means making students memorize a lot of formulas and do a lot of test questions,” said Jiang Xueqin, a Chengdu-based educator. “It is making students into human calculators, uniquely designed to make them irrelevant in the age of AI.” In contrast to math education that emphasizes the teaching of math principles and strategies, which would enable students to tackle problems they had not yet previously encountered, in China, the commonly used approach is to cover as many individual math problems as possible and memorize shortcuts without grasping the fundamental logic underlying the solutions, Jiang explained.
Furthermore, the schools that will have the resources to introduce innovative new pedagogies will most likely be the nation’s most elite. The government has only issued a vague directive. Educators across the country must interpret and implement that directive according to their own resources and capabilities. For the top-ranking schools, that mandate might mean creating an interdisciplinary robotics course; for others, it may translate into more coding classes; for many, especially the poorly equipped schools in the countryside, it could simply entail having students memorize propaganda extolling the virtues of patriotic AI.
A 2019 report by the Mercator Institute for China Studies on creativity in Chinese schools points to a widening gap between China’s poorer and more prosperous regions. Wealthy schools “offer a variety of novel teaching approaches, like maker spaces, escape rooms and advanced computer classes. … [Their poorer counterparts] are struggling with limited resources and are struggling to establish creativity-fostering environments.” Jiang, who trains teachers at both elite and average kindergartens and primary schools in Chengdu, said the difference is “night and day.” Whereas teachers at elite schools tend to be open to new ideas and training students, typically the offspring of party cadres in Chengdu, to be “confident and creative,” Jiang said, average public schools “don’t care about creativity and just teach to the test.”
Jiang described a two-tier system in China in terms of access to quality high school education: one level of “elite schools trying to model themselves off Exeter and Hotchkiss, and the rest.” The fictional Ming Ming of SenseTime’s textbook is described as the son of a leading AI engineer. Chances are, he would attend an elite high school in Shanghai and then either enroll in China’s top universities, Tsinghua University or Peking University, or study abroad at Harvard University or the University of Oxford. “Perhaps in the future, there will be a minority 1 percent class of elite intellectuals who rise above the rest,” Jiang said.
Ultimately, Lee believes that the potential of education reform in China will come not from the public sector but from the private sector. “The education system is too stubborn to reinvent itself,” he said. “If you see education as a product, it has a life cycle of around 30 or 40 years. On the other hand, if you’re a technology company like Amazon, you have a short product cycle and can change your website 50 times a day.”
Although the government can articulate its priorities, Lee said that innovation will depend on parents’ spending power. In China, Lee says, some parents invest up to half of their monthly salary in their child’s education. Total annual spending by households on preschool and primary education reached 1.9 trillion yuan ($296 billion) in the 2016-2017 school year, comprising some 2.5 percent of China’s GDP in 2016, according to a survey conducted by the China Institute for Educational Finance Research, with household expenditure in first-tier cities such as Shanghai and Beijing being the highest.
Parents’ pocketbooks are to new education technologies “like eyeballs were to the mobile internet, or big data is to AI,” according to Lee. Due to the growth of the private sector, China has become a technological lab, spurring, in the last two decades, monolithic companies such as Tencent, Baidu, and Alibaba, not to mention a slew of AI companies such as SenseTime, iFlytek and Pony.ai. By extension, it could also become a pedagogical lab—but likely only affecting sectors of the population.
This starkly contrasts with the United States, where, overall, education technology does not attract the most investment. “The school systems have their own bureaucracies, parents don’t feel like paying tutors,” Lee said. For that reason, the use of AI education is more advanced in China when it comes to sophisticated adaptation to the needs of its user base, as algorithms are fed by immense quantities of data. Lee compares two AI education companies that his fund, Sinovation Ventures, has invested in: China-based VIPKid and U.S.-based WriteLab. VIPKid, which began as a low-tech video conference app that connected American teachers with Chinese students, began to accumulate and store more user interactions, which allowed it to make better matches to maximize user happiness, which in turn fueled a better understanding of AI. WriteLab, which began as an essay-correcting company, ended up selling for $20 million, while the former is now worth $5 billion, according to Lee.
Then, there is the greater challenge of teaching ethics: To what extent are schools responsible for teaching students how to ethically engage with new AI technologies, within the constraints of a system that encourages political obedience and policy conformity? In the SenseTime textbook, Ming Ming lives in an era in which driverless cars track optimal pathway, and automated assistants sort through an abundance of information, delivering only news perfectly curated and relevant to him. There is no discussion of the Trolley Problem, fake news, or the regulation of data privacy—or mention of the unseen curators of China’s news, its tens of thousands of censors.
If a nationally endorsed curriculum articulates a vision of individual data as a commodity regulated and owned by the state, that could become an accepted norm. And if popular textbooks are written by corporate start-ups such as SenseTime, a company that also happens to be one of the pioneering producers of facial recognition and surveillance technologies, its readers could be influenced by the vested interests of its authors.
This September, 35 universities in China will become among the first in the country to offer an undergraduate AI major. Educational institutions such as the University of Chinese Academy of Sciences, Shanghai Jiao Tong University, and Nanjing University have set up AI colleges. But overhauling the Chinese grade school system to teach creativity, collaboration, and critical thinking in both an equitable and ethical way is a monumental task. Perhaps the system can be reformed, but only in the private sector, and for students who can pay. SenseTime’s hypothetical Ming Ming, taught from his earliest classroom experiences to work as an elite AI engineer in cosmopolitan Shanghai, will most likely be ready to go. But whether China is able or willing to teach and prepare the rest of the nation’s citizenry through its public school system is entirely up to question.