I write columns on news related to bots, specially in the categories of Artificial Intelligence, bot startup, bot funding.I am also interested in recent developments in the fields of data science, machine learning and natural language processing ...
I write columns on news related to bots, specially in the categories of Artificial Intelligence, bot startup, bot funding.I am also interested in recent developments in the fields of data science, machine learning and natural language processing
Kai-Fu Lee is very bullish about the future of AI in China. He started his lunch keynote at MIT's AI and the Future of Work event by predicting that self-driving cars will become a mass phenomenon in the U.S. in 15 to 20 years. But in China it will be "more like 10 years."
That may come as a surprise to many American observers of and participants in the rush to capitalize on recent progress in artificial intelligence. Especially when such predictions come from someone with the stature of Kai-Fu Lee and his deep familiarity with the state-of-AI in both North America and China. Lee completed his PhD in artificial intelligence (speech recognition) at Carnegie-Mellon University in 1988, worked at Apple, SGI, Microsoft and Google and, in 2009, established Beijing-based Sinovation Ventures, an early stage venture capital firm.
Lee concluded his talk at MIT by stating that "in the age of AI, a US-China duopoly is not just inevitable, it has already arrived." That was a polite conclusion to a presentation that marshalled plenty of evidence pointing to China moving beyond catching-up to becoming number one. Here's my summary of Lee's observations about the reasons for China's ascendancy:
Lots of Brains
There have been only two Chinese researchers out of 289 elected as Fellows of the Association for the Advancement of Artificial Intelligence (AAAI), "but the number of Chinese AI researchers is growing unbelievably fast," said Lee.
Chinese researchers' contributions to the best 100 AI journals/conferences rose as a percent of total papers from 23.2% in 2006 to 42.8% in 2015 and as a percent of cited papers from 25.5% to 55.8%.
Face++, a Chinese face recognition startup (and one of Sinovation's investments), recently won first place in 3 computer vision challenges, ahead of teams from Microsoft, Facebook, Google, and CMU.
The AAAI postponed its 2017 annual meeting by a week when it found out the planned date coincided with the Chinese New Year. A nearly equal number of accepted papers came from researchers based in China and the U.S.
Lots of Data
"There is no data like more data," Lee quoted AI pioneer Fred Jelinek. Most of the AI algorithms are open source and well-known, he said, you just need a few smart people to tweak them. Companies such as Facebook and Google owe their success to a virtuous cycle: More data leads to a product that is better trained with AI which leads to more users and making more money, enabling the hiring of more scientists and the acquisition of more machines, processing and mining even more data.
As Lee pointed out, the data gap between the US and China is "dramatically larger" than the actual gap between the respective populations or the number of active mobile users. Chinese use their phones to pay for goods 50 times more often than Americans, he says, and orders for food delivery are 10 times greater than in the US. All of this is data available for training AI models. "China is leapfrogging the way mobile is and should be used," says Lee. US used to have the best payment infrastructure in the worlds credit cards. But China got the chance to leapfrog, says Lee, with mobile payments: frictionless, peer to peer, instantaneous. This mobile infrastructure and all the applications built on top of it generate lots of data feeding into the AI virtuous cycle.
For MIT's Distinguished Lecture series, wrote Robert Buderi and George T. Huang in their 2006 book Guanxi: Microsoft, China, and Bill Gates' plan to win the road ahead, Lee was asked to leave a sentence with the students that they could put on a plaque for posterity. "The first thing that came to my mind [Lee recalled] was What really matters is not innovation, but useful innovation.'"
This is where China is now. The days of being just a copycat are over. But copying taught entrepreneurs how to build products and they became good product managers, says Lee. Then they quickly moved to the next stage "when they were inspired by American innovation" and developed new product features. As a result, "Weibo is a better product than Twitter, same for Taobao and eBay, WeChat and Facebook Messenger. Better features, more robust business model." Today, Chinese companies are coming up with innovative products not seen in America such as customized news or distance learning using "underpaid American teachers" to teach English. We are now entering the age of copying from China, says Lee.
Scale and scope made US companies flourish in the vast American market in the 19th century and then successfully compete globally in the 20th century, wrote historian Alfred Chandler. Japanese companies had no experience of scale and scope in their relatively small domestic market but honed their competitive skills fighting each other at home before they went to conquer world markets in the second half of the 20th century. Chinese companies experience both much larger scale than anything seen before in the US and no holds barred domestic competition.
China provides a "conducive environment for fast launch and fast iterations," says Lee. Compared to the US, there is less focus on data privacy and "less expectation of consensus on moral issues." There is more focus on execution and a dedication to a singular objective and the vision for the company. That vision, more often than not, includes global ambitions.
Government support (of the right kind)
The aforementioned Face++ recently raised $460 million in a C-round of financing led by China's state-owned Capital Venture Investment Fund. Matching funds from the Chinese government increased dramatically last year to $353 billion, up from $146 billion in 2015 and $0.129 billion in 2014.
"The government becomes an LP [Limited Partner] in a VC fund with a willingness to take a lower return than other LPs," says Lee. "That's an encouragement to top VCs to take government money and spend less time raising funds and more time building great companies. It's inspired by Israeli and Singaporean practices which are very smart practices," he adds.
Most important, government funds are guided by clear government policies. The Chinese government is pro-tech, pro-experimentation, and pro-speed, according to Lee, "as opposed to using policies to forbid things." Last July, China's state council announced that by 2020 China will catch up in AI technologies and applications and by 2030 it will become "the world's primary AI innovation center." Lee compared this "very straightforward" approach with the Obama administration's 2016 report on AI which was "very rational and reasoned about long-term resources, regulation, thinking about ethics and things like that."
Quoting the October 2017 speech by Xi Jinping to the 19th Congress which set the goal of promoting "further integration of the internet, big data, and artificial intelligence with the real-world economy," Lee says "He gets what we are talking about in terms of AI and combining the virtual world with the real world."
For those in the audience in the habit of discounting statements from government officials, Lee noted that in 2010, China said it will be the world leader in high-speed rail. Six years later, 60% of the world's high-speed trains were deployed in China. In 2015, Premier Li Keqiang announced a program called "mass entrepreneurship and innovation movement," making everybody in China an entrepreneur. There are now 8,000 Chinese incubators/accelerators (up from 1,700 in 2014), said Lee, and 89,000 startups have already graduated from these incubators.
An adventurous spirit of risk-taking, experimentation, and audacity characterizes Chinese consumers, many companies, even government agencies. This was evident in Lee's talk especially when he ventured to predict the future along the lines of his "four waves" or stages in the digital transformation of the economy.
In the first wave, companies succeed by smartly using online data and Lee predicts that "Chinese internet companies will leapfrog American ones." Not only because of their lead in mobile and payments, but also because they "are less restricted by anti-trust laws and are tenacious in terms of territory expansion. Every one of them [i.e., Baidu, Alibaba, Tencent] is going to have a bank, and insurance company, and so on."
In the second wave, companies succeed by developing and selling software that mines commercial data. Because large, traditional Chinese companies, unlike American enterprises, "never bothered to data warehouse their important content," China lags behind the US in this area. "But the government is pushing Chinese companies forward and they may have a chance to catch up in the future," Lee hedged his bets.
In the third wave, digitizing the real world, "China will leapfrog the US," because of the quantity of data, less concern by consumers about privacy, and cheaper hardware. A product similar to the Amazon Echo is selling for $15 in China, says Lee.
The fourth wave is about full automation, as in autonomous vehicles and robots. In Lee's estimation, the US is way ahead in technology, "meaning 2 years which is really a long time." But Chinese companies, with government support, are moving faster, "so we will see who ends up being ahead I would say it's fifty fifty at this point."
But Lee was a bit more precise and confident in Chinese superiority in another part of the presentation when he discussed retail automation or autonomous stores. Amazon acquiring Whole Food? "We think this is a rather primitive activity," he asserted. While acknowledging that Amazon Go is the most technologically advanced example, "autonomous stores are coming up everywhere" in China, said Lee. This is part of a larger trend, Online-Merges-with-Offline (OMO), the merging of the digital world with the physical world, which Lee says will impact also education and many other areas. If you visit China in another year or so, he promised, half of your shopping will be done in autonomous or semi-autonomous stores. "This is the speed in which Chinese companies implement things," Lee says.
Free-wheeling, competitive, bold. "That sprit will launch China forward to become a very strong AI power," concluded Lee.
The world's most dominant spirit of enterprise has relocated and been reimagined before, transforming from the Merchant Adventurers of 16th century England and their wave-ruling successors to the "American System" which first meant protectionism (and IP theft from the British), and later became the buzzword for the mass production of industrial goods with interchangeable parts and, after 1945, the globe-trotting ideology "of free enterprise and limited government" which reached its apex with the "end of history" in 1989.
Except that history did not end. Instead, the future is being invented elsewhere. The same motivating force that has animated economic growth in the West is now producing economic systems with completely new features in China and India. In the preface to the 2010 Chinese translation of Nationalism: Five Roads to Modernity, Liah Greenfeld wrote: "Nationalism has finally taken root within the general populations of China and India, becoming the decisive collective motivation of these two giant peoples. This has made them economically and politically competitive."
Lee featured prominently in the story Buderi and Huang told about establishing Microsoft's research lab in China. "In talking to Lee," they wrote, "his sense of familial duty and respect for his mother and father are inescapable." Lee's father, Li Tien-Min, a legislator in the governing Kuomintang administration when Mao came to power in 1949, had to flee to Taiwan. "My father always felt China had not realized its potential," Lee told Buderi and Huang.
Today, China is realizing its potential to become a leader in many fields, including artificial intelligence.