The explosion of data and computing power has led to considerable advancements in A.I. over the past decade. Machine Learning algorithms rely on vast amounts of data to train models that are then used to guess various things, like who’s present in the photos that you clicked, or whether the voice your Alexa heard was yours or not.
Now let’s look at China.
First and foremost, China has the lead in internet user numbers by country. According to FT, almost 800M users sent digital greetings for Lunar New Year in Feb 2018. At the same time, Alibaba reported 580M mobile monthly active users. The U.S giants, Google and Facebook, are primarily cut off from this population and don’t have access to their data thanks to China’s great “internet firewall.”
Second, Chinese platforms are feature-rich, and hence, data-rich. Tencent’s WeChat spans various segments of consumer usage like photos, social networks, games, finances, etc. Alibaba is the e-commerce and retail giant in the country with data on searches, retail, payments, among others.
Third, compared to other countries, Chinese internet users are more willing to share their data. According to GfK Global Survey 2017 Report, China leads in percentage of people who responded that they are willing to share their personal data (health, financial, driving records, energy use, etc.) in exchange for benefits or rewards like lower costs or personalized service. Baidu’s CEO, Robin Li, came under fire earlier this year for claiming that people in China don’t much care about what’s done with their personal data.
Fourth, the Chinese Government is highly focused on developing AI. Last summer, China’s State Council issued an ambitious policy blueprint calling for the nation to become “the world’s primary AI innovation center” by 2030. It’s unknown how much China is going to spend over the next decade or so, but so far, the government of Tianjin, an eastern city near Beijing, said it planned to set up a $5 billion fund to support the A.I. industry and Beijing has committed $2 billion to an A.I. development park in the city. These are significant investments that overshadow other countries.
Fifth and finally, the Chinese Government is interested in collecting as much data on its citizens as possible. Across China, a network of 176 million surveillance cameras watch the country’s 1.3 billion citizens. To test the full capabilities of the system, the BBC sent a journalist to Guiyang, a city of 3.5 million people, to see if he could get lost in the crowd. The Surveillance cameras readily identified the journalist and police had him in custody within just seven minutes. Chinese AI startup SenseTime, a company that develops people surveillance software for the country’s law enforcement, recently finished a new funding round worth $600 million raising its total valuation to over $4.5 billion in May 2018.
If you put the above five factors together, you have the perfect brew for rapid A.I. advancements: a country with the largest number of internet-connected users who are more likely to share their data with feature-rich platforms in a walled garden served by a government focussed on developing A.I. and interested in collecting as much data on its citizens as possible.
Moreover, China’s progress over the past decade shows that this is actually working. In the Stanford Question Answering Dataset, an A.I. prediction benchmark that researchers are actively competing on right now, Alibaba and Microsoft tied for the top spot in January this year. Compare this to just eight years ago, when in Large Scale Visual Recognition Challenge 2010, another benchmark at the time, the top Chinese entrant was at the 11th place.
According to MIT Technology Review, Donald Trump’s chief technology advisor, Michael Kratsios, speaking at EmTech Conference at MIT in June last month said that that the government is looking for ways to open up federal data to A.I. researchers. The government has access to large amounts of data, and it’s possible that it could be used to train innovative algorithms to do new things. If done properly, this could be a boon to independent research groups.
In this discussion, It’s also important to note the importance of upcoming techniques that make our reliance on a large data set lesser. Transfer Learning is one such technique which focuses on storing knowledge gained while solving one problem and applying it to a different but related problem. For example, knowledge gained while learning to recognize cars could apply when trying to recognize trucks, so you don’t need to start afresh. Transfer Learning makes it possible to reuse learnings from one domain in another, thereby reducing the dependence on fresh training data. Transfer Learning is still under active research and development, but it will end up playing a significant role in maintaining U.S.’s lead in A.I. Deservedly, it’s also one of the most exciting research areas in the field of Machine Learning today.
Eric Schmidt, Executive Chairman of Google’s parent company Alphabet Inc., at A.I. & Global Security Summit 2017 said, “I’m assuming that [USA’s] lead in [AI] will continue over the next five years and that China will catch up extremely quickly. In five years, we’ll kind of be at the same level, possibly. It’s hard to see how China would have passed us in that period, although their rate of improvement is so impressively good.”
The United States is still ahead for now, but China is focused, organized, and gaining rapidly.