Everyone’s talking about AI: what’s next?
Updated 10:29, 28-Jun-2018

Suddenly everyone is talking about artificial intelligence; the victory of AlphaGo over a human Go champion made the world realize that scenarios from science fiction films are close to being replicated in real life. Under this context, 35-year-old computer scientist and tech startup entrepreneur Xu Li shared his professional outlook of AI’s transformation through the data age and explained to us why the darling of the scientific world is still at the toddler stage.

‍Stepping into the reception area of SenseTime’s Beijing office in TusPark (scientific research park of Tsinghua University), your face is immediately scanned by a camera - physical characteristics are analyzed and matched to information stored in the system, before being displayed on a screen. Staff from this computer vision startup call the device their “clock-in machine.’
The machines can score people’s looks, estimate their age, feature face swapping applications, and contain gesture recognition sensors… the range of remarkable technologies based on visual data are not intended for entertainment only - they can be utilized in multiple business areas, including security supervision, financial sector and mobile Internet. 
Stepping into the reception area of SenseTime’s Beijing office in TusPark (scientific research park of Tsinghua University), your face is immediately scanned by a camera Input Words.

Stepping into the reception area of SenseTime’s Beijing office in TusPark (scientific research park of Tsinghua University), your face is immediately scanned by a camera Input Words.

Deep Learning

This is deep learning content.
      "We are using artificial intelligence’s deep-learning technology to dig out value from visual data to help our clients."
 -  Xu Li, CEO of SenseTime.
From its cooperation with companies and organizations - including China Mobile and UnionPay - over 300 million users are active on SenseTime’s visual image processing technology. SenseTime has just obtained high exposure in the media about its collaboration with StarVC, a venture capital firm run by several top Chinese celebrities. Before StarVC, IDG Capital Partners invested over 10 million USD in SenseTime in November 2014.  
In a bid to concentrate on original technology research— to build the brain, in CEO Xu Li’s words - he set up technology-oriented teams in October 2014. Doctors from top education institutes across the world joined senior staff from giants corporations like Google, Microsoft and Lenovo. Multiple office locations were chosen to research a variety of functions: Beijing for its engineering talents, cultivated by the specialized Chinese education system; Hong Kong for its advanced roles on AI R&D and culture fusion; Shenzhen for its mature hardware manufacturing industry; and Kyoto as a key location to explore the Asian market. “It is pretty convenient to do business in Japan,” Xu said, adding that he appreciates the country’s respect for copyright.
It may seem risky for SenseTime to enlarge its territory this fast. But that concern is not in Xu’s mind. “Right now it’s the opportune time for AI’s development, both in academic and industry practice, we need to drum up good quality teams to stay on the edge of the field.” 
There are good reasons for Xu’s confidence. He started researching computer vision as an intern on visual and text identification in Motorola’s research lab in Shanghai in 2004, and received his doctorate degree in computer science from the Chinese University of Hong Kong. “One could directly see the results from the identification research, although at that time it was just a period of human-oriented intelligence,” Xu recalled. He explained that although the academic world had been studying AI for many years, it was not ready to practically apply it until 2014, when the method of “deep learning” pushed AI to mass popularization. Both data and algorithm systems are crucial to data-oriented artificial intelligence - the development of AI has now arrived at “the beginning of steam engine age.”
Talking about human’s efforts on AI research, Andrew Ng, a key figure in machine learning and now chief scientist at Baidu Research, applies the concept of rocket and spacecraft relationship to machine learning during an interview with Baidu. “Countless data is being generated every day, but only in recent years did we get the ability to build big engines (mainframe computers) to absorb (analyze) the fuel (data).” 
AI has been written into China’s 13th Five-Year Plan in 2016, which implies the technology will be applied more frequently as the country seeks to take advantage of its vast technical talent pool and the data advantages of a large population. According to Beijing research company iResearch, there are nearly a hundred startups in China entering the AI field and 65 of them have gained venture capital. The market is indeed soaring.
Critics claim that unlike big companies who hold fundamental resources or produce terminals themselves, most of the startup companies in AI field are technology-oriented and are still in the stage of exploring proper business model.
Xu counters that his company enjoys decent cash flow with a business model that provide services to terminals and cloud. “We are using a business to business to customer model, charging by the quantity of terminals and volumes of service requests by customers.”
“A passing rate over 90% could meet the common need in the facial recognition market, and our faulty rate is only one in a million,” Yang Fan, co-founder and head of SenseTime’s Beijing office said.
To Startups

To Startups


We missed Doctor Xu in SenseTime’s Beijing office when we first visited the computer vision startup. Then at a scheduled interview time in Shenzhen after his business trip to the US, he lost his voice. Finally we caught up with him at an enterprise service summit in Beijing’s 798 art district.

Deep Learning

Has been carefully studied by scientists, deep learning is a branch of machine learning, a method to deal with big data based on a set of algorithms. It could compose a neuro network system in imitation of the human brain to analyze data in the goal of achieving artificial intelligence. This is deep learning content.
Interview venue credited to:Artofark