Zhu Pinpin: Making Robots with “Brains”

By XIE FEIJUN



Zhu Pinpin: Making Robots with “Brains”

By XIE FEIJUN

Data analyses show that Xiaoi Robot users now surpass 50 million.

IN July， more than 1，700 top AI experts gathered at the 25th International Joint Conference on Artificial Intelligence （IJACI-16）in New York. A grand meeting for both AI researchers and practitioners，IJCAI is also one of the major global academic conferences in this field. One Chinese scientist stood out among the speakers. His name is Zhu Pinpin，founder and CEO of Xiaoi Robot. In his speech， Zhu shared his insight into the “Bot Economy” and development of the Xiaoi Robot.

A customer service bot for a bank.

Transformation of Xiaoi

A “chatbot” named Xiaoi made the first public appearance of one of Zhu’s products on MSN in 2004. Before this，only human beings were able to chat on MSN.The introduction of Xiaoi meant that a user could chat with a bot. Different from human beings， Xiaoi gave answers to every question asked. “Some people were intrigued to chat with Xiaoi， so we accumulated a pool of users very quickly，” said Zhu. Back then，users meant money， and you didn’t need a closed cycle to make profit on the Internet.

It was just at that time that Zhu， as China’s first pioneer of commercializingchatbots， began to ask himself and his team the question， how could this type of chatting generate commercial value？

This question was followed up with a series of explorations， during which the company continued developing Xiaoi applications，adding weather reports and stock checking to its function list. In 2006， specialized field knowledge was installed into Xiaoi to make it a virtual customer aide. The advantages of this kind of virtual aid are obvious: communication and labor costs associated with customer service staff who answer phone calls was rising in the telecommunications， banking and airline industries； more important， this kind of work is so boring that staff turnover rates are very high. Automation is Zhu’s solution to this problem. He envisioned that 80 percent of the work， like inquiries and similar monotony， could be done by robots， leaving customization to real operators.

Despite this grand vision， it has been hard to find customers. Since domestic companies introduced telephone service systems as late as 2000， they were not eager to reduce costs. “Innovation is much more difficult than one could imagine. During the whole process，there may be more onlookers than participants or supporters.” Jiangsu Mobile became Xiaoi’s very first corporate customer， and a company to whom Zhu is still grateful. He and this trial user see eye to eye in many aspects. They first applied bots to its website， and then to text messages. Fortunately， feedback data proved encouraging， with labor costs reduced by RMB 2 million and new revenue from text message bots increasing by RMB 2 million.

Xiaoi became an overnight hit. In a short period of time， many branches of China Mobile， telecommunications companies， banks， and airlines contacted Xiaoi with the aim of using bots for monotonous labor. Today， several siblings of Xiaoi are operating in our daily life. For example， Xiaozhao， a customized bot for China Merchants Bank， offers 24-hour service in all kinds of business， including smart customer service，cellphone applications and offices. Data show that the application of bots saved China Construction Bank 6，000 seats every year. Statistics from the Bank of Communications indicate that in December 2014 alone， RMB 10 million was saved thanks to bots.

During an interview it was revealed that Xiaoi Robot has currently provided intelligent bot services to over 500 million people worldwide， making it one of the top AI enterprises and a major engine driving the “Bot Economy.” Furthermore， it occupies at least 70 percent of the domestic customer service robot market share.

Zhu Pinpin exchanges views with Tom Austin， world leading “smart machine”technologies analyst.

OVERNIGHT HIT

Xiaoi became an overnight hit. In a short period of time，many companies contacted Xiaoi with the aim of using bots for monotonous labor.

Earth-shaking Development

At the New York International Joint Conference on Artificial Intelligence，leading experts from around the globe reached a consensus. The third eruption of AI currently taking place is not simply a repetition of the previous two，but should bring about earth-shaking changes. Zhu says he realized this long ago and that the third eruption differs from the previous two in three aspects. First， technological

breakthroughs have been made and the application of cloud computing and big data has vastly enhanced computing capabilities and offered better data. Second， in terms of industrial applications， many projects are underway， generating much commercial value. Third， people in the industry now have a more rational attitude than before.

Rational attitudes play a vital role in the development of AI. The term AI was first proposed in the 1956 Dartmouth conference. The computing capabilities of computers are indisputably much stronger than those of human beings，but there is still no clear way that computers could simulate human thinking to create similar intelligence.

In Zhu’s opinion， the human thought process is more than logic. It depends on people’s common sense and experience. Around 1970， AI encountered its first low tide when it fell short of expectations. Knowledge was used to drive its development， and AI language specifically designed for logical thinking was invented. However， before 2000，another failure struck and the second low tide followed for the same reason. AI development still could not meet our high expectations and no products were commercialized. This led to a discouragement as regards investment.

Recently， AI has entered a new round of development following the success of AlphaGo. This has greatly boosted human confidence. “Machines are able to check and use the vast amount of datawhich we have accumulated， but their IQ is much lower than ours. This means that at most they can only become champions in certain specific field.”

Having realized that AI and human thinking are based on different logic， we will not let them compete with us in wisdom， nor be obsessed with “the Turing Test，” and will not shackle them to our standards. Instead， we will focus on one problem， solve it and then target the next. This is the right direction. The prospects for AI will thus be even brighter.

Next Stop

In the Xiaoi Robot exhibition space，Xiaoi is widely used in customer services，smart homes， smart driving and many other fields. This reporter also saw the same type of robot greet Premier Li Keqiang at the Big Data Industry Summit.

This intelligent banking robot is the first in the world. Since she commands a substantial presence， she is like a lobby manager， socially skilled and able to communicate verbally with customers， and assist in queuing and guiding. What’s more， by means of “ID card plus facial recognition，” she can check an account balance within a second.

Three important factors make a robot tick: moving， feeling and thinking. There are many intelligence problems to be solved in terms of moving， for example，self-balancing in kinematics. As regards feeling， voice recognition， face recognition and image recognition， all belong to emotional intelligence. Thinking is most important， and could be regarded as cognition intelligence， also the major research focus of Xiaoi.

“What we need to do is to add ‘brains’to the robot hardware. To explain further， the core of thinking is knowledge，which supports thinking and which constitutes the biggest difference between humankind and animals. Humankind is capable of creating new knowledge and passing it down. This means that future generations will not need to obtain new knowledge through practice but education. Naturally， through constant expansion of knowledge， our thinking ability grows stronger and we become smarter.”

Technically speaking， is it possible for robots to undertake all repetitive work in daily life？ Unfortunately it is not. It is possible for one robotic arm to conduct all the work on an assembly line， but not to serve at a restaurant. Although it seems simple， spacial movement is involved， which makes it no simpler than operating precise instruments.

Internationally， robots are divided into virtual and real ones. The core technology of Xiaoi is natural language understanding， which is motivated by a knowledge repository. Worth mentioning is that all technologies in this respect are self-developed by Xiaoi.

The domestic and international strategies for AI are the same. Compared to top players in the world， Chinese companies are still at primary stage in core technologies. But in some specialized fields some Chinese enterprises perform quite extraordinarily， and have caught up with the world leaders. Xiaoi， which is focused on natural language processing， and Deep Glint， which is dedicated to computer vision， are two such examples.

The smartphone weather bot app.

Of course， the development road for Xiaoi is not always smooth. Zhu says there are ups and downs all the way and new challenges are sure to arise. But Zhu， who started the enterprise even before college graduation， has persevered. The year 2009 was the hardest period. He paid his employees through taking out a private loan.

Also in that year， Xiaoi was undergoing a tough transition from dealing with individual customers to corporate customers. Zhu himself transformed from being buried in developing technologies to planning overall corporate development and dealing with customers.

“Investors thought tinkering with company service was too slow， which led to a broken financial chain for some time.” During its toughest period， only 50 employees stayed with Xiaoi， but today the number has increased to 500.

Zhu said， “Looking back， we can see that innovation is a systematic project. It is not about just an idea or technologies， but also customer service and business operation.” The age of the people responsible for the products is vital to whether the innovation will be effective. In Zhu’s opinion， it’s reasonable that many people in charge of products in IT companies be 20 to 30 years old， because innovation really relies heavily on younger people.

Against the backdrop of economic globalization， multi-lingual semantic recognition and understanding technologies are becoming an inevitable trend. As a widely applied robot， Xiaoi has already gained an advantage in Chinese semantic understanding. Zhu believes that his team will actively develop multi-lingual semantic understanding technologies in the future and lay a foundation for his products to enter global market.

XIE FEIJUN is a reporter with Liberation Daily.