Satya Nadella, the current chief executive officer (CEO) of Microsoft, has recently discussed the prospect of the decline of the App era and the rise of the Robot era. He said that in the near future, if we were to give a command to a robot through our daily language, it would understand and respond properly by considering the context of our words. In fact, concierge services that use chatbots, which are robots that conduct conversations with humans via messenger, have been expanding as Artificial Intelligence (AI) has developed dramatically in the past few years. The Sungkyun Times (SKT) introduces the concept of a mobile concierge service, the global and Korean trends, problems, efforts to overcome the limits, and the prospect of mobile concierge.
What is Mobile Concierge?
According to the Cambridge Dictionary, a concierge is someone employed in a hotel to help guests arrange things, such as theater tickets and visits to restaurants. The word originates from a French word meaning a servant aiding a lord in the Middle Ages and it refers to a person that arranges customized services to clients in fields such as tourism, transport, restaurants, and shopping. Meanwhile, “mobile concierge” is a term that has emerged recently. It is a personal assistant that offers private assistance to people via mobile messenger. The chatbot behaves in the same manner as a human concierge. Chatbot is a compound word derived from “Chatting” and “Robot,” and indicates a system that replies to users’ requests according to certain response rules. Apple’s Siri is a notable example of a chatbot. The operation procedure of a mobile concierge service consists of two steps: we speak to the chatbot and then the conversational robot responds to the requests that we have made. There are several distinct methods for calling on the chatbot in the first phase. For example, when we add “@” and the name of a company on the messenger platform, the chatbot of the corporation appears. Kakao Talk, which is planning to launch the service during the first quarter of this year, is going to add a chatbot to the list of Kakao Plus Friends and users will be able to call on them by adding them to the friends list.
There are two main causes of the rise of mobile concierge. The first arises from the consumer side, which has arisen due to the difficulty of making choices due to the overflow of information. It is exhausting to determine what to purchase online due to the existence of numerous products and the ceaseless appearances of new items or similar goods. The large variety of choices has rendered some people tired and has led to an increased demand for one-to-one referral services. The second, on the supplier side, has arisen due to the advancement of AI. In Korea, in the early 2000s, there was a service that automatically replied to people through MSN Messenger called “Simsimi”. People, however, could not utilize Simsimi as a marketing channel or a platform of commerce because it showed limitations in accurately grasping the context of conversations. On the other hand, as a consequence of the rapid development of AI and machine learning, the foundation for mobile concierge services has been arranged.
The existing “human” concierge service is provided offline primarily for VIP customers staying at five-star hotels or department stores. Employees who have obtained special training help the clients by providing assistance while shopping or solving complaints. E-commerce, however, has a limit in providing personal care services. Because it is based on online, consumers cannot experience the kindness of clerks at offline stores and it usually takes a long time to resolve complaints. The emerging chatbot technology has opened the possibility of the concierge service to be provided to the public online.
Global and Korean Trends
Platform The basic platform of mobile concierge services is a mobile messenger. We can simply talk with chatbots as if we talk with friends through messenger and obtain recommendations on products by our assistants. It is efficient in that we do not have to install new applications for numerous purposes such as when shopping. For example, there is no need to install an airline application when purchasing flight tickets because we can make flight reservations by discussing the details with a chatbot.
|Facebook Meseenger unveiled Send/Receive API to external chatbot developer. / appfutura.com|
Several global IT corporations are currently leading the mobile concierge industry and they are classified depending on whether they possess their own chatbot technologies or not. The former only provides the messenger platform to external developers of chatbots, while not actually developing the technology itself. For example, in 2016, Facebook took over “Wit.ai”, a startup company that had Natural Language Processing (NLP) technology, and Facebook Messenger unveiled Send and Receive API to developers by utilizing the technology. API is an abbreviation for Application Programming Interface, which is the software used between operating systems and application programs. In this situation, the messenger platform is the operating system while the chatbot program is the application program. By communicating with the chatbots, people can easily book hotel reservations, look up the weather, or purchase products through the message service. Currently there are about 20,000 chatbots registered in the platform and at least a billion messages are sent to corporations by the chatbots. The global companies of the second type not only have their own messenger platform, but also develop chatbot technology in order to stay competitive. The chatbots they create are interlocked with other companies. For example, Google released a messenger “Allo” that has the mobile concierge “the Google Assistant” inside. When you add “@google” into you and your friend’s chat window and send requests, the assistant appears and replies promptly. It suggests restaurants nearby or shows Youtube videos right there in the chat window. In addition, you can speak one-to-one with the Google Assistant after adding it to your chat friend, too. After Allo was officially released in September of 2016, more than five million people downloaded the messenger application in the first week.
|Example of Mobile Concierge Service from Google Allo|
Compared to global IT companies, the mobile concierge technology in Korean companies is still in its fledging stage. The industry, however, is anticipating that it will take a long time for the global IT companies like Google to sweep the Korean market because they do not have a sufficient amount of Korean language data to implement the service. Korean companies are trying to take advantage of this and dominate the domestic market before foreign companies get the chance. Meanwhile, Korean enterprises engaged in the industry are also divided into two groups. The first type acquires external chatbot technology like Facebook. A good example is Kakao Talk, a Korean messenger application created by Kakao. 99.2% of message senders are using the application as of January this year and Kakao provides the platform to external chatbot developers. Kakao launched a web-based bank named Kakao Bank this February also, and the bank is planning to combine its chatbot and Kakao Talk in order to offer a mobile concierge service to its clients. The second type develops internal chatbot technology, while having its own messenger like Google. Naver’s chatbot called AMICA, which was released in 2016, is a notable example. The company is aiming to build an alliance named “AMICA Open Alliance” with internal chatbot developers. People joining the coalition will be provided with numerous mobile concierge services that utilize Naver’s Natural Language Understanding (NLU) technology. Naver Labs, the laboratory of Naver, is looking forward to open the Korean chatbot market as it has its roots in Korea.
Limits and Efforts to Overcome Issues
|Racist Comment by Tay / nextshark.com|
The mobile concierge service has several limits at present. Firstly, there can be a communication problem. Although chatbots can currently recognize dictionary definitions or standard language, they cannot perfectly analyze the languages we use in our daily lives. They can easily give information regarding simple things like our daily schedules, but they cannot effectively deal with complicated situations like call centers. Secondly, there is a possibility of learning data quality deterioration. As an example, Tay, an AI chatbot launched by Microsoft in 2016, uttered offensive and provocative comments due to the learning d a t a t h a t s o m e ultraconservative people deliberately inputted. They abused the fact that Tay adopts automatic machine learning, inducing the chatbot to make racist or sexist remarks. As a result, Microsoft had to cease the service after just 16 hours. Thirdly, there is a personal information security issue. Mobile concierge services offer one-onone services on the basis of clients’ personal information.
■Efforts to Overcome the Limits
The attempt to overcome the first limit is to develop AI. The core technology of mobile concierge is the ability of AI based chatbots analyzing the sentences in a chat window and conducting necessary work. Therefore, developers are engaged in developing NLU technology. The endeavor to solve the second limit is to reinforce the ethical aspect of chatbots, creating a mobile concierge service that can make moral judgments. Currently, AI behavior could be completely different according to the learning data, even though it starts from the same algorithm. People are trying to grapple with ethical issues by inputting moral values to the learning data, and trying to teach AI such imperative values. For example, Tufts University in the United States is researching how to make a chatbot that can say “no” in some circumstances after making its own judgments. The effort to solve the third limit is protecting private information via encryption. Companies like Facebook are adding End-to-end encryption (E2EE) in messenger platforms. E2EE is a security system that keeps messages secluded through encryption. As a result, even platform servers cannot read conversations held between clients and chatbots.
Future of Mobile Concierge
As the mobile concierge industry is growing, many experts are convinced of its bright future. First, the market is expected to expand rapidly, reaching an estimated value of eight billion dollars by 2024, which is nine times bigger than the market scale of the end of last year. The service would be widely used in various fields like the financial market, although it is presently commerce oriented. Moreover, chatbot technology will advance drastically in the near future. Until now, chatbots could only reply properly to simple questions, but the technology has made progress into the appearance of chatbots that can break down sentences grammatically and analyze complicated requests. For example, a domestic startup company, MoneyBrain, which released a mobile concierge service, inspects users’ natural language in four steps: breaking down a sentence in substantial parts, separating the chunks into smaller ones, analyzing the meanings of small parts, and finally understanding the sentence as a whole and the demand the client has made. Likewise, more advancements and improvements are expected. Additionally, the development of AI and data processing technology will bring out a voice recognition chatbot. In fact, at present, the primary platform of the mobile concierge service is Messenger because we use text-based chatbots, but someday when calling on our mobile concierge, we should be able to just use our voices.
A personal assistant in our hands is definitely going to bring us a more comfortable and convenient life. We should, however, keep reminding ourselves that technological development does not only have a bright side. Being aware of the ethical issues of chatbots and the security of private information is crucial because we have to prevent and minimize the potential problems.
<저작권자 © THE SUNGKYUN TIMES, 무단 전재 및 재배포 금지>