Robot customer service is coming: can not pass the "Turing test", but or an antidote to "Baumol's disease"?

（IC photo/图）

Unicom customer service calls the user, a kind female voice. The user suspected that she was a robot, and later asked her whether she was a robot, and the answer of the customer service was exactly the same every time: "How come, it must be the employee who called you, we Unicom are all trained in a unified and strict way..." Repeatedly use the same words to answer the customer's video, even the tone words are exactly the same, which has triggered a lot of ridicule on the Internet, and everyone unanimously believes that this customer service is a robot, and ridicules that this does not recognize that it is a robot robot.

In fact, answering the phone initially thought that it was a real person calling, but it turned out to be a robot, which is not a rare thing.

The reason for adopting AI as a customer service is simple.

In one hall, in one cubicle after another, many people wore headphones and answered one phone after another. The needs of these incoming customer calls are, in many cases, similar and repetitive. In the vast majority of cases, answering customer calls is a labor-intensive task. So, all along, the work has shifted to areas where labor prices are lower. For example, many of the company's customer service centers are outsourced to the Midwest. As early as many years ago, telephone customers in Europe and the United States moved across the ocean to India, because India has a relatively high number of English speakers and lower wages. Although it will increase the cost of transoceanic calls, it can greatly reduce labor costs.

Obviously, outsourcing to machines is a better approach than outsourcing to people. In fact, outsourcing to machines has long been available. As long as technological advances provide this possibility, people will think of using new technologies to improve efficiency and reduce costs.

The earliest calls were all manual, with manual customer service forwarded to various parts. Later, with the popularization of audio dialing technology, after different numbers corresponded to different audios, people came up with new ways to save manpower. First, a voice is played, and then distributed to different customer service according to the caller's button, and even a voice answer question is played directly. Now with the emergence of new technologies such as artificial intelligence and deep learning, semantic recognition has been greatly improved, and the accuracy is already very high.

After the adoption of the artificial intelligence customer service system, no one needs to listen to the voice and press the button to select, but directly take the form of dialogue, intelligent speaking, intelligent listening. The most immediate benefit of this is that the cost is greatly reduced. The marginal cost of AI is very low. If it is done by real people, the customer service service of 1,000 seats costs 100 times that of 100 seats, and it is the same every month. But for computers, when the whole system is developed, 1,000 customer service is more than 100 customer service, only a few more servers at the beginning, and only needs to have more electricity every month. In addition to reducing costs, there is also an advantage after adopting artificial intelligence, that is, people mistakenly think that they are real people, and often people are more patient with real people, which increases marketing opportunities.

In a way, artificial intelligence customer service is a realistic example of a Chinese room thought experiment.

In 1950, Alain Turing proposed the famous "Turing Test" as a criterion for judging artificial intelligence. He believes that if a computer can make it impossible to tell whether it is a computer in a question and answer, then the machine has intelligence. This test has also become a classic criterion for testing artificial intelligence. However, in 1980, John Searle, a philosophy professor at the University of California, Berkeley, proposed Chinese the "Chinese Room" thought experiment, which was used to reject the validity of the Turing test. In this thought experiment, an English-speaking man is in a room that is all enclosed except for a small window on the door. He carried with him a set of books that recorded enough rules for translation between Chinese and English. There was also enough paper and pencils in the room. Pieces of paper with Chinese written on them are sent into the room through small windows, and the people in the room can use his book to translate the words and reply with Chinese. Although he wouldn't Chinese at all, he could make anyone outside the room think he would say Chinese.

What John Searle wants to show through this thought experiment is that a person who does not understand Chinese and cannot really understand Chinese can make people think that he understands Chinese by rules alone, then even if the computer passes the Turing test and makes people think that it is a person, the computer is not really "understanding", but simply acts according to the rules.

Thought experiments are complex, the debate over artificial intelligence will continue, and there is new hope for quantum computers along the way. But for the current industry, whether to understand is not important, the important thing is that the technology can meet the needs of the application, as long as it can be "like a real person", it will definitely bring a breakthrough in the application, but artificial intelligence is still a long way from passing the Turing test. Now whether it is Microsoft Xiaoice, Xiaodu or SIRI, they are extremely mentally retarded and easy to identify, but with the development of technology, the future space is very large, and may even become a new entrance to the Internet.

When consumers have needs, they give instructions to their artificial intelligence assistants: Duoduo, Xiaodu, SIRI, help me book a ticket, ask for the cheapest, or help me find a restaurant or gas station. This may be the next generation of Internet portals. Although everyone is now working towards this aspect, it still needs to be developed because semantics and even artificial intelligence technology still cannot achieve basic communication.

At the Wuzhen Summit of the 2021 World Internet Conference, Vice Premier Liu He pointed out that the current Internet development has jumped to a new stage of full penetration and cross-border integration, and digital technology has deeply transformed production functions and continuously created new formats, bringing new development opportunities to all countries. In this process, it is necessary to overcome the "Baumol disease" and the "digital divide" and achieve inclusive economic growth.

In 1967, the American economist Baumol divided economic activity into two main sectors: one is the "progressive sector" with innovation, scale effect, large impact of technological progress, rapid per capita output growth, and rapid productivity growth, generally referring to manufacturing. The other is the "stagnant sector" with little innovation, weak technological impact, lack of scale effects, and slow productivity growth, which generally refers to the service industry, including education, municipal services, performing arts, catering, entertainment and leisure. With the development of the economy, labor continues to flow into the stagnant sector. Specifically, an auto factory can use automated production lines to produce cars, but the wages of manual telephone customer service are getting higher and higher. At the national level, the rate of economic growth across the country will slow. This is the so-called Baumol cost disease and growth disease, referred to as Baumol disease. Baumol's disease describes the problem of an overused share of the service sector.

However, the emergence of artificial intelligence provides a new perspective for solving Baumol's disease. To some extent, ai-powered customer service is a typical example of a new technology solving Baumol's disease.

(This article is only the author's personal opinion and does not represent the position of this newspaper)

Liu Yuanju