Over the past few years artificial intelligence has tried to teach, it seems, almost everything from complex calculations to a poker game. But Japanese researchers from NTT Resonant company focused their research in several different areas: they have developed artificial intelligence system Oshi-el now tries his hand at an online psychologist.
Like all currently existing systems of artificial intelligence, Oshi-el for a start, was forced to take a course. To do this, AI had to learn about 190 000 770 000 questions and answers in psychology. From the resulting database generated responses depending on the question asked, and giving the emotional response to color. Thus, the machine is able not just to give advice, but to Express sympathy and to offer some solutions to the problem. According to one of the founders of AI, Makoto, Nakatsuji,
“The vast majority of virtual companions, the so-called chat-bots, I can give you a short answer concerning certain factual matters. The answers to the questions concerning life and love in General, especially in Japan, can be very complex and contain a whole page of text. These responses usually contain a lot of the relationships between different categories such as family, school, work, etc., which makes replying very difficult task”.
Using as a basis the already existing answer, Oshi-el selects and combines the relevant proposals, selected from the database based on key words, and despite the fact that at the moment the answers are rather formulaic, the text is not devoid of meaning. The system continues to learn, and in the future, Japanese researchers plan to bring the AI up to a level where when replying to their “psychologist” will give more precise answers that will not be provided from a separate ready-made phrases, and is based on an “experience” of the entire system. Professor Di Wang from the University Carnegie Mellon actually has, after observing the work Oshi-el, in an interview with New Scientist, said:
“Of course, this system is not yet able to write an essay on abstract theme. For that she really need to understand the question. But we still do not even have a precise definition of what is meant by the term “understanding” in the context of artificial intelligence, so intelligence is able to capture only “the tip of the iceberg” of issues lurking in a given issue. And the most surprising is the fact that in this case, people don’t really care about the correctness of given them the answer they want to hear what they need, not what is right. And this system is doing”.