What is the "possibility and danger possessed by artificial intelligence" the artificial intelligence guru speaks?

Artificial intelligence using machine learning has developed and technology of artificial intelligence is used in various fields. Meanwhile, German online news mediaSPIEGEL ONLINEIs a professor at the University of Washington, which is a leading expert on artificial intelligencePedro DomingosWe are disclosing to Mr. interviewed about "possibility and danger possessed by artificial intelligence".

Pedro Domingos on the Arms Race in Artificial Intelligence - SPIEGEL ONLINE
http://www.spiegel.de/international/world/pedro-domingos-on-the-arms-race-in-artificial-intelligence-a-1203132.html

Domingos said "The Master AlgorithmIt is a person who wrote an introductory book for artificial intelligence for beginners. "The Master Algorithm" became the best seller in the world, without using too difficult mathematical expression, in relation to artificial intelligence and SNS, business and politics military technology and so on, "How can artificial intelligence of our lives Do you change it? "Is explained. In addition to acclaimed by Microsoft founder Bill Gates, this book is the best Chinese leaderXi JinpingIt became famous by being in line with Mr. 's bookshelf.

Russian President Vladimir PutinThose who control AI rule the world"I said in a speech to the students, Mr. Domingos also fully agrees with President Putin's opinion. "While AI is still in the process of competition among countries, Russia and China are concentrating very much on artificial intelligence and eventually China or Russia may win the AI ​​war," Domingos Mr. mentioned, the Xi Jinping's own book to the bookshelf had been decorated, China believes that it will shows the attitude to the ongoing development of the country of AI technology.

In response to the question "Why does control of AI lead to world domination?" By SPIEGEL ONLINE, Domingos answered, "AI can drastically reduce the cost of acquiring knowledge" and effectively By exploiting the machine learning system, I explained that the machine can easily obtain knowledge gained by working with one million people. In other words, dominating AI will lead to domination of the world because one country can have significant knowledge advantage over AI's country by properly managing AI.

Mr. Domingos who know that to Xi Jinping's bookshelves are lined up, "The Master Algorithm" is so "and excitement I felt fear at the same time."China raises "AI Dream"It is under the leadership of the state and is promoting AI's research, and if China, which is rapidly developing, pushes the development of AI technology to a large extent, benefits will spread not only to Chinese citizens but also to the rest of the world Mr. Domingos says it may not. Meanwhile, China is a very strong government authority, and it says that it is worth the threat to focus on controlling A citizen using AI. Domingos said, "As with many technologies, AI can be used for both good and bad things," AI said that AI can be a human threat.

byGlobal Panorama

Domingos received the popularity of "The Master Algorithm" and had the opportunity to talk with AI researchers and government officials around the world. In doing so, Domingos said that he was interested in different ways of thinking about AI for each region such as America, Asia, and Europe.

"The vision of AI that Silicon Valley of the United States has isLibertarianismAlthough it was optimistic affected by Europe, people from Europe received the opposite reaction to Silicon Valley. In many sessions that took part in Europe, people seemed to have a pessimistic idea against AI and seemed to have a consciousness of "humanity collapses by AI", Domingos said. Applications utilizing AI are being used in rural and national levels in the United States and China, medical research and traffic management, etc., while AI is not used much in Europe.

Also, "The Master Algorithm" has been translated into Russian and Chinese, but it has not been translated into German or French at the time of the interview. When SPIEGEL ONLINE told Mr. Domingos, the agent concerning the rights of Mr. Domingos's book said that "this book will not be sold in France or Germany".

byFrédéric BISSON

People in Europe have a threat to AI, Domingos says that work to regulate AI is preceded by the United States. However, it pointed out that understanding with respect to AI will not proceed as a result of precedence of regulation, and "EU Data Protection Regulation (GDPR)I am seeking accountability of "how data is used", but this idea is nonsense. " Mr. Domingos alreadyDiagnosis of cancer where machine learning is being usedAs an example, I have a cancer diagnostic system that can explain how I use data with 80% accuracy and 90% accuracy but how I use data I will ask myself what to do when a cancer diagnostic system that can not be explained is developed. And, from the EU philosophy, we have to use a system with accuracy of 80%, but originally you should choose a system with 90% accuracy, of course Domingos himself I will choose.

Regarding this point, Domingos said "The best algorithm is a neural network inspired by human beings and animals, and it has high accuracy to recognize the world from a very complicated situation, but what about the neural network Even experts can not fully understand whether we are doing the work or not, we just know "what the system does," he says, We should not allow only algorithms that can explain the outline.

bySavannah River Site

Domingos says regulation of AI should regulate what is developed for the purpose of pursuing profits, like Facebook algorithms to maximize profits by keeping users at the site. On the other hand, like data protection by GDPR, data protection regulation that hinders data collection useful for AI development is a bad hand, and despite the fact that there are excellent AI researchers in Europe, it can not be used altogether And that. In the present situation, although the quality of Europe is higher than that of China by China, China has established a system to support AI research entirely in the state, and it is easy for researchers to use data necessary for machine learning I pointed out that it is in the environment.

"However, there are" data sample diversity "in Europe, data tends to be homogeneous and uniform in China, but in Europe with people's diversity, diversity of data suitable for machine learning "Domingos says that European data is valid as a data sample.

Furthermore, it is said that it is no longer a state but a company that leads the development of AI, and it is Google that is currently running the world's most advanced AI development. Domingos said that Microsoft is following it, and that Facebook and Amazon are about to catch up from behind. In China, Alibaba, an online shop and Tencent for a huge Internet related company, are effective, but the search engine Baidu also seems to be a big bet on AI.

"FacebookUser information leakage scandalWhat do you think about? If SPIEGEL ONLINE asks questions, it is unlikely that data will be leaked from the current Facebook, but he said that it is no doubt that Facebook is a risky company from a privacy point of view. Also, "Facebook is trying to maximize profits with AI, but in the process, it deals damage such as showing fake news to the user etc. At some point, the government needs to intervene I will come. "

byLawrence OP

And against the allegation that AI threatens democracy, "While AI can be a useful tool to promote democracy, it can also pose a threat." As a tool to promote democracy, it is possible to tell politicians the demands of people by developing algorithms using AI, and to conduct politics reflecting more people's will. Meanwhile, the risk that AI is used as a tool to monitor people, and that people's actions can be abused as a means of operating the government as they wish may also occur in the future.

"For autocratic governments where democracy is a threat, computers that include AI will never be the best bureaucrat who will not go against government orders, unlike the Western army who has the right to refuse unwilling orders , The computer does not say "no", it will execute the order correctly, "Domingos said, and I acknowledged that AI is useful for dictators.

Mr. Domingos is worried that the authoritarian government may take control of AI and that people's rights may be infringed. "People may not be interested in AI, but AI is interested in people." Politicians can design AI in his or her preferred direction and so attractive There is not much for existence for politicians. "


"Highly developed AI is indistinguishable from God," Domingos talks in "The Master Algorithm", but this is a science fiction writerArthur C. ClarkMr. 's advanced technology seems to have been tampered with by not being able to tell from magic. "Some people are afraid that AI will be rebellious, but there is little possibility that there is a high possibility that we will abandon the voluntary control of AI and people are more likely to have AI It is necessary not to throw away the consciousness to control AI at all times in development, "Domingos said.

"AI is very clever, but it is not AI to control the influence on our lives, it is a person who controls AI, whether it is a democratic person or not to control AI, That will be the point, "Domingos said.