A common "misunderstanding" for AI (artificial intelligence) and answers to it



In recent years, it evolves rapidlyArtificial intelligence (AI)With technology, it is thought that automatic driving cars and perfect translation tools will be born in the near future. Not only that, I heard someone wary of the evolution of AI such as "AI will take away most of the work human beings are doing?" "Does not the risk of stopping AI's runaway?" It is. About AI which is getting familiar, it is like this when we summarize common misunderstandings and responses to it.

Q & amp; A: The future of artificial intelligence
http://people.eecs.berkeley.edu/~russell/temp/q-and-a.html

◆ What is AI?
AI is a generic term for methods that give intelligence to computers. Roughly speaking, to be able to distinguish between "good things" and "bad things". By "good thing" here is to take actions that are most likely to reach the goal (objective), application, inference, learning, planning, perception, language, robot technology etc. Yes.

· General misunderstanding: "AI is a specific technology"
As mentioned above, AI is a general term for "research to make intelligence to a computer", this is a mistake to capture "physics" as "steam engine".

◆ What kind of benefits does AI bring to human society?
All the civilization's rituals are the products of human intelligence. And the purpose of the instruments of all civilizations was to expand human capabilities. AI also extends human capability of intelligence, making it a better understanding of human value as long as it is properly designed.


· General misunderstanding: "AI is inhumane"
In a malicious scenario, AI is used in the wrong way, robotization, monitoring progresses, even justice is created by AI and it tends to be thought that a controlled economy will be born. However, a powerful means of expanding human intelligence is AI, and AI realizes much greater value as long as it is properly designed. For example, it seems to be said that situations where people using different languages ​​can communicate in real time with AI are very human-like.

· General misunderstanding: "AI promotes inequality"
There is certainly the possibility that wealth will concentrate on a small number of people by advancing work automation. However, it is possible to choose how to use AI, activate collaboration between individuals and small groups, and there is also the advantage that you do not necessarily need to ask jobs from large companies.

Machine learningWhat is?
Machine learning is one of the AI ​​technologies to explore ways to enhance the performance of computers based on experience.

· General misunderstanding: "Machine learning is a new field to replace AI"
Machine learning is one AI technology that improves performance and does not replace AI technology. Machine learning is one of the technologies that is constantly the center of AI research since Turing published a paper in 1950, one of the techniques studied in AI.

· General misunderstanding: "You can not learn a machine, but only do what you are told"
To put it briefly, the programmer can command the machine to learn. Many AI applications are made using machine learning which deals with large amounts of data.

neural networkWhat is?
A neural network is a computer system devised as a model of the characteristics of brain neurons. The neural network consists of many units, each receiving inputs and sending output. Output is a form of weighted sum of inputs and it is the key that the weight of the link connecting the units is corrected by experience.


· General misunderstanding: "Neural network is a new kind of computer"
A neural network is a computing system considered as a model of human brain neurons, and the neural network itself is implemented on a general purpose computer.

· General misunderstanding: "Neural network works like a brain"
As a matter of fact, neurons in the brain are not as simple as units used in AI 's neural network. There are various kinds of neurons in the brain, connection changes with time.

Deep learningWhat is?
Deep learning is a special machine learning using a multilayered neural network. Deep learning is a technique that became famous at once in recent years as a very important technology for image recognition and speech recognition.

· General misunderstanding: "Deep learning is a new technology replacing machine learning"
Since deep learning is a type of machine learning, it is incorrect to think deep learning and machine learning separately. Although deep learning itself existed for over 20 years ago, technological progress has made great progress in recent years due to improvements in algorithms and models and the acquisition of large amounts of data.

◆ What is "strong AI" and "weak AI"?
The expressions "strong AI" and "weak AI" are those used by John Searle to explain the hypotheses of the two AIs. Originally, "weak AI" was used in the hypothesis that "human intelligence can be programmed". In contrast, "strong AI" was used in the hypothesis that machines can think and understand like human beings.

· General misunderstanding: "Strong AI is to realize AI that can do everything on the human level"
A weak AI is commonly understood as a specific thing, a strong AI is a human being able to do everything on a human level, but it is different from what Sarl first defined in 1980.

◆ "Moore's lawWhat is?
Moore's Law is to predict the future based on an empirical rule that semiconductor integration density grows exponentially. "Semiconductor integration density doubles every 18 months" is simplified.


· General misunderstanding: "Moore's law is a physical law"
Moore's Law is a prediction that comes from experience with technology progress. No one knows whether this will continue and will not last forever.

· General misunderstanding: "Computers are already fast enough that development of efficient algorithms is futile enough"
In fact, the improvement of the algorithm is much more important than the improvement of the hardware.

◆ What is IQ (Intelligence Index) of a machine?
There is nothing like IQ on the machine. IQ can not be assumed for machines specialized in the field even though IQ can be handled by multiple people. Computers that defeat the chess champion can not win with other games such as Othello. A computer that can win the quiz competition can not even answer the simple question "What is your name?"

· General misunderstanding: "According to Moore's law, machine IQ continues to rise"
There is no IQ on the machine. Also, Moore's Law has nothing to do with the existence of an algorithm that can handle any task.

◆ What can the AI ​​system do now?
The range of tasks that a machine can handle is much wider than it was several years ago. In addition to doing board games and card games and answering simple questions, you can extract facts from news articles, build complex objects, translate languages ​​into different languages, recognize voices You can also determine the type of image and drive a car. Many of the functions of search engines also use AI.

· General misunderstanding: "A task like playing chess is the same on human being and machine"
Human learns by watching and listening to chess ways, but the chess program has no such ability. The algorithm of motion of pieces is directly programmed, and the computer does not know as human beings know the rules of chess.

◆ In the near future, how will AI affect human society?
Several innovations may emerge in the near future. A typical example is an automatic driving car. We can also predict the appearance of robots working in the agricultural field and elderly nursing care field and evolution of the personal assistant function of smartphones. In addition, it is expected to be utilized in the scientific field by AI that can deal with massive and complicated information.

· General misunderstanding: "Robot is trying to deprive human's role"
The majority of progress in the AI ​​field is to make computers and robots convenient, and progress has been made incrementally little by little. Nonetheless, from a long-term perspective, it is important to control by humans.

◆ Does AI and robot evolution deprive work from human beings?
If AI and robot evolution will continue in the future, it is unlikely that many occupations will be affected. This does not mean a large amount of unemployment right away, but it may lead to changes in the economic structure. And you may be asking for a new way of thinking about the way of work and reward itself.


· General misunderstanding: "All work done by the robot reduces human work"
The work is not zero sum. With the help of the robot, productivity may be increased or new demand may be created by the robot. Work without a robot may not be economically feasible. Although it is common to use a brush or a spray when repainting a house, this is like not finishing until completion, if ink is dropped one by one.

◆ Why did people suddenly worry about AI?
After 2014, media have regularly reported that celebrities such as Steven Hawking, Earon Mask, Steve Wozniak and Bill Gates are "concerned about AI." Another reason is the fact that the progress of AI technology seems to be accelerating. There, various AI technologies are tied together, and the achievements that were being held in the laboratory reached the level at which the problems in the real world can be solved, investment in research has increased, and further progress has been made as research progresses There are likely to be multiple factors such as acceleration.

· General misunderstanding: "The birth of superintelligent AI with intelligence beyond human intelligence is approaching soon"
Superintelligent There are only a few researchers who think that the birth of AI is close. However, this does not mean that you do not think seriously about the appearance of superintelligent AI. Assuming that a meteorite colliding with Earth was found within 50 years, there should be no one saying "Let's pay attention after 5 years distance."

◆ What does the AI ​​community do to the risk of AI that can become reality?
Despite the negative response from most of the AI ​​researchers, much discussion on the existential risk of AI is out of the subject of the AI ​​community. However, recently Earon Mask has provided $ 10 million (about 1.1 billion yen) to support research in this field, and for the long-term research that Eric Horbitz leads to making policy recommendations We do activities such as investing in. Then,AAAIHas established a standing committee dealing with "the degree of influence and ethical issues of AI".

· General misunderstanding: "It is impossible to control AI research"
Some people think that there is no way to avoid a bad future due to AI because progress by research is not stopped and regulation is impossible. However, this claim is a mistake. For example, it was held in 1975Asilomatic conferenceAs a result of giving a time grace to make a conclusion, succeeded in making an international norm for subsequent genetic recombination technology. If studies that realize human level AI are done without checking, the importance of research on techniques to ascertain that the AI ​​system is still under human control will be further increased.

in Note,   Software, Posted by darkhorse_log