In this blog post, we will look back at the criteria for evaluating artificial intelligence through the Turing test and the Chinese room argument, and consider whether it is possible to become useful intelligence without resembling humans.
Computers have developed at a speed that we cannot imagine in the last few decades. Artificial intelligence, a product of computer science, is accelerating as computers develop. As seen in numerous related novels and movies, people have been interested in artificial intelligence, sometimes even fearing it. The artificial intelligence we are currently exposed to only understands what we say and gives limited answers, but there is a possibility that artificial intelligence that thinks like humans will appear in a few years. As artificial intelligence has developed, there have been many efforts to evaluate it. The most famous measure is the Turing test.
In 1950, not long after the first computer was released, Alan Turing proposed the Turing test to evaluate whether a computer is as smart as a human. This test method is simple. The judges chat with the computer or a human on the Internet, and if more than 30% of the judges cannot distinguish the computer from the human after five minutes of question and answer, the AI has passed the test. The first instance of passing this test occurred 64 years after the test was devised, with a computer called “Eugene” developed by the University of Reading in the UK. “Eugene” was tested on the basis of a 13-year-old boy whose native language was not English to avoid awkwardness when taking the test. Some have pointed out that this limitation means that the AI cannot be considered to have passed the test, but some scholars predict that an AI that passes the Turing test without setting a limit will appear as early as 2029. Considering the pace of AI development, future AI will be much more similar to us than we think, and it is expected to replace humans in the task of analyzing and processing vast amounts of information that is difficult for humans to handle.
However, some people question the development of artificial intelligence. Does artificial intelligence have to resemble humans? Is it accepted that the more it resembles humans, the better the artificial intelligence, and if it does not resemble humans, is it considered to be inferior? Of course, it is easier to set goals and move forward if there are certain standards for creating artificial intelligence. And it cannot be denied that humans are almost the only standard. However, humans are not appropriate as a standard for artificial intelligence.
The Turing test does not measure how close an AI is to human intelligence in the first place. While it is acknowledged that the Turing test can test whether a computer looks human, some scholars question whether a computer that passes the test can be said to have human intelligence. For example, the Chinese room argument, a famous counterargument to the Turing test, makes us rethink the meaning of the Turing test. The Chinese room argument is a test in which a person who does not know Chinese is placed in a room and given a list of questions in Chinese and the answers to them. The person in the room can answer the judge’s questions by looking at the questions and answers even if he or she cannot speak Chinese at all. In this argument, the person in the room has no understanding of Chinese, and even if he gives the correct answer, there is no way to know if he understands Chinese. Therefore, the Turing test, which determines whether a machine has intelligence, cannot tell whether the machine actually has intelligence or is simply giving a stored answer. Seol’s argument is not that artificial intelligence cannot exist, but that a stricter standard is needed when judging artificial intelligence. Of course, the principle by which humans understand questions and answer them has not yet been elucidated, but I am confident that it will be different from the way that artificial intelligence currently receives questions and provides stored answers. Therefore, if true artificial intelligence is to think in a way similar to that of humans, the first step is to understand how human intelligence works.
There have been efforts to understand and program human intelligence for a long time. But before understanding it, can we be sure that human intelligence works the same way? For example, the difference in the way people see the controversial dress earlier this year raises this question. The dress controversy has impressed upon the world that different people see colors differently. While some see white and gold, others see blue and black. If this had not become a global controversy, some people would have believed that white and gold were real colors, while others would have seen them as blue and black. This controversy shows that even the ability to recognize color, which is part of intelligence, varies from person to person, and it is not known whether everyone’s intelligence works the same way.
If we don’t even know how intelligence works and don’t have a good understanding of it, how can we measure a human being by computer standards? Also, even if we know how intelligence works, does it really matter? Can artificial intelligence be called artificial if it does not think like a human? Let’s assume that technology has advanced to the point where we can make contact with an extraterrestrial civilization. Let’s assume that this extraterrestrial civilization has developed in a different environment from humans, and that it communicates, thinks, and perceives the world in a different way from us. How would we evaluate the intelligence of this extraterrestrial civilization? Just because they think in a way that we cannot understand does not mean that they have not achieved civilization. If we use humans as the standard for evaluating intelligence, we may encounter situations that we cannot understand. However, if we determine that aliens, whose way of thinking is different from ours and who we cannot understand, also have intelligence, then computers that react similarly to humans or are useful to humans may be recognized as artificial intelligence.
When determining whether a device created by humans, i.e. a computer, has artificial intelligence, it is not necessary to use the same criteria as when determining whether a human has artificial intelligence. Even if it does not function in the same way as human intelligence, if it is sufficiently useful to humans and helps our lives, it can be recognized as artificial intelligence, and it should be.
