The development of artificial intelligence is rapid, but does it have to resemble humans to be recognized as useful intelligence? This book highlights the limitations of the Turing Test and the Chinese Room Argument, and raises the question of whether it is appropriate to judge AI based on human intelligence.
Computers have been advancing at a faster pace than we can imagine in recent decades. Artificial intelligence, a product of computer science, is accelerating as computers become more advanced. People have always been fascinated, sometimes even afraid, by artificial intelligence, as evidenced by the countless novels and movies about it. While the AI we have today can only understand us and give us limited answers, it’s possible that in a few years we’ll have AI that can think like humans. As AI has evolved, there have been many efforts to evaluate it, the most famous of which is the Turing Test.
In 1950, not long after the first computer, Alan Turing proposed the Turing Test to evaluate whether a computer was as smart as a human. The idea is simple. A panel of judges have an internet chat with a computer or human, and after five minutes of questions and answers, if more than 30% of the judges cannot distinguish between the computer and the human, the AI passes the test. The first to pass the test came 64 years after it was devised: a computer called Eugene, developed by the University of Reading in the U.K. Eugene was tested on a 13-year-old boy whose first language was not English to avoid awkwardness. While some have argued that this limitation does not make it a true success, some scholars predict that an AI that can pass the Turing test without setting limits could be as early as 2029. Given the rate at which AI is advancing, it is likely that the AI of the future will be much more like us than we think, and will eventually take over from humans in analyzing and processing vast amounts of information that is difficult for humans to process.
However, there are some who question the progress of AI. Does A.I. really need to resemble humans? The more it resembles humans, the better it is, and the less it resembles humans, the more backward it is? Of course, it’s easier to set goals and move forward if you have a certain standard for creating an AI, and it’s undeniable that humans are almost the only standard. However, humans are not the right standard for AI.
In the first place, the Turing test doesn’t measure how close an AI is to human intelligence. While it’s true that the Turing test can test whether a computer looks like a human, there are scholars who question whether passing the test means that a computer has human intelligence. For example, the Chinese Room Argument, a famous argument against the Turing Test, makes us rethink the meaning of the Turing Test. In the Chinese Room Argument, you put someone in a room who doesn’t speak Chinese and give them a list of questions and answers in Chinese. The person in the room doesn’t need to speak any Chinese to see the questions and answers and answer the judges’ questions. In this argument, the person in the room has no understanding of Chinese, and even if he answers perfectly, there’s no way to tell if he understands Chinese. So the argument is that the Turing test, which determines that a machine has intelligence, doesn’t tell us if it actually has intelligence or if it’s just giving stored answers. The argument is not that AI can’t exist, but that we need a more rigorous standard for judging it. Of course, we still don’t know how humans understand and answer questions, but we’re pretty sure it’s not the same way that current AIs ask questions and give stored answers. So if we truly want AI to think in a similar way to humans, we need to understand how human intelligence works in order to get there.
Efforts to understand and program human intelligence have been around for a long time. But before we can do that, can we be sure that all human intelligence works the same way? For example, the differences in how people viewed the controversial dress from earlier this year begs the question. The dress controversy reminded the world that different people see colors differently. What one person might see as white and gold, another might see as blue and black. If it hadn’t become a global controversy, some people would have believed that white and gold were the real colors, while others would have seen them as blue and black. This controversy shows that vision, which is part of intelligence, and even the ability to perceive color, varies from person to person, and we don’t know if everyone’s intelligence works the same way.
If we don’t know how intelligence works, and we don’t have a good grasp on it, how can we hold humans up to the standard of computers? And even if we did know how intelligence works, would it matter much? Does an AI have to think like a human to be called an AI? Suppose technology advances and we make contact with an alien civilization that evolved in a different environment and communicates, thinks, and perceives the world in a different way than we do. How would we evaluate the intelligence of this alien civilization? Just because they think in ways we don’t understand doesn’t mean they aren’t civilized. Using humans as the standard for evaluating intelligence can lead to incomprehensible situations. However, if we decide that aliens who think differently from us and are incomprehensible to us have intelligence, then computers that react similarly to humans or are useful to humans could be recognized as artificial intelligence.
To determine whether a human-made device, or computer, has artificial intelligence, it doesn’t have to behave in the same way as humans. Even if it doesn’t behave in the same way as human intelligence, it can and should be recognized as artificial intelligence if it is useful enough to humans and helps us in our lives.
