Will AI Defeat Humans?
- Hits : 481
- Registration Date : 2016-04-06
Humans continue to pursue a more convenient life. Achievements in scientific civilization resulted from their desire to minimize physical efforts. Today, people want technology to carry out everything so that they would do nothing but lie down in bed. This is why computers and communication technologies have been invented, and why an artificial intelligence (AI) having a very similar thinking system with humans has appeared.
AI enters the human world
In the five-round Go match between the professional Go player, Lee Sedol, and an AI, AlphaGo, from March 9 to 15, 2016, AlphaGo won in four games. Lee Sedol obtained one invaluable win, but this victory could not mitigate the shock of human beings. From the match, AlphaGo, which had recorded 504 wins and 1 defeat, added 4 more victories to its record. Go is far different from Gomoku, which has limited s. Go had been regarded as a game requiring human intuition, and thus, it was difficult to imagine that a machine would win a Go match against a human.
Another surprising news came out. A robot that can predict human actions was invented. This astonishing achievement was made by a research team in the Georgia Institute of Technology. Last October, the team developed an AI that analyzes the images of a human’s actions taken for six months and predicts the next actions of the human. The team categorized the images into 19 types, including cooking and working, and made the AI computer learn them. According to the research team, the AI can predict actions with 83 accuracy. They added that an AI robot that predicts the next actions of humans and provides all information by itself will come out in the future. All these inventions are based on the in-depth learning of technology.
In fact, AIs have already penetrated our daily life. The LA Times writes earthquake articles with “Quakebot,” which automatically collects information with regard to earthquakes. As it produces articles in real time, humans do not have to try writing articles quickly. Other news agencies, such as Reuters, use AIs for breaking news in the fields of sports and finance. In the United Kingdom, a robot delivery man that can on the streets without being bumped into a pedestrian and deliver products is going to be on test operation. In the United States, some hospitals are already using robots to prepare medicine.
There have been AIs before but not in this level. In 1956, an optical character reader (OCR), which can read characters, was called an AI. However, no one considers an OCR as an AI these days. In modern science, AI study is defined as follows: a field of computer engineering or intellectual technology to study in order to make a computer that can think, learn, and do self-improvement just like a human being. That is, an AI is an algorithm system configured to think, sense, and act like a human.
The key function of an AI is prediction through classification. This starts from data collection by a computer. A massive amount of data is analyzed to calculate almost all numbers of cases in order to discover an optimum value based on the analysis and to categorize data. The learning method of an AI is divided into “supervised learning” and “unsupervised learning.” In supervised learning, a computer is taught information. On the other hand, unsupervised learning does not include the step of teaching. An AI under unsupervised learning learns all by itself. AlphaGo started to learn from a database including about 30 million s, connecting the s of Go players from records of historical Go matches, in order to how a human being plays the game. After mastering the data, AlphaGo trained itself and improved its Go performance by playing the games with a number of other selves through reinforced learning.
AI is a human product
Development in AI is bringing about concerns of people. They fear that, one day, a machine might surpass a human being and dominate the world, as in the movie Matrix. However, experts dismiss possibilities of such a case. They explain that an AI technology is the fruit of mathematics and statistics, and differs from human intelligence as well as the structure or operational principle of the human brain. In other words, an AI is still a result of a development in computing speed.
On the day when Lee Sedol won over AlphaGo, AlphaGo was puzzled by one of the s of Lee. It was not fully prepared to deal with uncertainty, so it made vain attempts. An expert in Google commented that AlphaGo also had limits.
Both Lee and AlphaGo made mistakes. Actually, machines also make mistakes, just like humans. It is natural because an AI is made by a human. An AI created by an imperfect human cannot be perfect.
There is a difference in the way by which a human and a machine react after making a mistake. Lee retraced his s after he was defeated by AlphaGo and sought a way to deal with it. Only humans have the ability to ponder over the past and the present in order to predict the future. AlphaGo does not, and cannot, examine its defeats. It just collects more data to try to increase its winning rate.
A machine still stays on the level of a machine. It cannot make a comprehensive decision on numerous uncertainties in the human world. This is why a human still sparkles more than a machine.
For human beings, an AI is not an opponent to defeat but a standard to estimate human development and a companion to improve human life. Following developments in computer and communications, a new revolutionary era has come to face us, humans.