When it comes to the "man-machine war", I think many people still remember it and are familiar with it. That refers to the pinnacle of the game between two Go world champions (South Korean Go nine-dan player Lee Sedol and Chinese Go nine-dan player Ke Jie) and Google's artificial intelligence Go program Alphago. The first was a five-game chess match played in Seoul, South Korea from March 9 to 15, 2016, in which Alpha Dog defeated Lee Sedol by an aggregate score of 4-1; the second was a three-game chess match played from May 23 to 27, 2017 in Wuzhen, Jiaxing, China, where Alpha Dog Go defeated world number one Ke Jie by an aggregate score of 3-0, causing great repercussions and shocks around the world. People are asking the same question: Is the human brain stronger than the computer, or will the computer surpass the human brain?
Image source: baidu.com
In fact, before the Go Man-Machine War, the confrontation between computer programs and human players was staged in other chess games. For example, 24 chess pieces of checkers have been the first choice for Westerners for centuries, and this game, dating back to ancient Egypt, is relatively simple to play, but it is not easy to win. After 18 years of intensive research, Canadian computer scientists have shown that if two equally matched players compete, the result of the game must be a draw.
Overall, checkers are a strategy game that can't be won by luck at all. So computer scientists chose this game to test whether the human brain, which relies on its own ability to intuitively grasp the situation and make decisions, is superior to a computer that looks for a winning strategy among billions of possible solutions.
In 1989, Jonathan Schaeffer, a computer scientist at the University of Alberta in Canada, developed a computer program called Chinook and invited the top chess players at the time to compete with it. In the first checkers man-machine battle, Marion Tinsley, the mathematician's world champion of checkers, narrowly defeated Chinook by a narrow margin.
But two years later, the results changed. Professor Tinsley, who has only lost 7 games in his life, and Chenouk have played seven consecutive draws, and the cunning and exhausted mathematician has to choose to withdraw from the game, which is completely understandable, because his machine opponent does not know what fatigue is. Since then, there has been a lot of debate over whether computers are superior to the human brain in checkers.
Tinsley Image source: baidu.com
So Shafer continued to refine and develop his procedures, which was an astonishingly difficult task. Because there are 5 × 1020 possible scenarios during the game. The data is so huge that the computer's "brute force method" (also known as exhaustive algorithms—strategies that rely on the computer's computing power to exhaust every possible game situation and sift through it to sift through the best moves) are also difficult to work, which requires the human brain to think of smarter algorithms.
Schaffer and his team built a database to analyze the endgame when there were 10 or fewer pieces left on the board. Reduce the number of possible scenarios that need to be analyzed to only 39 billion, and then classify these endgames according to the black game wins, the white game wins and the draw. By 1996, they had spent seven years sorting out endgames of 8 or fewer pieces. The team then paused to wait for a more processing and faster computer to come out. When more advanced supercomputers emerged, the research team began to analyze the endgame of 9 and 10 sub-computers, often requiring up to 200 desktop computers to work at the same time.
In addition, the research team also analyzed various situations that will occur after taking 3 steps at the beginning. A specially developed program gives the best move for both participants to have a chance to win. That is to say, if two players play in this way, they will definitely end in a draw.
Up to now, in addition to checkers that have been challenged by computer programs, chess with a high number of matches of 1040 and more complicated situations has also been played against computers. The most famous chess man-machine battle was counted in two matches between 1996 and 1997 chess world champion Gary Kasparov against IBM's calculator "Deep Blue". In 6 sets of matches, Kasparov won the 1996 man-machine battle with a total score of 4-2, but the "Deep Blue" also won the first time against the top chess players, and in the 1997 man-machine war, the improved "Deep Blue" won for the first time with an aggregate score of 3.5:2.5. Later, chess masters constantly challenged the top artificial intelligence of chess, and at most they could only win a draw or a single victory, and they could no longer win in the overall score
Next, computer programs began to attack the only remaining highland of human intellectual games- Go. This process can be called difficult, because the number of changes that Go needs to calculate is too large, far more than the number of atoms in the universe that has been observed, and it is completely impossible to use computers to brute force cracking, but humans can rely on some kind of algorithm or intuition that is difficult to copy, and it is not difficult to understand that the starting Go program cannot even beat amateur chess players. When "deep learning" gradually enters the field of artificial intelligence research, the Go skills of computer programs have been improved at an unprecedented speed, and Alpha Dog is a product developed based on the catalog of deep learning technology, and finally won the highly anticipated Go man-machine war after continuous improvement.
Board games have long been regarded as the touchstone of top human intelligence and artificial intelligence. Although ARTIFICIAL programs already have an overall advantage in various board games, it seems not convincing enough to assert that human intelligence will be subordinated to ARTIFICIAL intelligence, and it is even sooner to judge that "computers are stronger than human brains". The only thing that is certain is that the development and game of the two will continue for a long time, let's wait and see!
Call for Papers
China Science Exploration Center WeChat public account welcome to give the manuscript!
The content of the manuscript takes anti-counterfeiting and breaking the mystery as the core idea, and can be involved in the fields of scientific popularization, scientific culture, philosophy of science and technology, science and the public, secular humanism, and scientific and technological ethics, aiming to inform the public of the results of scientific exploration without prejudice and avoid the public being deceived.
Once the manuscript is adopted, it will be paid.
Submission Email:
China Science Exploration Center
Advocating science, anti-pseudo-delusion