On the battlefield, will artificial intelligence defeat human intelligence？

　　On March 15th, 2016, Google’s artificial intelligence "AlphaGo" defeated Go world champion Li Shishi with a disparity score of 4-1. On May 27th, 2017, China’s Ke Jie, who is recognized as the world’s No.1 Weiqi player, lost 0-3 in a three-piece chess contest with Afa Dog.

　　It is reported that in the research team of "Afa Dog", only Huang Shijie can play Go for four amateur paragraphs. The level of amateur four-stage Go may only be to understand the basic rules and some simple skills of Go. And once he is combined with the artificial intelligence "Alpha Dog", he can actually play a super-class chess player without winning. This fact reminds us that a commander who understands basic operational rules and tactical principles, if combined with the corresponding operational command-aided decision-making artificial intelligence, will defeat the first-class military commander without artificial intelligence?

　　Where does artificial intelligence come from?

　　Electronic computers and the Internet are born in response to military needs, and so is artificial intelligence, a branch of computer science. In 1939, World War II broke out and Britain declared war on Germany. Alan mathison turing, a mathematician and logician engaged in research work in Cambridge, enlisted in the army and joined the code-breaking station ordered by Churchill. The deciphering machine named "Turing Bombe" developed by them successfully cracked the communication code of the German Supreme Command Center, code-named "Enigma". Later generations commented that the "Turing Bomb" advanced the end of World War II by at least two years. After the war, Turing mainly engaged in the pioneering research of computer programming theory, neural network and artificial intelligence. He was called "the father of artificial intelligence" because of a series of outstanding achievements.

　　According to the Chinese definition, "intelligence" includes people’s "wisdom" and "ability", that is, intelligence and ability. The English expression of artificial intelligence is: artificial intelligence; Artificial means "artificial" and "imitation", which is strictly translated. Only Artificial Intelligence can be translated into "artificial intelligence" or "artificial intelligence". In fact, the current "Alpha Dog" is really just a "Go Brain", which decides how to play chess, and the chess movement is done manually.

　　The superb Go level of "Alpha Dog" benefits from the deep learning method provided by new technology with big data science and technology as the core. It can be described visually as follows: the program only describes the learning and thinking methods of "Alpha Dog", and the improvement of its intelligence level depends on the accumulated data. For "Alpha Dog", the data include the chess score of Go masters, the chess score of real opponents and the chess score of "Alpha Dog" playing by itself. These chess scores are the source of "intelligence" for "Alpha Dog" to improve its chess skills. In terms of digesting the chess score and learning progress, the ability of "Alpha Dog" is beyond anyone’s reach. Ke Jie, the world champion, can play up to 1,000 chess games a year, while Alpha Dog can play 1 million chess games a day.

　　This makes the improvement of "Alpha Dog" chess power open, and there is almost no "ceiling". On the premise of not modifying the "Alpha Dog" program, the "Alpha Dog" can achieve the progress and improvement of "new day by day" or even "new from time to time" as long as the chess score of high-quality chess is continuously increased. It is objectively proved that the research team of "Alpha Dog" has basically mastered the winning mechanism of Go and described it in programming language.

　　This remarkable ability includes two aspects: one is to judge the advantages and disadvantages of the existing disk; Second, based on this judgment, the possible position of the next move is decided, and a decision is made after the value judgment of several possibilities. Whether it is to judge the value of a certain move or the quality of the whole disk, it is necessary to construct a certain mathematical model and accurate algorithm.

　　The success of "Alpha Dog" shows that its R&D team’s efforts in this regard are very close to human cognition, and machine learning has stood on the threshold of human brain learning. Super-first-class chess players found that "Alpha Dog" can even discover the laws and knowledge that human beings have not discovered for thousands of years from big data, which means that human beings have opened up new channels for expanding their knowledge system.

　　What is the use of artificial intelligence

　　Because of the realistic ability and potential ability displayed by artificial intelligence, all countries in the world are trying to use artificial intelligence for national defense and army building. As far as reality is concerned, the application of artificial intelligence in military field is mainly reflected in three aspects.

　　The first is an unmanned combat platform. The existing unmanned combat platforms, whether unmanned aerial vehicles, robots on land or unmanned submarines at sea, have low intelligence. Generally, it only replaces people for some specialized operation. On the battlefield in Iraq, the US military used more than 4,000 robots to detect mines, roadside bombs and operate in some dangerous areas. The US military has more than 8,000 aerial unmanned systems and more than 12,000 ground unmanned systems, which have become an indispensable part of the US military’s operations. The development of artificial intelligence technology can make unmanned platforms more "smart" and improve the combat effectiveness of weapons and equipment, but it is not possible for unmanned platforms to make independent decisions so far.

　　The second is intelligent precision ammunition. The US military used advanced laser or TV guidance in missile technology, which began in the Vietnam War in the 1970s. Its "Gemstone Road" missile hit more than 60% on the battlefield in Vietnam. The JDAM "Joint Direct Attack Ammunition" equipped by the US military adopts a more advanced guidance mode: inertial guidance +GPS satellite guidance. The "Joint Direct Attack Ammunition" during the Gulf War can only add target data before taking off. In the Iraq war, its receiver can receive the target data from navigation satellites, and can change the target at any time to destroy it, and its "intelligence" is obviously improved.

　　The third is command-assisted decision-making. Think of "Alpha Dog" as a command and decision system of black-and-white confrontation, and its excellent performance in cognitive decision-making of black-and-white confrontation, and its ideas and methods should be used for reference for operational command and decision-making.

　　Situation awareness on the battlefield is very complex, and multi-source intelligence gathers to form massive data, which cannot be completed by relying on the commander’s manual analysis. Artificial intelligence has great advantages over human beings in searching, storing, calculating and mining massive data. Introducing a deep learning algorithm similar to "Alpha Dog" to intelligently analyze satellite images and radar data will greatly improve the efficiency of intelligence analysis.

　　In addition, at present, the cognitive understanding of the comprehensive battlefield situation depends largely on the commander’s experience and intuition. Using the "value" algorithm similar to "Alpha Dog" to judge the situation of Go and the probability of winning or losing, and learning from the "strategy" algorithm similar to "Alpha Dog" to determine the position of the player, it can provide auxiliary support for the commander’s operational decision.

　　However, the application of artificial intelligence in the military field is far from reaching the level of independent decision-making like "Alpha Dog" Go. In November 2001, Predator UAV launched an attack on Al Qaeda military commander Mohammed Latif in Afghanistan, which became the first anti-terrorism actual combat of UAV. But it is the drone operator who controls from the rear who presses the "fire" command. Even so, it will inevitably bring collateral damage to innocent civilians. In 2013, Pakistani officials announced that since 2008, the US military has launched more than 300 drone air strikes in Pakistan, killing more than 2,000 innocent civilians.

　　What is the future of artificial intelligence?

　　From the situation that the world’s first-class Go players, Li Shishi and Ke Jie, have little chance to win the game against "Afa Dog" and the result of complete defeat, it is not exaggerated to draw the conclusion that artificial intelligence has surpassed human beings in the field of Go. In the game of Go, the obvious advantages of artificial intelligence are as follows: without fatigue curve, intelligence will not decay with time; There are basically no mistakes, and no opportunities will be given to opponents where they can; Free from emotional interference, not angry and impulsive because of the provocation of opponents; Learned and memorized, learning and playing chess more than ten million times before the game, which is beyond the reach of human beings; Powerful computing power is beyond the reach of human beings.

　　But the victory of "Alpha Dog" is, in the final analysis, a human victory. The research on the mystery of human thinking and brain science is still superficial, and the development of human brain is far from exhausted, but the mechanism and composition of computers and artificial intelligence, no matter how powerful, are well known, because they were originally designed, manufactured and developed by people.

　　At least so far, any artificial intelligence has the ability given by human beings. It is not surprising that it surpasses human beings in some ways. Humans make up for their own shortcomings by making various tools, a process that began in the Stone Age. Today, the major means of transportation far exceed human’s walking ability, and various machine manufacturing systems far exceed human’s unarmed manufacturing ability. Digitalization, networking and intelligence are the basic tracks of the development of the information society. It is expected that artificial intelligence will show its great power in the mature stage of the information society.

　　Then, will there be a first-class military commander completely defeated by an artificial intelligence command system with strong deep learning ability in the battlefield in the future? To fully answer this question requires a masterpiece, but there are many essential differences between chess game and battlefield confrontation. For the following reasons, the author thinks the answer is no.

　　One is the one-way transparency of the battlefield and the two-way transparency of the chess game. In the game of Go, both sides of the game are equal and transparent, regardless of the chessboard structure or the disk situation. However, the modern battlefield spans the virtual and physical space of land, sea, air, sky, electricity and internet, and the weather, weather, rain and shine, flood and tide are unpredictable. Clausewitz’s "battlefield fog" covers the battlefield. Even if it is a close confrontation, the battlefield space can never be transparent in both directions. The party with strong perspective observation can realize that the battlefield is more transparent to itself. Two-way reciprocal transparency in battlefield space has never existed and will never exist.

　　The second is the certainty of the value of chess and the uncertainty of the effect of combat action. In the game of Go, the value of one player’s move is almost certain on the disk. An opponent can weaken the value of his next move, but he also has the same chance to fight back.

　　However, the effect of combat operations on the battlefield is uncertain. In October 2001, the United States Joint Forces Command published "Effect-based Operations", and the US military formally accepted the theory of "Effect-based Operations". In 2008, General james mathis, the current U.S. Secretary of Defense and then commander of the United States Joint Forces Command, issued the Guide to Effect-based Operations, which officially abolished the theory of Effect-based Operations. Matisse’s reason is: "All combat environments are full of countless changing factors, so it is logically impossible to accurately predict the outcome of an action."

　　The third is the authenticity and deception of the situation. On the board of Go, the situation of the board seen by both players is clear and true. But as Lenin said: "There is no war in the world without strategy." Strategic and tactical deception is a common tactic in war.

　　The fourth is the strict game between the two sides and the potential multi-party participation. On the chess board, the two sides of the game are strictly limited between the two sides. Although there is a team behind the "Alpha Dog", it cannot change the nature that they are one of the two sides. In the history of human wars, especially since World War II, many local wars are actually "proxy war", where two sides or three parties join in the battle. Behind them, the situation is complicated and the behind-the-scenes promoters are hidden. For example, in the battlefield in Afghanistan, al-Qaeda is the target of the US military; On the Syrian battlefield, however, they may get some support from the US military because they are anti-Syrian forces.

　　The fifth is the symmetry and asymmetry of space, time and weapons and equipment. In the game of Go, the time and space of both sides are symmetrical, and the weapons and equipment are black and white chess pieces. Although they are black first, they must pay back. In the real battlefield, preemptive strike is asymmetric in time, and air-to-ground attack is asymmetric in space. The wars of the United States against Iraq and Afghanistan are the wars of information-based weapons and equipment against mechanized weapons and equipment in the agricultural era, respectively, and there is intergenerational asymmetry in weapons and equipment between the opposing sides.

　　Sixth, the certainty of chess rules and the flexibility of operational principles. In the game of Go, "Jinjiao’s silver-edged grass belly" is the basic value judgment criterion, and it is the basic chess rule … … The connection is broken, the dragon is encircled, and the eye position is broken, which is to avoid passivity with all one’s strength. These rules are basically certain and must be observed. Militarily, victory is no longer an important tactical principle. However, in the Anti-Japanese War, Liu Bocheng made a second ambush in Qigen Village, which became a classic of military command and a stroke of genius.

　　Clausewitz said: War is a kingdom of uncertainty. The most essential difference between Go and military confrontation is the difference between certainty and uncertainty. The success of artificial intelligence is an amazing scientific achievement; But military command is not only science, but also art. The superb command art is more derived from the commander’s imagination and inspiration, generate, who said, "The wonderful operation is single-minded".

　　Generally speaking, the application of artificial intelligence in the military field will enter an unprecedented active era with the development of technology, but the leader of war or battlefield will always be people.

　　(Author: Wu Minwen Unit: National Defense Information College)