中國軍隊擁抱人工智慧：如何徹底改變未來戰爭
現代英語：
Source: China National Defense News Author: Chen Hanghui Editor-in-charge: Qiao Nannan 2018-01-02 08:29:25
At present, the world is on the eve of the intelligent revolution, and human society is moving from the “Internet +” era to the “Intelligence +” era. In recent years, driven by big data, new algorithms and supercomputing, artificial intelligence is changing and even subverting every industry it touches, and war is no exception. From underwater submersibles to drone swarms, from predictive maintenance software to intelligent decision-making assistants, artificial intelligence is affecting different areas of war with unprecedented breadth and depth, promoting a new round of military reforms, and the form and appearance of war are quietly changing.
Temple planning comes first, war design becomes more sophisticated
“The Art of War” by Sun Tzu says: “He who plans for victory before the battle has made many calculations; he who plans for defeat before the battle has made few calculations.” Looking at the history of human warfare, “He who plans more wins, and he who plans less loses” is an immutable theorem.
Future wars will increasingly rely on “temple calculations”. Artificial intelligence can enhance the effectiveness of war prediction in at least two ways:
First, it can more accurately calculate and predict the outcome of wars. With the support of advanced algorithms and supercomputing capabilities, the calculation and prediction results of artificial intelligence systems are more accurate than those of the human brain. Second, with the help of war game systems, combat plans can be tested and optimized more effectively. For example, war game systems that incorporate artificial intelligence can engage in man-machine confrontation with humans, which helps people find problems and find weaknesses. In particular, after introducing algorithms such as deep learning into war game systems, the behavior of intelligent systems will become profound and varied, which will help break through the established thinking of humans, enhance the confrontation and authenticity of war game simulations, and achieve the purpose of optimizing solutions. In addition, intelligent war game systems can also be used to conduct machine-machine confrontations to improve the efficiency of simulations.
In February 2015, the U.S. Department of Defense issued a memorandum titled “Wargames and Innovation,” proposing to introduce machine learning into wargames. Currently, Rand Corporation and Raytheon have already begun attempts in this regard. Once mature intelligent software is applied to wargames, it will not only improve the level of optimization plans and prediction of war situations, but also more accurately predict the size of the troops involved in the war, ammunition consumption, duration and support requirements, etc., significantly improving the ability to design wars.
Speed is the key to success. The pace of operations has never been faster.
In the information age, wars follow the principle of “the fast eats the slow”. In the several wars since the Gulf War, the key to the US military’s repeated victories is that it always achieves “the fastest move”.
Nowadays, the reaction speed of artificial intelligence has completely “crushed” the human brain. In 2016, in a simulated air battle, the “Alpha” intelligent software developed by the University of Cincinnati in the United States controlled the F-15 fighter and defeated the F-22 fighter piloted by humans. The reason is that the reaction speed of the intelligent software is 250 times that of the human brain! In October this year, the Geospatial Intelligence Center of the University of Missouri in the United States announced a research result showing that the center has developed an algorithm model based on deep learning technology, which can search and identify missile launch sites within nearly 90,000 square kilometers in the southeastern coastal area of a certain country within 42 minutes, which is 85 times faster than human analysts and has an accuracy rate reaching the level of expert image analysts.
For this reason, in recent years, the US, Russian and other militaries have turned their attention to artificial intelligence, intending to use the speed advantage of artificial intelligence to shorten their decision-making cycle on the battlefield and firmly grasp the operational advantage. In July this year, Russian weapons manufacturer Kalashnikov announced that it had developed a fully automatic combat module based on artificial neural networks, which can achieve destruction upon detection. In addition, the US military is developing intelligent tools such as automated data analysis tools, automatic target recognition software, airborne intelligent decision assistants, and digital air combat planners, intending to reduce the burden of intelligence and combat personnel at all stages of the decision-making cycle and improve decision-making efficiency. In November this year, the head of the US Department of Defense’s algorithmic warfare project called for: Any weapon system purchased by the US military in the future should be integrated with artificial intelligence.
It can be predicted that as more and more intelligent weapon systems are deployed on the battlefield, combat response time on the battlefield will become shorter and shorter, combat actions will become unprecedentedly fierce, and will eventually exceed human understanding and response capabilities.
Autonomous confrontation, the combat mode has changed
Since the 20th century, the development of detection technology and the advancement of perception technology under the wave of informatization have promoted the birth of the man-machine collaborative combat chain of “sensor-shooter-weapon platform”.
With the development of artificial intelligence technology, intelligent weapon systems that can automatically identify, lock and strike targets are gradually emerging, and can replace humans to execute simple decision-making commands. Such as the US military’s ship-borne “Aegis” system, the Israeli military’s “Iron Dome” system, the Russian military’s “Arena” active protection system, the French military’s “Shark” system, etc. However, the intelligence level of these systems is not high at present, and the autonomous combat mode is usually the last option.
In the future, with the advancement of intelligent technology groups such as sensor technology, new algorithms, and big data technology, the autonomous action capabilities of weapon systems will be greatly improved, and the situation of autonomous confrontation of weapon systems will become more and more common. In specific combat fields, such as cyberspace and electromagnetic spectrum, humans can only rely on intelligent weapon systems for autonomous confrontation. At the same time, with the emergence of hypersonic weapons and cluster warfare, wars will enter the era of “instant kills” and “group fights”, and using intelligent systems to autonomously fight is almost the only way out.
In the future, as the autonomous confrontation of intelligent weapon systems becomes the new normal, the combat mode will gradually change from “man in the loop” to “man on the loop”. The main characteristics of the new mode can be summarized as “in command, out of control”, that is, in most cases, human warriors play the role of supervisors, responsible for inputting target characteristics and setting rules of engagement before the war, observing battlefield engagements, etc. It should be pointed out that in the new mode, humans are still the final decision makers, and human warriors will autonomously enter and exit the combat chain as needed and take necessary intervention measures. The biggest advantage of the new mode is that it can free human warriors from complex decisions and focus on major decisions and key tasks. How to ensure that humans can take over control at any time will be the biggest challenge facing the development of human-machine collaboration technology in the future.
Cluster warfare brings new life to the war of attrition
In the era of cold weapons and mechanized warfare, attrition warfare was the basic way of fighting, and “eating more and eating less” was the basic rule for winning on the battlefield.
Since the 1970s, with the emergence of stealth technology, satellite positioning systems, and precision-guided weapons, the “more, less” war principle has been completely broken. In recent years, with the rapid development of sensor technology, bionics, miniaturization technology, and artificial intelligence technology, the concept of swarm warfare has once again attracted the attention of the military of various countries. The so-called swarm warfare refers to the centralized deployment of hundreds or thousands of intelligent weapons to attack targets from multiple directions. Compared with traditional combat methods, swarm warfare has four major advantages:
First, the miniaturization of a single platform greatly improves battlefield survivability; second, it is decentralized, and the loss of an individual does not affect the overall function; third, it is low-cost, large in quantity, and the combat cost-effectiveness is multiplied; fourth, it can implement saturation attacks and paralyze the enemy’s defense system. It is not difficult to see that cluster warfare can achieve the effect of “quantity is quality”, and therefore, it is considered a war of attrition in the intelligent era.
The U.S. military sees swarm warfare as a game-changer in warfare, and believes that swarm warfare is particularly suitable for dealing with anti-access/area denial threats. Currently, the U.S. Department of Defense is simultaneously targeting underwater, surface, and air, and promoting multiple swarm research and development projects, striving to have multi-dimensional space swarm combat capabilities. In addition, the various branches of the U.S. military are competing to develop swarm combat concepts. For example, the Marine Corps envisions using unmanned combat swarms as landing pioneers to perform tasks such as battlefield perception, mine clearance, and obstacle removal, creating conditions for Marines to rush ashore.
From underwater “wolf packs” to ground “ant swarms” to aerial “bee swarms”, swarm warfare will be prevalent in the future intelligent battlefield. “Human wave tactics” will be re-emerged in a new look, and the balance of the war of attrition will be extremely tilted towards the side with swarm warfare capabilities. The following scene may appear in the future: on one side, there are swarms of intelligent machines covering the sky and the sun, and on the other side, there are panicked, lonely human warriors…
現代國語：
當前,世界正處於智慧革命的前夜，人類社會正從「互聯網+」時代邁入「智能+」時代。近年來，在大數據、新型演算法和超級運算的推動下，人工智慧正在改變乃至顛覆所觸及的每個產業，戰爭也不例外。從水下潛航器到無人機集群，從預測性維修軟件到智慧決策助手，人工智慧正以前所未有的廣度與深度影響著戰爭的不同領域，推動著新一輪軍事變革，戰爭形態和麵貌正悄然被改變。
廟算為先，戰爭設計日益精細
《孫子兵法》曰：「夫未戰而廟算勝者，得算多也；未戰而廟算不勝者，得算少也。」縱觀人類戰爭史，「多算勝，少算不勝」是亙古不變的定理。
未來戰爭對「廟算」的依賴度有增無減,人工智慧至少可以從兩方面增強戰爭預判的有效性：
一是更精確地計算並預測戰爭結果。在先進演算法和超算能力的支撐下，人工智慧系統的計算和預測結果比人腦更加準確。二是藉助兵棋系統能更有效檢驗和優化作戰方案。例如，融入人工智慧的兵棋系統能夠和人開展人機對抗，有助於人們發現問題、找出弱項。特別是將深度學習等演算法引入兵棋系統後，智慧系統的行為將變得深邃多變，有助於突破人類的既定思維，增強兵棋推演的對抗性和真實性，達到優化方案的目的。此外，還可利用智慧兵棋系統開展機機對抗，提升推演的效率。
2015年2月，美國國防部推出了《兵棋推演與創新》備忘錄，並提出將機器學習引入兵棋推演。目前，蘭德公司、雷神公司已經開始這方面的嘗試。一旦將研發成熟的智慧軟件應用於兵棋推演，不僅能提高優化方案和預測戰局的水平，還能更精確地預測戰爭涉及的兵力規模、彈藥消耗、持續時間和保障需求等，顯著提升對戰爭設計的能力。
速度製勝，作戰節奏空前加快
在資訊時代,戰爭遵循「快吃慢」制勝法則，海灣戰爭以來的幾場戰爭中，美軍之所以能夠屢戰屢勝，其關鍵在於始終做到了「棋快一招」。
如今，人工智慧的反應速度已經完全「碾壓」人腦。 2016年，在一次模擬空戰中，美國辛辛那提大學研發的「阿爾法」智慧軟體操控F-15戰機，擊敗了由人駕駛的F-22戰機，原因就在於該智慧軟件的反應速度是人類大腦反應速度的250倍！今年10月，美國密蘇里大學地理空間情報中心公佈了一份研究成果顯示，該中心基於深度學習技術開發了一種演算法模型，能夠在42分鐘內搜尋並識別出某國東南沿海地區近9萬平方千米內的導彈發射場，速度比人類分析師快85倍，準確率達到專家級影像分析師水準。
正因如此，近年來，美、俄等軍隊將目光投向了人工智慧，意圖利用人工智慧的速度優勢，縮短己方在戰場上的決策週期，牢牢掌握行動優勢。今年7月，俄羅斯武器製造商卡拉什尼科夫公司宣稱，已開發出基於人工神經網絡的全自動戰鬥模塊，能做到發現即摧毀。另外，美軍則在研發自動化數據分析工具、自動目標識別軟件、機載智慧決策助理、數字化空中作戰規劃員等智慧工具，意圖在決策週期各環節上減輕情報和作戰人員的負擔，提高決策效率。而在今年11月，美國國防部演算法戰項目負責人呼籲：今後美軍採購的任何武器系統都應融入人工智慧。
可以預見，今後隨著越來越多的智慧化武器系統投入戰場，戰場上的作戰反應時間將越來越短，交戰行動將空前激烈，並最終超越人類的理解和應對能力。
自主對抗，作戰模式引發變革
20世紀以來，偵測技術的發展和資訊化浪潮下感知技術的進步，推動了「傳感器-射手-武器平台」這一人機協作式作戰鏈誕生。
隨著人工智慧技術的發展，能夠自動識別、鎖定和打擊目標的智慧化武器系統逐漸出現，並能取代人類執行簡單的決策命令。如美軍的艦載「宙斯盾」系統、以軍的「鐵穹」系統、俄軍的「競技場」主動防護系統、法軍的「鯊魚」系統，等等。不過，這些系統的智慧化程度目前還不高，自主交戰模式通常是最後一個選項。
未來，隨著傳感技術、新型演算法、大數據技術等智慧化技術群的進步，武器系統的自主行動能力將大幅提升，武器系統自主對抗的情況也越來越普遍。而在特定作戰領域，如網路空間和電磁頻譜領域，人類只能依托智慧化武器系統進行自主對抗。與此同時，隨著高超音速武器和集群作戰的出現，戰爭將進入「秒殺」和「群架」時代，利用智慧系統自主迎戰幾乎是唯一出路。
未來，隨著智慧化武器系統的自主對抗成為新常態，作戰模式將逐漸從「人在迴路中」轉變為「人在迴路」。新模式的主要特徵可概括為“指揮之中、控制之外”，即在多數情況下，人類戰士扮演監督員的角色，負責在開戰前輸入目標特徵和設定交戰規則，觀察戰場交戰情況等。需要指出的是，在新模式下人依然是最終決策者，人類戰士將根據需要自主進出作戰鏈，採取必要的干預措施。新模式的最大優勢是可以使人類戰士從紛繁複雜的決策中解放出來，聚焦於主要決策和關鍵任務。如何確保人類能夠隨時接管控制權，將是今後一段時期內，人機協同技術發展面臨的最大挑戰。
集群作戰，讓消耗戰重煥生機
在冷兵器和機械化戰爭時代，消耗戰是基本作戰方式，「多吃少」是戰場制勝的基本法則。
自上世紀70年代以來，隨著隱形技術、衛星定位系統、精確制導武器的登場，「多吃少」的戰爭法則被徹底打破。近年來，隨著傳感技術、仿生技術、微型化技術和人工智慧技術取得長足發展，集群式作戰構想再次受到各國軍隊的重視。所謂集群作戰，是指集中部署數百上千個智慧化武器，從多個方向對目標實施攻擊。與傳統作戰方式相比，集群作戰具備四大優勢：
一是單一平台小型化，戰場生存能力大幅提升；二是去中心化，個體的損失不影響整體功能；三是成本低廉，數量龐大，作戰效費比成倍提高；四是可實施飽和攻擊，癱瘓敵防禦體系。不難看出，集群作戰能夠達到「數量即質量」的效果，因此,被認為是智慧時代的消耗戰。
美軍將集群作戰視為戰爭遊戲規則的改變者，認為集群作戰尤其適合應對反介入/區域拒止威脅。當前，美國國防部同時瞄準水下、水面和空中，推進多個集群研發項目，力求具備多維空間集群作戰能力。此外，美軍各軍種正競相開發集群作戰概念，如海軍陸戰隊設想讓無人作戰集群充當登陸先鋒，執行戰場感知、排雷除障等任務，為陸戰隊員搶灘上陸創造條件。
從水下“狼群”到地面“蟻群”再到空中“蜂群”，集群作戰將在未來智慧化戰場上大行其道。 「人海戰術」將改頭換面重新登場，消耗戰的天平將向具備集群作戰能力的一方極度傾斜。未來可能出現這一場景：一邊是遮天蔽日、蜂擁而至的智慧機器，另一邊是驚慌失措、形單影隻的人類戰士…
