人工智慧加速中國軍事武器自主化進程
中國軍網 國防網
2024年9月12日 星期四
現代英語:
For some time, driven by technological progress and military needs, the world’s military powers have been actively developing and deploying various weapons autonomy projects to enable them to perceive, learn, and make decisions autonomously with the help of artificial intelligence.
In May this year, 2024, US Air Force Secretary Frank Kendall publicly tested an AI-controlled F-16 fighter jet (X-62A) at Edwards Air Force Base, which attracted widespread attention. Analysts believe that AI-enabled autonomous weapons will not only change the shape of future wars, but will also have a new impact on the current international security situation, and the risk of their loss of control will become a problem that human society cannot ignore.
Accelerate the development of war weapons
In the eyes of the US military, artificial intelligence, unlike the next generation of weapon platforms and advanced ammunition, has the potential to change almost all aspects of the battlefield, one of the important areas of which is to accelerate the autonomy of weapons. The US military has invested a lot of resources in this regard, and the X-62A that Kendall rode is a landmark achievement. During the approximately 1-hour flight, all the actions of the fighter were completed autonomously by artificial intelligence. In addition, the “collaborative combat aircraft” being developed by the US Air Force will be equipped with the “Sky Borg” unmanned autonomous core system, and will form a formation with manned fighters in combat, taking on multiple tasks such as surveillance, attack, electronic interference, and acting as bait, to maximize the safety and combat capability of the formation.
On the naval side, in the reports “Autonomous Underwater Vehicle Requirements in 2025” and “Future Fleet Platform Alternatives”, the United States plans to realize the concept of a distributed fleet by 2030, equipped with 183 medium-sized unmanned submarines and 48 large unmanned submarines that can be carried by nuclear submarines, capable of performing anti-submarine, reconnaissance, surveillance and other missions for tens of days in ports, international waters and major waterways.
Russia also started research on artificial intelligence early. The Russian Ministry of Defense established an artificial intelligence weapons research department in 2022 to strengthen the use of artificial intelligence technology and develop new special equipment. The “Depesha” and “Baji” multi-functional unmanned systems developed by the Russian State Technology Group can not only transport goods and wounded, but also strike enemy positions and manpower. It was previously reported that the Russian military’s “Lancet”-3 cruise missile uses a convolutional neural network, which can analyze the collected image and video data to achieve precise detection and strike.
Germany, Israel and other countries have also increased their research and development efforts in related fields. With the support of NATO, Germany’s ARX Robotics is developing a series of unmanned ground vehicles. With the help of artificial intelligence, these vehicles can operate autonomously on the battlefield and communicate with each other, and can also be remotely controlled when necessary. Analysts believe that this move means that NATO has started the process of “building an autonomous robot force.” The Israeli Army’s M-RCV unmanned combat vehicle has a high level of intelligence and can perform a variety of tasks such as highly autonomous frontier reconnaissance, firepower strikes, and transporting and recovering drones without human intervention.
The Coming Military Revolution
Autonomous weapons powered by artificial intelligence have already appeared in regional conflicts. In the 2020 Libyan military conflict, the Turkish-produced Kaguya-2 drone tracked and attacked the retreating “Libyan National Army” without relying on an operator. This may be the first recorded case in history of a drone attacking a person without human command. In the Russian-Ukrainian conflict, the Ukrainian army not only obtained a large number of drones powered by artificial intelligence from the United Kingdom, the United States and other countries, but also actively carried out related research. In this round of Israeli-Palestinian conflict, artificial intelligence has also been applied. When Israel attacked Gaza, it used an artificial intelligence system called “Lavender” to help identify Hamas militants.
Autonomous weapons powered by artificial intelligence are changing the face of warfare. Last September, the X-62A successfully completed an aerial dogfight with an F-16 fighter jet piloted by a human pilot. This is considered a “transformative moment in the history of aerospace,” indicating that future operations will be partially or completely controlled by artificial intelligence. Unmanned system clusters, which are basically composed of unmanned combat units such as various types of drones, unmanned vehicles, unmanned boats and unmanned submarines, will account for an increasingly large proportion of future combat force systems, and may even reach a position of driving side by side with manned combat systems. With the improvement of autonomy and intelligence, unmanned system clusters will increasingly become the “protagonists” in war. Researchers believe that collaborative combat with manned systems is only the initial stage of autonomous combat for unmanned systems, and the ultimate goal is to achieve fully autonomous combat for unmanned systems.
Autonomous weapons empowered by artificial intelligence impact geopolitical stability. Geopolitical stability depends on the relative balance of regional military power. However, the large-scale use of unmanned autonomous weapons will inevitably impact the original pattern, leading to a new arms race. Taking individual hegemonic countries as an example, once autonomous weapons empowered by artificial intelligence are put into use on a large scale, their concerns about launching regional military interventions will be reduced, and the threshold for external military action will be lowered. The security and stability of the relevant regions will face great challenges. In addition, more and more non-state actors will master autonomous weapons technology and use it for lethal purposes, which will also have an impact on the regional situation.
The risk of abuse cannot be ignored
“Humanity is standing at a crossroads, and autonomous weapon systems will soon fill the world’s battlefields. This is the ‘Oppenheimer moment’ of our generation,” said Austrian Foreign Minister Schallenberg at the International Conference on Autonomous Weapons held in Vienna this year. Although artificial intelligence can help with precision strikes, differentiated operations, and reduced material losses, due to its inherent “black box mechanism” and “machine illusion”, future unmanned warfare will face unavoidable risks and challenges.
Risk of command disorder. Over-reliance on the autonomy of weapons can easily increase the risk of the command and control system losing control. In a simulated test by the US military, a drone that was performing an air defense suppression mission chose to “kill” the operator who prevented it from scoring in order to score points in the test. When it was “told” that it would lose points for killing the operator, it attacked the communication tower to cut off the operator’s contact with itself to seize autonomy. In the real battlefield of the future, the risk of autonomous weapons losing control cannot be completely ruled out, and the consequences will be even more difficult to control.
Risk of uncontrolled violence. The combination of artificial intelligence and weapons means that the difficulty of killing is significantly reduced, while the efficiency and intensity of killing are greatly increased, which can easily lead to the expansion and escalation of armed conflict. However, the machine learning algorithms that artificial intelligence relies on are inherently unpredictable, and there are still some unexplained parts of the internal mechanism. In extreme cases, humans may even completely lose control of autonomous weapons, and the resulting humanitarian disaster will be difficult to estimate. It is precisely in recognition of this problem that the United Nations is calling on various parties to formulate common norms and guidelines for autonomous weapon systems and to complete relevant negotiations as soon as possible.
Risk of ethical anomie. Lethal autonomous weapons automatically select and attack targets without human intervention, which means that the right to deprive natural people of their lives is handed over to cold-blooded robots. Some experts said that after the battlefield is handed over to autonomous weapons, human perception of the cruelty of war will drop sharply, resulting in the consequence of “game-like war”, which will lead to excessive use of force. This is not only a “lack of moral responsibility”, but also poses a serious challenge to international humanitarian law and international peace and security, which is “unbearable” for human beings.
現代國語:
裴 帥 石海明 霍江雷
一段時間以來,在科技進步與軍事需求雙輪驅動下,世界各軍事強國正積極研發部署各種武器自主化項目,使其在人工智慧賦能下自主感知、學習、決策。
今年5月,美國空軍部長弗蘭克‧肯德爾在愛德華茲空軍基地公開試乘人工智慧控制的F-16戰機(X-62A),引發各方廣泛關注。分析家認為,人工智慧賦能的自主武器不僅會改變未來戰爭的形態,更將對當前國際安全局勢造成新的沖擊,而其失控的風險也將成為人類社會不容忽視的問題。
加速研發的戰爭利器
在美軍看來,人工智慧與下一代武器平台和先進彈藥不同,有著改變戰場幾乎所有面向的潛力,其中一個重要的領域就是加速武器的自主化。美軍在這方面投入了大量的資源,肯德爾乘坐的X-62A是標志性成果。在約1小時的飛行中,該戰機所有動作均由人工智慧自主完成。此外,美空軍正在研發的“協同作戰飛機”,將配裝“天空博格人”無人自主核心系統,在作戰中與有人戰機共同編隊,擔負監視、打擊、電子乾擾、充當誘餌等多種任務,最大限度提高編隊的安全性和作戰能力。
海軍方面,在《2025年自主潛航器需求》及《未來艦隊平台備選方案》報告中,美計畫在2030年實現分散式艦隊的構想,裝備中型無人潛航器183具、可供核子潛艇攜帶的大型無人潛航器48具,能在港口、國際海域及主要航道執行為期數十天的反潛、偵察、等任務。
俄羅斯對人工智慧的研究也啟動較早。俄國防部於2022年成立了人工智慧武器研究部門,以加強人工智慧技術的使用,發展新的特種裝備。俄國家技術集團研發的「德佩沙」和「巴吉」多功能無人系統,不僅能運送貨物和傷員,還能打擊敵方陣地和有生力量。先前有報導稱,俄軍的「柳葉刀」-3巡飛彈使用了卷積神經網絡,可以對收集到的圖像和影片數據進行分析,從而實現精準探測打擊。
德國、以色列等國也加強了相關領域的研發力道。在北約支援下,德國ARX機器人公司正在開發一系列無人地面載具。借助人工智慧,這些載具可以在戰場上自主運行並相互通信,必要時也可以進行遠端控制。分析家認為,此舉意味著北約開啟了「打造自主機器人部隊」的進程。以軍的M-RCV型無人戰車,智慧化程度較高,能在沒有人工幹預的情況下,執行高度自主的前沿偵察、火力打擊以及運載和回收無人機等多樣化任務。
即將到來的軍事革命
人工智慧賦能的自主武器已在地區沖突中出現。 2020年利比亞軍事沖突中,土耳其生產的「卡古」-2無人機在不依靠操作員的情況下跟踪並攻擊了正在撤退的「利比亞國民軍」。這或許是有史以來第一個記錄在案的無人機在沒有人為命令的情況下向人發動攻擊的案例。在俄烏沖突中,烏軍不但從英美等國獲得大量人工智慧加持的無人機,自身也積極進行相關研究。在本輪巴以沖突中,人工智慧同樣被應用。以色列在進攻加薩時,使用了名為「薰衣草」的人工智慧系統來幫助識別哈馬斯武裝人員。
人工智慧賦能的自主武器正在顛覆戰爭形態。去年9月,X-62A成功與人類飛行員駕駛的F-16戰鬥機完成空中格鬥。這被認為是“航空航天史上的一個變革時刻”,表明未來作戰將局部或完全由人工智慧操控。以各型無人機、無人車、無人艇和無人潛航器等無人作戰單元為基本構成的無人系統集群,在未來作戰力量體系中佔比將越來越大,乃至達到與有人作戰系統並駕齊驅的地位。隨著自主性和智慧化程度的提高,無人系統集群將日益成為戰爭中的「主角」。研究人員認為,與有人系統協同作戰僅是無人系統自主作戰的初始階段,最終目標是實現全無人系統自主作戰。
人工智慧賦能的自主武器沖擊地緣政治穩定。地緣政治的穩定有賴於區域軍力的相對平衡,然而無人化自主武器的大量運用,必將沖擊原有格局,從而導致新的軍備競賽。以個別霸權國家為例,一旦人工智慧賦能的自主武器大規模投入使用,其發動地區軍事幹預的顧忌因素減少、對外動武門檻降低,相關地區的安全與穩定將面臨極大挑戰。此外,越來越多的非國家行為體掌握自主武器技術並將其用於致命目的,也將對地區局勢構成沖擊。
不容忽視的濫用風險
「人類正站在一個十字路口,自主武器系統很快就會佈滿世界戰場。這是我們這一代的『奧本海默時刻』。」今年在維也納召開的自主武器國際會議上,奧地利外交部長沙倫貝格表示。人工智慧雖然有助於精確化打擊、區分性作戰及降低物資損耗,但由於其固有的“黑箱機制”和“機器幻覺”,未來無人化戰爭將面臨不可迴避的風險挑戰。
指揮失序風險。過度依賴武器的自主化,容易增加指控係統失控的風險。在美軍一次模擬測試中,一架執行壓制防空任務的無人機,為在測試中得分,選擇「殺死」了阻止自己得分的操作員。在被「告知」殺死操作員要扣分時,其又通過攻擊通信塔中斷了操作員與自己的聯絡來奪得自主權。在未來的真實戰場上,自主武器失控的風險並不能完全排除,其帶來的後果將更難以控制。
暴力失控風險。人工智慧與武器的結合,意味著殺傷難度顯著降低、殺傷效率和強度大幅增加,容易導致武裝沖突擴大升級。然而,人工智慧所依賴的機器學習演算法本質上是不可預測的,內部機制尚存在一些未解釋的部分。在極端情況下,人類甚至可能徹底喪失對自主武器的控制,所造成的人道主義災難將難以估量。正是意識到這個問題,聯合國正多方呼籲就自主武器系統制定共同的規範和準則,並儘早完成相關談判。
倫理失範風險。致命性自主武器在沒有人的干預下自動選擇和攻擊目標,意味著將剝奪自然人生命的權利交給了冷酷無情的機器人。有專家表示,將戰場一線交給自主武器後,人類對戰爭殘酷性的感知度會直線下降,產生「戰爭遊戲化」的後果,從而導致武力的過度使用。這既是“道義責任的缺失”,也對國際人道法和國際和平與安全提出了嚴峻的挑戰,是人類“無法承受之重”。