This essay argues that there is a path to the replacement of armies with a national defense system consisting entirely of automated responses and weapons. Based on the future many national armies already see in unmanned weapons, this culmination of military evolution may be closer than it seems. Should we continue reacting with horror to unmanned weapons? They may be inevitable, and we could hope they will form part of a strong deterrent structure that minimizes violence and helps push us to pacifism.
It is well-known that many countries manufacture and operate unmanned aerial vehicles (UAVs) as weapons,[1] and this is leading to an interest in solutions to guide similar armed drones without the need for remote control by humans. We can see this already in the concept of "drone swarms", groups of semi-autonomous drones that could participate in battles.[2] Weapons equipped with artificial intelligence (AI), are pursued by the US, Russia and China.[3] Russia's military chief Valery Gerasimov considered in 2013 that "a fully robotized unit will be created, capable of independently conducting military operations."[4]Taken together, such news reports suggest the seemingly fictitious pursuit of robot armies by states is real.
Relying mostly on news reports, this essay will consider the merger of two different areas of military technological innovation. The first is the development of reliable unmanned aerial, naval and ground combat units. The second is the more hypothesis-laden topic of military artificial intelligence (AI), which may be applied to coordinate individual combat vehicles and eventually entire units on at least a tactical level. Military experts have not suggested the breakthrough of artificial general intelligence referred to by AI experts[5] is necessary to hit targets or outperform a human tactician, so that concept will be irrelevant here. This essay will be new in its attempt to make possible the concept of a complete national defense system that can consist wholly of automated units, reliant on automated responses and measures, to entirely replace a manned army. In doing so, this essay will address some military and political challenges such a national defense system could face and argue that its creation is achievable.
Before addressing the present direction of the two major areas of innovation concerned, we can consider an example of a similar defensive initiative. Unmanned defenses and automated retaliatory measures are not new ideas, although similar projects have never been seen as desirable. Nuclear weapons in Russia have been considered as possible components of an automated retaliatory system, in response to the close proximity of NATO forces to the Russian capital. In such a system, launches of the opposing (US) side's nuclear missiles would be detected and automatically trigger a retaliatory strike from Russia without requiring any human authorization.[6] If such powerful weaponry can be launched automatically, why not much less drastic military options deployed in a purely defensive manner?
On the less ambitious end of the spectrum, automated defenses including self-aiming stationary defenses like turrets with machine-guns already exist[7] and intelligent mines have already been advanced as ideas.[8] What seem to be persistently absent from this mix are unmanned anti-tank guided missile (ATGM) launchers and other towed or man-portable launchers. Even the most advanced such weapons known to exist only gained "fire-and-forget" technology and still require crews exposed to danger[9] to perform loading and firing. If better types of stationary defenses were to be relied on more heavily in future wars, they could significantly reduce the need to endanger human personnel with the low-value task of guarding a single point and help deter the most common low-tech threats.
Beyond stationary point defense is the adoption of mobile unmanned weapons. Russia,[10] China,[11] the US,[12] the UK[13] and others[14] all fund programs that consider drones as essential players in the future of warfare. Unmanned weapons such as these are best examined when divided into the different domains of ground, naval, aerial, and space combat and will be addressed in that order here. For a state to fully automate its defense, it would need to have capable responses to threats in all military domains.
The biggest current obstacles to unmanned warfare exist on the ground. Hostile personnel, being flexible and fast, can present a never-ending source of confusion and challenges to drone tanks. Such remote-controlled tanks, called unmanned ground vehicles (UGVs), have been trialed in combat by the Russian Federation but are hindered by difficulties in signal,[15] cheap and unreliable design features,[16] and optics.[17] One summary of such problems may simply be that engineers have so little historical experience of creating full-scale unmanned vehicles, in contrast with manned vehicles, that almost all the problems are new and the solutions are still decades away.
Plans to overcome signal problems and the threat of jamming by making drone tanks drive autonomously may be complicated by the "cluttered and unpredictable" nature of battlefields.[18] Whatever targeting software is involved also appears to be too simple to deal with the huge array of threats from a speedy and proficient enemy.[19] The answer to such problems seems to lie in machine learning analogous to the same elusive way image recognition bots are developed.[20] For weapons, that would mean testing them repeatedly with the help of massive volumes of recorded data from real and simulated combat situations until results are promising enough for them to learn in real situations.
Done tanks appear to be the most challenging of all types of military robot to develop. Commercial self-driving cars are likely to remain drastically ahead of them for the foreseeable future. Due to their relationship, advances in self-driving car software can be expected to result in successes that filter down into the technologies carried aboard drone tanks.[21] The weapons will become more reliable as this happens. Self-driving car software appears to already be based on the same machine learning mentioned previously,[22] verifying that this is the likely route that will be taken to create smarter autonomous weapons. Machine learning used for image recognition should be just as applicable to all other data, including audio, radar and sonar, that may be recorded and acted upon in a combat situation.
The biggest current obstacles to unmanned warfare exist on the ground. Hostile personnel, being flexible and fast, can present a never-ending source of confusion and challenges to drone tanks. Such remote-controlled tanks, called unmanned ground vehicles (UGVs), have been trialed in combat by the Russian Federation but are hindered by difficulties in signal,[15] cheap and unreliable design features,[16] and optics.[17] One summary of such problems may simply be that engineers have so little historical experience of creating full-scale unmanned vehicles, in contrast with manned vehicles, that almost all the problems are new and the solutions are still decades away.
Plans to overcome signal problems and the threat of jamming by making drone tanks drive autonomously may be complicated by the "cluttered and unpredictable" nature of battlefields.[18] Whatever targeting software is involved also appears to be too simple to deal with the huge array of threats from a speedy and proficient enemy.[19] The answer to such problems seems to lie in machine learning analogous to the same elusive way image recognition bots are developed.[20] For weapons, that would mean testing them repeatedly with the help of massive volumes of recorded data from real and simulated combat situations until results are promising enough for them to learn in real situations.
Done tanks appear to be the most challenging of all types of military robot to develop. Commercial self-driving cars are likely to remain drastically ahead of them for the foreseeable future. Due to their relationship, advances in self-driving car software can be expected to result in successes that filter down into the technologies carried aboard drone tanks.[21] The weapons will become more reliable as this happens. Self-driving car software appears to already be based on the same machine learning mentioned previously,[22] verifying that this is the likely route that will be taken to create smarter autonomous weapons. Machine learning used for image recognition should be just as applicable to all other data, including audio, radar and sonar, that may be recorded and acted upon in a combat situation.
The future of naval drones shows promise, with the adoption of unmanned underwater vehicles (UUVs).[23] With the most concerning challenges to armed drones arguably existing in the domain of ground warfare, potential dangers to naval drones are not as significant. Naval armed drones are being trialed in various countries,[24] including a Russian drone torpedo called the Poseidon that acts as an autonomous nuclear-armed, nuclear-powered torpedo[25] and has strengthened Russia's nuclear deterrent.[26]
UUVs may have some of the best prospects as unmanned weapons in actual combat, since a non-operating drone underwater could at least be difficult for an enemy to locate. Such a drone could lose signal and be useless for a time, but not be truly lost or surrendered like its air and land counterparts. It can be hypothesized that an underwater drone could be created with the option to remain inert on the seabed, in order to gain an opportunity to reactivate and attack when the enemy is not ready.
Air-to-ground drone warfare is often thought to be fully developed already, but is totally ineffective in contested airspace.[27] Present technology appears to be inadequate to challenge a human pilot. Air-to-air drone warfare may remain too difficult at present, but development in that direction does seem to be of interest to China.[28] The most common use of lethal drones so far seems to be in ethically questioned assassinations by the US.[29] Because these are aimed at non-state targets, with the expectation that they will be undefended and unaware that they are being targeted, the effectiveness of drones as front-line weapons remains untested by the US. What is clear is that current drones and other unmanned weapons are effective only when acting in uncontested and remote spaces. It can be concluded that current unmanned weapons are very far from being flexible enough to compete with enemy manned weapons or enemy personnel in direct combat. Just as with drone tanks, however, this would be overcome by making the weapons autonomous and capable of learning.
Space warfare can already be considered entirely unmanned, because all space objects with potential military value are unmanned satellites[30] heavily dependent on computers and mission control centers on Earth. The fact the only presence held by states on other planets is robotic, too,[31] warns us that robots will be central to space militarization if and when it speeds up. Creating a "space corps", meaning military combatants stationed in orbit,[32] could prove to be a catastrophic mistake while space flight continues to be such a fragile task. This is because opposing states may instead solely focus on robotic satellites[33] that are much less vulnerable and much more menacing in the vacuum of space than any pressurized vehicle or suit. Small satellites, new missiles,[34] and intentionally placed debris[35] created by a desperate enemy could easily inflict overwhelming losses on a space corps and bring them burning down to Earth. Unmanned craft could also survive longer away from the Earth without resupply. Of all the different areas of warfare today, therefore, space warfare would be the most likely to see effective dominance of unmanned weapons, and that is because personnel would be too hard to sustain there. Space is already favorable to the non-living.
Lack of autonomy in drone weapons is their greatest handicap. They are little more than oversized remote-controlled toys in their present form, not much different than their World War 2 era predecessors, remote-controlled "Goliath" mines.[36] They may be even more vulnerable because signals can be disrupted and jammed[37] and obstacles can block signals,[38] potentially surrendering vehicles and their armaments to the enemy or leaving aerial drones vulnerable to interception by a much cruder enemy aircraft.
We can see the creation of lethal autonomous weapons (LAWs) is being pursued by states despite the push for international prohibitions against them.[39] Engineers are likely to consider autonomy important to overcome aforementioned problems inherent to remote-control by human operators, especially in the case of drone tanks. Given the current situation, military competition between states can reasonably be expected result in significant strides in the development of autonomous drone weapons in coming decades, possibly as commercial self-driving software filters into military software.[40] If Russia's military chief is correct,[41] this will lead to robotic units that can continue to carry out decisive combat actions even when their signal is severed, and perhaps fight to restore their connection with their commanders. These designs could, we can speculate, be able to defend their own technology from capture and even self-destruct when parts of their casing are opened by unauthorized personnel. Whether this is possible or useful will have to be explored by the relevant professionals.
Beyond the future engineering breakthroughs in lethal autonomous weapons, it is predictable that a whole different form of warfare will emerge in the long-term. This form of warfare will be the result of artificial intelligence continuing to be incorporated onboard military drones[42] or in the signals infrastructure used to control them from afar, or both. Although a completely automated army is not a stated goal of any country, evidence of amazing military engineering will make it increasingly plausible to propose much heavier funding for such projects in the future. It may not even emerge as a specifically funded project at all, but a simple reorganization of robotic forces and integration of future weapons and other systems into that national defense system once the machines are reliable.
Whether leadership of tactical-level decisions for drones could be handed over to a form of artificial intelligence like a robot general is a much more difficult topic to address than the drones themselves, so the following is highly speculative. The breakthroughs to artificial general intelligence and "super" AI[43] are almost certainly not necessary, although the state achieving these first may have a big advantage.[44] The success of computers in chess[45] and various games[46] can be viewed as precedent they can already make swifter and superior decisions than a human expert, but we must respect the fact real wars are much more complicated than any game.[47] It is possible some biological elements of creative thought and aggressiveness are required to act decisively in the fog of war and turn the tide of a battle, but it is not clear if these require anything like a human mind. What is clear is that, because war is so complex, it will be hard to maintain the flow of sufficiently detailed data between drones and an artificial general to allow it to make the type of fast or well-informed decision it would make in a game. This brings us back to the issue of maintaining signal. Whereas food and medicine would never be issues, a whole set of new baggage and responsibilities would drive the actions of an army consisting only of robots.
The biggest engineering problem, even in the exceedingly hypothetical idea of a fully automated military, may not be the creation of competent artificial generals to manage artificial armies but the already discussed problem of maintaining communication in a chaotic environment. It makes sense that such a problem would only be further compounded if a remotely-based artificial intelligence was giving instructions to robotic units and waiting for their feedback. The data being transmitted to and from the units would be immense. Time delays would exist. However, it is possible that the challenges would then be overcome simply by building ever more specialized and larger drones. Antenna-carrying models could be assigned to carry signals over buildings and other troublesome objects and overpower jamming attempts. These hypothetical vehicles may need to be very large, costly, and high-powered to provide maximum coverage. They would become high-value targets that require significant protection from other drones. Other types of supporting drone could be created specifically to attack sources of interference detected by them. Some drones could be adapted to act as commanders by interpreting many signals from lower-ranking drones and issuing orders back to them.
What is not speculation is the way funding and policy preferences of governments are going to be the deciding factors in whether defense becomes more automated, and to what degree. Current military robots and drones can be accurately described as being of limited variety and low cost,[48] indicating governments fund them grudgingly and wait for results before deciding what to do next. The production of war machines capable of acting autonomously and being led by other machines will likely depend on the creation of many more ambitious variants of military drone, with ever increasing size and cost. Whether this is purely science fiction talk or the next big revolution in warfare will depend on the choices of political leaders in the country prepared to take that leap.
The crossover from human warriors and their horses to machines as the main agents of warfare has been in the works for hundreds of years already, but may be close to its ultimate conclusion. It can be argued that drone armies and AI will eventually overtake humans in their ability to win wars. The end result, although still remote, may be that humans become obsolete in all domains of war. Over sufficient generations, it can be hoped that this reduces the militaristic upbringing and values of much of the population and reduces warlike sentiments and policies. Whether that is a helpful prediction has not yet been discussed in any literature so far and is worth further inquiry.
Harry Bentham
Exclusively for The clubof.info Blog
[1] Dillow, C., "All of These Countries Now Have Armed Drones", http://fortune.com/2016/02/12/these-countries-have-armed-drones/, Fortune, 12 February 2016, Retrieved 3 October 2018
[2] Lendon, B., "U.S. Navy could 'swarm' foes with robot boats", https://edition.cnn.com/2014/10/06/tech/innovation/navy-swarm-boats/index.html, CNN, 13 October 2014, Retrieved 3 October 2018
[3] O'Connor, T., "Russia's Military Challenges U.S. and China By Building a Missile That Makes Its Own Decisions", https://www.newsweek.com/russia-military-challenge-us-china-missile-own-decisions-639926, Newsweek, 20 July 2017, Retrieved 3 October 2018
[4] LaPointe, C., and Levin, P. L., "Automated War", https://www.foreignaffairs.com/articles/2016-09-05/automated-war, Foreign Affairs, 5 September 2016, Retrieved 3 October 2018
[5] Dickson, B., "What is Narrow, General and Super Artificial Intelligence", https://bdtechtalks.com/2017/05/12/what-is-narrow-general-and-super-artificial-intelligence/, TechTalks, 12 May 2017, Retrieved 3 October 2018
[6] Bender, J., "Russia May Still Have An Automated Nuclear Launch System Aimed Across The Northern Hemisphere", https://www.businessinsider.com/russias-dead-hand-system-may-still-be-active-2014-9?IR=T, Business Insider, 4 September 2014, Retrieved 3 October 2018
[7] Parkin, S. "Killer robots: The soldiers that never sleep", http://www.bbc.com/future/story/20150715-killer-robots-the-soldiers-that-never-sleep, BBC, 16 July 2015, Retrieved 3 October 2018
[8] Bergstein, B., "'Smart' land mines, with remote control", http://www.nbcnews.com/id/4664710/ns/technology_and_science-science/t/smart-land-mines-remote-control/, NBC News, 4 April 2004, Retrieved 3 October 2018
[9] "Antitank guided missile", https://www.britannica.com/technology/antitank-guided-missile, Britannica.com, Retrieved 3 October 2018
[10] Majumdar, D. ,"Russia Is Developing a Mysterious Unmanned Strike Aircraft", https://nationalinterest.org/blog/the-buzz/russia-developing-mysterious-unmanned-strike-aircraft-23941, The National Interest, 4 January 2018, Retrieved 3 October 2018
[11] Huang, K. "The drones that have become part of China’s military strategy", https://www.scmp.com/news/china/diplomacy-defence/article/2161354/drones-have-become-part-chinas-military-strategy, South China Morning Post, 26 August 2018, Retrieved 3 October 2018
[12] "Predator C Avenger Unmanned Aircraft System (UAS)", https://www.airforce-technology.com/projects/predator-c-avenger-unmanned-aircraft-system-uas/, Airforce Technology
[13] "Taranis", https://www.baesystems.com/en/product/taranis, BAE Systems, Retrieved 3 October 2018
[14] Dillow, C., "All of These Countries Now Have Armed Drones", http://fortune.com/2016/02/12/these-countries-have-armed-drones/, Fortune, 12 February 2016, Retrieved 3 October 2018
[15] Mizokami, K., "Russia’s Tank Drone Performed Poorly in Syria", https://www.popularmechanics.com/military/weapons/a21602657/russias-tank-drone-performed-poorly-in-syria/, Popular Mechanics, 18 June 2018, Retrieved 3 October 2018
[16] "Russian Uran-9 unmanned combat vehicle tested in Syria", https://www.youtube.com/watch?v=0MiiUOjmqLo, Binkov's Battlegrounds, YouTube, 20 July 2018, Retrieved 3 October 2018
[17] "Combat tests in Syria brought to light deficiencies of Russian unmanned mini-tank", https://defence-blog.com/army/combat-tests-syria-brought-light-deficiencies-russian-unmanned-mini-tank.html, Defence Blog, 18 June 2018, Retrieved 3 October 2018
[18] "Unmanned Ground Vehicle Technology Issues", Autonomous Vehicles in Support of Naval Operations, The National Academies Press, 2005, pp. 148-153
[19] "Russian Uran-9 unmanned combat vehicle tested in Syria", https://www.youtube.com/watch?v=0MiiUOjmqLo, Binkov's Battlegrounds, YouTube, 20 July 2018, Retrieved 3 October 2018
[20] "What is the Working of Image Recognition and How it is Used?", https://www.marutitech.com/working-image-recognition/, Maruti Techlabs, Retrieved 3 October 2018
[21] Wall, R. "Armies Race to Deploy Drone, Self-Driving Tech on the Battlefield", https://www.wsj.com/articles/armies-race-to-deploy-drone-self-driving-tech-on-the-battlefield-1509274803, WSJ, 29 October 2017, Retrieved 3 October 2018
[22] "What is the Working of Image Recognition and How it is Used?", https://www.marutitech.com/working-image-recognition/, Maruti Techlabs, Retrieved 3 October 2018
[23] Rees, M., "General Dynamics Demonstrates Naval Unmanned Systems C3 Capabilities", https://www.unmannedsystemstechnology.com/2018/09/general-dynamics-demonstrates-naval-unmanned-systems-c3-capabilities/, Unmanned Systems Technology, 11 September 2018, Retrieved 3 October 2018
[24] "Unmanned Warrior", https://www.royalnavy.mod.uk/unmannedwarrior, Royal Navy Website, Retrieved 3 October 2018
[25] "Russia Begins Sea Trials of Nuclear-Capable ‘Poseidon’ Underwater Drone", https://thediplomat.com/2018/07/russia-begins-sea-trials-of-nuclear-capable-poseidon-underwater-prone/, The Diplomat, 21 July 2018, Retrieved 3 October 2018
[26] "Russia’s new weapons, nuclear parity and arms race: What’s going on?", https://www.youtube.com/watch?v=Q4LejOtYiyw, RT, YouTube, 26 March 2018, Retrieved 3 October 2018
[27] "NATO report highlights drone limitations in 'contested environments'", http://www.natowatch.org/newsbriefs/2014/nato-report-highlights-drone-limitations-contested-environments, NATO Watch, Retrieved 3 October 2018
[28] Axe, D., "Dark Sword: China's Mysterious (and 'Robotic') Stealth Fighter Has Arrived", The National Interest, https://nationalinterest.org/blog/the-buzz/dark-sword-chinas-mysterious-robotic-stealth-fighter-has-26175, 8 June 2018, Retrieved 3 October 2018
[29] Masters, J. "Targeted Killings", https://www.cfr.org/backgrounder/targeted-killings, Council on Foreign Relations, 23 May 2013, Retrieved 3 October 2018
[30] Adams, D., "Weaponized Satellites and the Cold War in Space", https://www.digitaltrends.com/cool-tech/weaponized-satellites-and-the-cold-war-in-space/, Digital Trends, 1 May 2018, Retrieved 3 October 2018
[31] Mindell, D. A., "Robotic exploration of Mars is equivalent to human presence on Mars.", http://www.slate.com/articles/technology/future_tense/2015/10/robotic_exploration_of_mars_is_equivalent_to_human_presence_on_mars.html, Slate, 23 October 2015, Retrieved 3 October 2018
[32] Gould, J., "US Space Corps could launch in 3 years, key lawmaker says", https://www.defensenews.com/space/2018/02/28/2021-a-space-odyssey-space-corps-could-launch-in-three-to-five-years-key-lawmaker-says/, Defense News, 28 February 2018, Retrieved 3 October 2018
[33] Gertz, B., "China’s Space Weapons Threaten US Satellites", https://freebeacon.com/national-security/chinas-space-weapons-threaten-us-satellites/, Washington Free Beacon, 26 February 2015, Retrieved 3 October 2018
[34] Erwin, S., "U.S. intelligence: Russia and China will have ‘operational’ anti-satellite weapons in a few years", https://spacenews.com/u-s-intelligence-russia-and-china-will-have-operational-anti-satellite-weapons-in-a-few-years/, SpaceNews.com, 14 February 2018, Retrieved 3 October 2018
[35] Stenger, R., "Scientist: Space weapons pose debris threat", https://web.archive.org/web/20120930100948/http://articles.cnn.com/2002-05-03/tech/orbit.debris_1_low-earth-orbits-space-junk-international-space-station?_s=PM:TECH, CNN, 3 May 2002, Retrieved 3 October 2018
[36] "Back to the Drawing Board – The Goliath Tracked Mine", https://www.military-history.org/articles/back-to-the-drawing-board.htm, Military History Monthly, 12 July 2012, Retrieved 3 October 2018
[37] Kube, C., "Russia has figured out how to jam U.S. drones in Syria, officials say", https://www.nbcnews.com/news/military/russia-has-figured-out-how-jam-u-s-drones-syria-n863931, NBC News, 10 April 2018, Retrieved 3 October 2018
[38] Mizokami, K., "Russia’s Tank Drone Performed Poorly in Syria", https://www.popularmechanics.com/military/weapons/a21602657/russias-tank-drone-performed-poorly-in-syria/, Popular Mechanics, 18 June 2018, Retrieved 3 October 2018
[39] " Pathways to Banning Fully Autonomous Weapons", https://www.un.org/disarmament/update/pathways-to-banning-fully-autonomous-weapons/, UNODA Website, 23 October 2017, Retrieved 3 October 2018
[40] "Russian Uran-9 unmanned combat vehicle tested in Syria", https://www.youtube.com/watch?v=0MiiUOjmqLo, Binkov's Battlegrounds, YouTube, 20 July 2018, Retrieved 3 October 2018
[41] LaPointe, C., and Levin, P. L., "Automated War", https://www.foreignaffairs.com/articles/2016-09-05/automated-war, Foreign Affairs, 5 September 2016, Retrieved 3 October 2018
[42] Lee, P., "Drones will soon decide who to kill", http://theconversation.com/drones-will-soon-decide-who-to-kill-94548, The Conversation, 11 April 2018, Retrieved 3 October 2018
[43] Dickson, B., "What is Narrow, General and Super Artificial Intelligence", https://bdtechtalks.com/2017/05/12/what-is-narrow-general-and-super-artificial-intelligence/, TechTalks, 12 May 2017, Retrieved 3 October 2018
[44] Allen, G. C., "Putin and Musk are right: Whoever masters AI will run the world", https://edition.cnn.com/2017/09/05/opinions/russia-weaponize-ai-opinion-allen/index.html, CNN, 5 September 2017, Retrieved 5 October 2018
[45] Gibbs, S. "AlphaZero AI beats champion chess program after teaching itself in four hours", https://www.theguardian.com/technology/2017/dec/07/alphazero-google-deepmind-ai-beats-champion-program-teaching-itself-to-play-four-hours, The Guardian, 7 December 2017, Retrieved 3 October 2018
[46] McConnell, M., "The AIs Are Winning: 5 Times When Computers Beat Humans", https://www.makeuseof.com/tag/ais-winning-5-times-computers-beat-humans/, MUO, 10 May 2016, Retrieved 3 October 2018
[47] "What Computer Games Get "Wrong" about War", https://www.youtube.com/watch?v=peNU5EffPYU, Military History Visualized, YouTube, 22 November 2016, Retrieved 3 October 2018
[48] Mclean, W., "Drones are cheap, soldiers are not: a cost-benefit analysis of war", https://theconversation.com/drones-are-cheap-soldiers-are-not-a-cost-benefit-analysis-of-war-27924, The Conversation, 26 June 2014, Retrieved 3 October 2018
