(THIS ARTICLE IS MACHINE TRANSLATED by Google from Norwegian)
In December last year, the state parties to the UN Convention on the Prohibition of or Restrictions on the Use of Certain Conventional Weapons (CCW), including Norway, decided to initiate formal talks on autonomous weapons systems, after three years of informal expert meetings. The hope among many organizations, and an increasing number of countries, is that these talks may lead to an international ban on these types of weapons.
A fully autonomous weapon system is a weapon system that can select and attack targets without direct human intervention. Autonomy, even in weapon systems, is not a constant size, but a spectrum from none to full autonomy, and several weapons already have a degree of autonomy. One example is the South Korean cannon tower Samsung SGR-1, which guards the demilitarized zone between North and South Korea. The cannon tower is designed to detect an intruder and automatically shoot at it, but for the time being it needs human approval to fire. The most important question for CCW is whether or not a weapons system can make decisions to attack targets without human control over that decision.
The Pentagon's recent, successful test of a swarm of the microdrone Perdix is considered by many to be an important step on the path to the development of fully autonomous weapons systems. Designed to operate in a coordinated swarm and follow each other's movements and actions, the drones are capable of making collective decisions and repairing one another. The microns are no more than 15 inches long, but a drone swarm could overwhelm a military target's defense systems.
Dehumanization. There are many reasons why this type of weapon system will be problematic. For example, it is very difficult to imagine that an autonomous weapon will be able to respect international law. This means being able to distinguish between warring and civilian and understanding when a soldier is injured or surrendering, and requires the ability to interpret body language, facial expressions and tone of voice in complex contexts. In addition, such a weapon system must be able to assess whether the military advantage gained by an attack can justify the expected damage to civilians caused by the attack. A weapons system is unlikely to be able to distinguish between legitimate and illegitimate targets, especially in modern conflicts where parties operate without uniformity and with hidden identity. The principle of international law requires that the attacker must be able to assess the impact of an attack on all persons, units and objects in the area around the target. An assessment of military advantage is difficult to program, partly because it is highly context-dependent and must be decided on the basis of each situation.
Although an autonomous weapons system against conjecture should be able to make good assessments of what constitutes military advantage at all times, as well as possible damage to illegitimate targets, it is unlikely that it will be able to balance these two against each other. This type of analysis depends on the kind of judgment it is difficult to imagine programmed in a machine.
Should an autonomous weapon be used and a violation of international law or international human rights, it is unclear whether anyone will be legally liable for such an abuse.
It is also an important moral question as to whether it is okay to delegate decisions to bring weapons and machinery to life. A machine cannot make moral considerations related to a decision to take a life, and lacks pity, empathy and understanding and respect for human dignity. Many, including UN Special Rapporteur on extrajudicial, summary or arbitrary executions Christof Heyns, have expressed concern that the use of this type of weaponry will result in further dehumanization of warfare.
It is hard to imagine that an autonomous weapon will be able to respect international law.
Unpredictable. Many countries have developed – and are working to develop – weapons systems with varying degrees of autonomy. Long Range Anti Ship Missile, developed by Lockheed Martin, can autonomously fly to a predefined area, avoid obstacles that appear on the road, and then select and attack one of several ships at the destination. The Israeli Harpy drone can patrol a defined airspace for long periods of time in search of radar signals, and then autonomously attack the source of the signal. The British Taranis drone is being developed for surveillance and attack operations under human control, but BAE Systems, which is building the drone, claims they are also developing some fully autonomous features. The Chinese Sharp Claw vehicle can be armed and operated autonomously, while Russia's URP-01G tank and robot soldier Iron Man are other examples of such weapons.
The arguments for allowing these types of weapons are often cited for military benefits – including a high degree of precision, fast response time, lower costs than human soldiers, and the ability to operate without established lines of communication.
However, some of the factors highlighted as military benefits may pose a threat. A very low reaction time can prevent people from being able to see and control, the actions of a weapon system, and a conflict or threat situation can quickly get out of control and escalate as a result of the actions and counter-reactions of such systems. An autonomous weapon, especially one involved in the fight against another autonomous weapon, will have to adapt and change its behavior quickly to avoid losses as a result of predictability. This is a necessary adaptability, which also creates unpredictability and uncertainty, and it is expected that complex systems with a high degree of autonomy will behave in unforeseen ways. Especially systems with the ability to self-learn may develop in ways that are impossible to predict, and two types of systems on different sides of a conflict may interact with each other in unexpected ways. In addition, a software update can change the entire strategic balance in minutes. There is also a danger that the development of such systems will lead to an armor race, with further negative consequences for stability and voltage level. The technology is likely to spread quickly from more technologically advanced countries, both to other countries and to non-governmental armed groups and terrorist organizations.
Take the blade from your mouth. An international ban will provide standardized international rules with less scope for interpretation, and – by also prohibiting development and production – may limit the spread to states and groups that do not care about international law. An explicit ban will also help to stigmatize the use of such weapons systems even among states that do not ratify a possible ban.
So far, Norway has been cautious about commenting on an international ban. Now it is time to take the blade from your mouth. Autonomy is a technology that has many positive uses, but which can also be combined with virtually any weapon system. The world is not served by an autonomous weapon of mass destruction. It is short-term and very little constructive if Norway does not actively participate in this process and contributes to progress in international efforts for disarmament and arms control.