We have already seen how artificial intelligence has won chess … Obviously, some have become competent in much less harmless “playing areas”. Chinese scientists have announced that an artificial intelligence drone has destroyed a remote-controlled human device in record time.
The information is taken from an article in the South China Morning Post and has already spread like wildfire. The Chinese military researchers behind this experiment are working on drones capable of dogfighting not only without a pilot… but without a human. Thus, these scientists claimed that for the first time, an unmanned fighter aircraft, fully piloted by AI, defeated human pilots in a real dogfight at close range. They add that this performance would have been achieved in a very short time, as it only took the drone 90 seconds to destroy its target.
A research paper detailing the experiment was published last week in Acta Aeronautica and Astronautica Sinica. The team was led by Professor Huang Juntao of the Chinese Army Aerodynamics Research and Development Center in Sichuan, China. If this drone can be so efficient, it’s because of its computing power. This allows him to analyze the course of the battle and accurately anticipate its development in order to take appropriate initiatives.
A program capable of analyzing and combining field data
Recall that nowadays what is called “artificial intelligence” can often be described as a machine learning program. The program is supplied with a large amount of data related to the field of knowledge for which it is intended. There are artificial intelligences designed to create images, texts, music, voices, AI acting as opponents in games, or others capable of generating computer code … In short, many areas are now affected by AI. In this particular case, of course, a certain versatility was needed, because the analysis of the air battlefield requires the integration of a large amount of information: the location of opponents and their movements, obstacles, weather conditions …
To conduct the test under optimal conditions, the researchers used two nearly identical drones. The only difference, of course, is that one of them was controlled remotely by a human, while the other had an AI on board. They claim that some improvements still need to be made. However, for them, this mock combat proved that AI piloting technology is viable in real dogfights. They even argue that aircraft with autonomous decision-making capabilities could eventually completely out-compete humans, especially in reaction speed.
The Chinese army is clearly not the only one interested in introducing this technology into its arsenal. Recently, the US allocated at least $78 billion to the AMASS (Autonomous Multi domain Adaptive Swarm of Swarms) project. The idea is to deploy thousands of drones, air, ground and even underwater units to take control of the area. This is not even about swarms of drones, but about “swarms of swarms” capable of conducting coordinated and automated attacks.
All of these advances applied to combat drones are cause for concern. In the case of AMASS, operators must control the entire operation. Even then, experts questioned the controllability of these weapons: this was pointed out to the press by Zachary Kallenborn, a researcher in the department of non-conventional weapons and technologies of the National Consortium for the Study of Terrorism and Countering Terrorism (START). According to him, on the scale of thousands of drones, human control will be more difficult, which breeds fragility. In the case of drones designed to operate entirely without a human operator, as is the case in this experiment, the issue of AI reliability and risk may be even more relevant.