China’s satellite hunter AI may spark dangerous precedent for a flash war in space
A Chinese research team has reported an experiment in which it says an anti-satellite AI learned to successfully trick its target in a simulated space battle.
In a paper published on April 25 in the domestic peer-reviewed journal Aerospace Shanghai, Dang Zhaohui, professor of astronautics from Northwestern Polytechnical University, and his colleagues conducted an experiment in which an AI commanded three small hunter satellites to capture a high-value target, repeating the exercise thousands of times.
The researchers also set penalty parameters for the hunter satellites, such as consuming more fuel and colliding with a teammate. In contrast, the target satellite gained points for each penalty incurred by hunter satellites.
Both sides performed poorly in the first 10,000 rounds. However, according to Dang, the hunter satellites learned faster – as they were working as a team – and after 20,000 rounds they secured an advantageous position. But the target satellite gradually learned its pursuers’ simple tactics and became better at avoiding capture, firing boosters to evade the hunter satellites.
However, after 220,000 rounds the hunting AI reversed the situation and employed more complex tactics such as collaboration, forward planning and deception, which substantially increased the chance of capture. The target satellite was trapped when the AI ordered the hunter satellites to veer off their original trajectory, after which one of them suddenly changed course and deployed a capturing device from 10 meters away.
Dang also noted that past studies treated the pursuit problem as a mathematical optimization problem, under the assumption that a large target in orbit is slow, stupid and blind. He also mentioned that humans would have no role in this space war, as AI will control both the hunters and targets.
However, some experts cautioned of the impact AI has on space. According to Cai Cuihong, an associate professor of international relations at Fudan University, “the application of artificial intelligence in space will have a disruptive impact on global strategic stability.”
Cai explained that “AI can make anti-satellite measures more precise, destructive and harder to trace, increasing the likelihood that some countries will carry out a ‘pre-emptive’ strike.” She also added that “attacking satellites, especially early warning satellites, is often seen as a precursor to nuclear war.”
InMarch, Chinese researchers called for the development of anti-satellite capabilities against Elon Musk’s SpaceX’s Starlink satellite internet constellation, citing the broadband system’s potential military applications and threat to China’s national security. The researchers stated that a combination of soft and hard kill methods should be adopted to make some Starlink satellites lose their functions and destroy the constellation’s operating system.
They also suggested that military payloads could be launched alongside Starlink satellites, which would require China to upgrade its existing space surveillance systems to take super-high-resolution photos to identify any unusual features on the satellites.
In the same month, Chinese researchers unveiled an advanced AI system that enables low-cost commercial imaging satellites to become potent spy platforms. This advanced AI could potentially upgrade the capabilities of China’s commercial imaging satellites to monitor Starlink satellites for new features indicating potential military use.
In April, the US Defense Advanced Research Projects Agency (DARPA) awarded multiple contracts to General Atomics, Lockheed Martin and Blue Origin to develop a satellite nuclear propulsion system for low-Earth orbit demonstrations in 2025 as part of its Demonstration Rocket for Agile Cislunar Operations (DRACO) program.
Nuclear propulsion would give US satellites effective maneuver warfare capabilities in space, increasing their flexibility for both offensive and defensive operations, and increasing their survivability against Chinese or Russian anti-satellite weapons.
China has also been developing directed-energy weapons such as ultra-high-power microwaves and solid-state lasers for use on its satellites. These weapons use concentrated electromagnetic rather than kinetic energy to incapacitate, damage, disable or destroy enemy equipment, facilities, and/or personnel. They also have several desirable attributes, such as instantaneous engagement, high shot capacity limited only by electrical power supply, high difficulty of intercept, precision targeting and low cost per shot.
Chinese researchers have developed a new type of Relativistic Klystron Amplifier (RKA), a high-power microwave (HPM) source that can be used as a source of high-powered microwaves to burn out the electronics of enemy satellites, and as a defensive asset to absorb high-power microwaves before they can cause damage.
Another such weapon is a satellite-mountable solid-state laser capable of generating a megawatt of laser light and fire 100 times a second for half an hour without overheating in space. Such a laser is capable of dazzling satellite cameras or even permanently blinding satellites. Upon testing, the device reportedly generated a beam lasting only 5 nanoseconds, but it was powerful enough to permanently blind human beings or vaporize target surfaces.
But the real danger of space-based autonomous weapons may lie in the risk of “flash wars,” involving autonomously attacking and self-replicating cyber weapons. This speeding-up of operations could also lead to undesired chain reactions, when one autonomous system reacts with another in an uncontrolled chain reaction of escalation.