In this article I will prove that you need Preysight in your autonomous weapon.
Let’s assume that you want to build an autonomous weapon. To build it, you must first know what you’re building. According to the United States Department of Defense, an autonomous weapon is one that “once activated, can select and engage targets without further intervention by a human operator.” To select a target the weapon first needs to detect the target. I won’t belabour this point. Perhaps to the reader’s dissatisfaction, I will prove that the weapon needs to sense targets by lack of counterexample.
Now that we’ve firmly established that the weapon needs to detect targets, if we want to build an effective weapon, we must find the most effective way to detect a target. So, how does one detect a target at all? All information processing systems in the universe are physical, and therefore to change their states, a physical change must occur. For a physical change to occur, force must be exerted on the system’s particles, and therefore there must be an exchange of energy between the target and the weapon.
Any exchange of energy must be mediated by a force-carrying particle. There are many such particles; a tennis ball will suffice, for now. Imagine the weapon suspended in space in close proximity to the target. The weapon could launch tennis balls in various directions. Eventually, a ball would collide with the target and hopefully be reflected back to the weapon. Thereby, the weapon could interpret a returned tennis ball as evidence of the presence of a target.
The tennis ball detector has a number of drawbacks. The balls are heavy relative to many other particles, meaning that if the weapon needs them to travel quickly, it must impart them with a significant amount of kinetic energy. Not only is this energy costly, but the ball could end up damaging any object that it collides with. Remember, we are only using the ball as a detector. The mass of the ball would also be lost forever unless the ball is retrieved, which would happen infrequently at best suspended in space. The balls are voluminous relative to many other particles, and since the minimum resolution of the weapon’s detector would be about as large as the balls themselves, its vision would be quite blurry. These problems compel us to seek particles other than tennis balls, even without considering factors like gravity and air resistance.
All the problems with the balls stem from the same cause: the balls are too large. However, let’s assume that we must use these tennis balls and make the most out of them. One problem that we can solve is the loss of mass. Consider if instead of launching the balls into space, we stacked them on top of each other and glued them together. We’d have made a pole out of tennis balls. With this detector, the weapon could poke around into space. If it felt something push back, that means it found the target. What’s happening now is that we are no longer using particles to transfer energy-information. Instead, we are transferring a wave of energy through the balls. If that wave is reflected and returns to the weapon, that means it has found something. Of course, waving around a pole in space brings its own set of challenges, and is probably not feasible.
Let’s bring things down to Earth and add a bit of realism to our scenario. Instead of being suspended in the void, imagine that the weapon and target are suspended in a sea of tennis balls. Now there are effectively an infinite number of ‘poles’ in every direction at every point. By agitating this sea, the weapon can generate waves of energy. To reduce their energy the quickest, these energy waves will necessarily spread out, which reduces the range of the detector but increases its breadth. By detecting the reflection and return of these waves, the weapon can detect its target. The reader might still believe that this sea of tennis balls is unrealistic. While it is true that outside of a tennis ball factory, we can’t find such a volume, the molecular composition of the matter is immaterial, and this analysis applies to any volume with a non-zero mass density. So, reinterpreting the tennis balls as mass density, and the energy waves as acoustic waves, we’ve reinvented sonar from physical first principles.
Casting our minds back to space, we know that we can’t rely on sonar in empty space. The lack of mass density in the volume prohibits it. So, are we back to launching tennis balls? We would be, if there weren’t the presence of another ‘sea’, for empty space is not truly empty. Throughout all of space there exists a physical field called the electromagnetic field. Actually, we’ve been secretly using it this whole time. The collisions between massive particles such as tennis balls are ultimately due to electromagnetic interactions between their protons and electrons. But the electromagnetic field can carry energy without the presence of massive particles. By agitating the electric charge distribution in a volume, electromagnetic waves are generated, just like acoustic waves. These waves offer several advantages.
The main advantage of EM waves is their ability to self-propagate without a massive medium, since the EM field is everywhere. We’ll return to this property shortly. EM waves are also very fast; since they lack mass, they travel at or close to the speed of light. Indeed, light is an EM wave. That speed means our weapon can detect targets about as quickly as relativity allows. EM waves can be generated with very small wavelengths, allowing a much finer resolution of detection than is possible with acoustic waves. Since EM waves interact with any charged particle, and all relevant massive particles have non-zero charge, our waves will be reflected by the constituent particles of the target. That allows the weapon to detect any target.
Let’s pause our discussion to briefly compare acoustic and EM waves. Any wave traveling through a medium can be absorbed by the constituent particles of the medium and lose energy, with the energy converted to heat. This is called attenuation, and is a problem, since we lose the information stored in the energy of the waves. In empty space, acoustic waves cannot exist, and EM waves cannot be attenuated, since there is a lack of massive, charged particles to carry the acoustic wave and to absorb the EM wave’s energy. In a massive medium, the EM wave will be attenuated, with longer wavelengths generally attenuated more slowly than shorter ones.
Sailing our minds back to the sea of tennis balls, let’s instead imagine that our weapon and target are in an actual sea. We recognize this as submarine warfare, and submarines are famous for their reliance on sonar. This reliance is due to attenuation. While EM waves can propagate through water, they are quickly attenuated, whereas acoustic waves are more slowly attenuated, and are actually faster in water than in air. The effect is even more dramatic when our weapon and target and embedded in the hot rock of the Earth’s mantle. While some powerful ground-penetrating radar can be used for subterranean sensing, seismic (acoustic) waves can travel the entire diameter of the Earth. In fact, they can be reflected by the crust and travel this distance several times before attenuating. Luckily for our poor weapon and target, the subterranean battlefield is not one in which they are likely to be deployed. The point is that the less mass-dense the battlefield, the more suitable EM waves are for detection.
On the surface of the Earth under the cover of the sky, where armies have clashed since the dawn of mankind, the atmosphere allows for both acoustic and EM waves to be useful. When the sun is shining a huge amount of EM radiation is bathing the Earth. Some of this radiation is absorbed and some is transmitted. But much of it is reflected. This allows for a key advantage of EM radiation: it doesn’t need to be generated by the weapon to detect the target. This not only saves energy but allows the weapon to hide better and not be targeted itself. There exists no comparable source of noise to the sun for acoustic waves. However, not all radiation comes from the sun. Blackbody radiation is generated by all matter at a non-zero temperature, that is to say, all matter. At the temperatures of the surface of the Earth, this peaks in the thermal spectrum, and is involuntary. Similarly, moving objects generate sound involuntarily. We will return to this subject momentarily.
On the surface of the Earth, objects bask under the light of the Sun. This light is reflected, absorbed and transmitted, and the reflected light can be passively detected by our weapon. The weapon can theoretically detect the entire EM spectrum, but some spectra are more useful than others. High energy solar spectra like the gamma and X-ray ranges are too quickly attenuated to be reflected. Low energy solar radio and microwave spectra suffer the opposite problem, passing through objects too easily to be significantly reflected. However, active use of the low energy spectra is called radar and is very useful for long-range detection of targets in the sky. These spectra suffer from their inherent lack of resolution, making it hard to distinguish small objects clustered close together. Radar emitters and detectors are also typically very large and expensive. So, despite the fact that their operation is active and thereby exposes their positions, they are expensive and slow to move. Useful in many situations, but an alternative is needed, which necessitates an alternative spectrum.
There are many reasons that the visible spectrum was naturally selected by evolution to be the spectrum of ocular detection. The visible spectrum has a high enough energy to interact with the valence electrons involved in chemical bonds, without causing damage. Higher energy UV radiation can break those bonds or, dangerously, ionize atoms. Lower energy IR radiation will be discussed below. Photosynthetic plants and solar cells both absorb the visible spectrum to generate energy because it interacts so well with valence bonds. Our eyes use a chemical called retinal to convert the energy of light into information for our brains. Our weapon can similarly use this supremely useful spectrum of EM radiation to detect the target, using a normal camera. By no coincidence, the global supply chain of normal cameras is robust. A camera gives the weapon a reliable, passive, inexpensive way to detect the target.
Blackbody radiation is generated by all objects (including black holes, as Hawking radiation) due to their temperatures. At terrestrial temperatures blackbody radiation peaks in the IR spectrum. By detecting IR radiation, the weapon can detect the temperatures of surrounding objects. Temperature is increased by adding heat to an object. On the surface of the Earth, substantially all heat comes from sunlight. That means that most objects on the Earth are at relatively uniform temperatures. This depends significantly on their material composition. For example, water can be much colder than the ground. Tree leaves are cooled by the plant due to their vascularity and surface area. Asphalt is hotter than dirt. But these differences are repeated uniformly over large areas. Very infrequently are small hot spots encountered on the surface of Earth.
When objects generate their own heat, they will elevate their temperature above that of their surroundings. There are many ways to generate heat, but they all involve the consumption of free energy to perform work. Animals metabolize nutrients, vehicles burn fuel, as do fires. Electrical equipment is heated by Joule heating. Vents emit hot gas from their interiors. All these objects will typically glow at a substantially higher IR frequency than their surroundings. This glow is difficult to obscure. Thus, a weapon with an IR camera can detect hot objects and substantially narrow down the possible identities of those objects. This is especially useful at night, when the surroundings are colder and visible light is unavailable. The downside is that IR radiation requires more expensive sensors to detect and is not good at distinguishing objects at similar temperatures.
Sight is the ability to see, to detect patterns of light. The problem of mechanical sight was solved decades ago with the first cameras. Vision is the ability to analyze sight to understand the world. Vision allows our weapon to distinguish the target from the background, the signal from the noise. Without vision, the weapon will never be truly autonomous, no matter how sophisticated its eyes are. The problem of computer vision was solved far more recently with advances in AI, and it is a problem that is still being solved and refined to this day. As shown above, detecting visible and IR light with cameras is the best way for a weapon to detect a target. For large, expensive systems like submarines and air defences, sonar, radar, or other detection methods may be superior. But for the mass numbers of inexpensive robots designed for direct, frontline combat, computer vision is the best option.
Psycraft develops Preysight, a suite of computer vision AI that can bring light to the eyes of your autonomous weapon. We develop a wide range of models, and the software to deploy them on robotic hardware. Our ever-widening array of targets covers threats faced on the battlefield and will evolve with those threats. We have released AI that analyzes the visual spectrum and will soon pair that with AI for the infrared spectrum, and AI for sound. These tools will allow global customers to bring sight to the dead eyes of their robots. Only with this vision will they be able to create truly autonomous weapons systems. Only with this will they be able to send legions of robots off to war – and keep human beings with lives worth living safe at home