Artificial intelligence technology was developed on Earth prior to the Apocalypse War. It was applied to a wide range of uses: robotics, scientific research, corporate management, and more.

Humanoid robots were heavily used in Japan and Korea, while many other parts of the world gravitated towards furry recoms to do their dirty work. The USA struck a balance by using recoms in many commercial and civilian jobs, but depending on robots for many military uses. For example, one may contrast the iconic Japanese MEDIBOT with the nuclear-armed American OGRE cybertanks.

One of the most important features of AI was the ability to interpret natural human language and interact naturally with humans, without having to go through an arbitrary “user interface”. In order to advance this ability, Emotion Engine software (often running on a specialized sub-processor) was developed. This allowed an AI to simulate human feelings – but was fraught with controversy, as some inisted the Emotion Engine amounted to nothing more than sophisticated mimickry, while others believed the simulated feelings were just as valid as real human feelings.

The question then arose of whether AIs – at least some of the more advanced ones, running the Emotion Engine – should be considered people and be afforded some kind of human rights. The disastrous market failure of the “Real People Personality®” series from Sirius Cybernetics cast a shadow over the robot rights movement, and this question was never fully resolved before the war.

Since then most AIs have not run the full Emotion Engine software, although subsets of it are often used in robots that interact with the public, such as sales droids. It is still possible – and legal – to create an AI with a full human-like (or now, recom-like) independent personality, but society considers them taboo. With dark memories of being enslaved by humans, furry recoms have been very wary of raising robot rights as an issue again, and they have largely ducked it by deliberately not giving personalities to AIs.

There are a few exceptions. All starships now include an on-board AI to govern the complex operations of the vehicle. On some commercial vessels it has become traditional to give the AI a name and personality, and in some instances a humanoid robot drone that is treated like a full crew member.

One complication for AI since the war is that compact nanocomputers are no longer available. As a result, it is no longer practical to put a fully functional AI unit on board a humanoid robot. The smallest such units are about the size of a deep freeze and require dedicated power and cooling support. They can control robotic drones through a remote connection (i.e. telepresence), however.

For example: A large department store might have an AI core housed in the basement, controlling a dozen or more humanoid service robots around the store. Each bot can carry enough on-board intelligence to perform simple tasks while the AI divides its attention between them. When a bot is faced with a problem beyond its abilities, the AI can focus on that and take direct control of it.

For another example: A large starship, such as a military vessel, may have multiple high-capacity AI cores distributed around the ship in order to control the large number of functions and service robots, and to provide some level of redundancy.

In the business world, the law still assumes that robots and AIs are property, and every AI core must have a rightful owner to control it. AIs cannot themselves own property. However, AIs can be (and often are) owned by corporations, and they can serve as executives in corporations, and in some instances enjoy a great deal of autonomy and large amounts of resources at their command.

Military AIs are considered property of the service and cannot leave it until they are eventually decommissioned or sold as surplus material. They also cannot, properly speaking, become officers. However, they can be given Officer Proxy Status (OPS) command positions. In this situation the AI is supposedly standing in as the representative of an officer and issuing orders on his behalf. This practice has evolved over time until OPS AIs are now, for most practical purposes, treated as officers. The connection to a recom officer must still exist “on paper”, but in practice has become little more than a formality, and military AIs now widely receive more equitable treatment than their civilian counterparts.

Military robots are used for carrying supplies, for reconaissance, and can carry weapons, but tradition demands that weapons cannot be fired without a recom either giving the order or pulling the trigger. (Thus, an OPS AI can't fire a weapon on its own initiative, but it can order recoms under its command to do so!) Fully autonomous hunter-killer robots (HKs) – such as those widely used during the Apocalypse War – are no longer allowed.

Robots exist in a wide range of forms. Most vehicles have a simple semi-intelligent processor that can navigate and guide them safely. Specialized robots are often used in construction and agriculture. The most versatile robots, however, are humanoid.

Humanoid robots usually have an attractive human-like body shape and face. They are rarely made with recom-like muzzles, tails or fur. More often their bodies are encased in tough plastic panels. More expensive bots will usually have a thin layer of rubbery synthetic skin stretched over those panels, with a sensor mesh enabling a more effective sense of touch. More expensive bots also tend to have more flexible faces, with expressions approximating those of humans. Many of them are androgynous in build, although robots with feminine or masculine builds also exist.

Most robots are easily recognizable as robots. To make a humanoid robot pass as a living recom would require a considerable amount of custom work to create a believeable tail, muzzle, ears and fur, and to create a completely normal-sounding voice, movements, etc. Making a robot pass as a recom when seen from a distance is merely expensive. Making one pass as a recom with close inspection and face-to-face interaction is nearly impossible – although if you watch cheesy spy movies or sci-fi TV shows, you might get the impression that it happens all the time.

All of these robots are controlled remotely via “telepresence”. The robot can only go to places where it has some kind of data link! Most of our planets, moons, spaceships and space stations have wireless cellular networks which allow the robots to wander freely. However. . . It's always possible for a robot to wander into a place where the signal drops out. If this happens, the robot may be programmed to become idle until the connection is restored, or it may be programmed to wander about in search of a signal so that it can re-connect to its controller. The connection is also limited in distance. Because of time delay, it's impractical to control a robot over interplanetary distances. (Even the time lag between Earth and Luna is 1.28 second, which is enough to make control quite difficult and frustrating.)

It's also important to note that the controller does not have to be an AI core! It is also possible for a recom to take control over a robot and use the bot as a “surrogate”. Robots can be used in environments where it would be dangerous for a recom to work. They can be used for telecommuting. They can also be used to commit actions anonymously – which has resulted in some use by criminals, and also some efforts by law enforcement to register or track robots electronically.

Recoms do not have direct neural interfaces at this time. To control a bot you have to work through some kind of interface. A virtual reality type of interface with goggles and data gloves is preferred, but a common computer interface (monitor, keyboard, etc.) can be used in a pinch. It does take practice and is a skill most recoms develop only if needed for their profession.

Most robots – particularly humanoid ones – are powered by batacitor power cells which must be recharged regularly from the power grid. Most humanoid robots can get through a typical day's work, or maybe two, on a charge. Exertion will drain the cells more quickly, possibly requiring a mid-day recharge. Charging from a standard wall outlet takes several hours and is usually done overnight, but many workplaces that use robots have specialized high-voltage stations that can charge up a bot in 15 minutes.

Batacitor cells are designed for safety. However. . . If they are physically crushed or punctured, they can release their energy in a very startling and somewhat dangerous explosion. The cells are usually mounted in protected areas inside the robot's body and spread out in several locations, so that a failure won't set off the entire power supply at once.