The Role Of Computers

Computers are ubiquitous in the Empire. Computing and display equipment is so nearly massless and free that nearly every object more complicated than a crowbar has at a minimum enough computing power and interface to display and search its operator's manual, and obtain upgrades from the network. Most tools are able to at least offer useful advice in realtime, if not actually operate themselves.

This really means all tools: not only do vehicles have autopilots, buildings climate control systems, and telephones integral name lookup, but kitchens produce food automatically, clothes adjust themselves to local temperature and fashion, and vacuum cleaners clean and dust houses while the owner's not at home. Most manual labor, in fact, is done by robots, although most such are fixed in place or part of another object, rather than the humanoid or miscellaneously mobile constructions of fiction.

Information not associated with a particular object is normally accessed through datapads. A datapad is no more than a few millimeters thick, depending on quality, and can be anything from a few centimeters wide to an entire wall of a room. A typical personal datapad is about 20x30cm, has audio and holo display and recording, weighs only a few grams, and has fusion fuel to last much longer than the mean time it takes for the unit to be lost or broken. Memory and processing are mostly delegated to larger fixed systems such as a building or ship.

Computers are capable of voice I/O as a matter of course, but the lack of privacy and speed and the susceptibility to poor environmental conditions make voice an unpopular choice. Pressure-sensitive handwriting surfaces or configurable keypads are more common. Motion sensors to pick up writing, typing, or gesturing in air are also used sometimes, although they are prone to misinterpreting fidgets.

Flat display surfaces with full color and 0.1mm resolution are thin as cloth or paint and almost as cheap; three-dimensional ("holo") displays are a bit thicker and noticeably more expensive, but still very common. Video pickups are about as thin as holo displays; mating the two with enough memory to store holo pictures at a reasonable frame rate results in a holocamera about a centimeter thick.

Networking of portable tools is entirely wireless, over frequencies from radio to IR depending on the situation. Optical cable is used only for major branches and trunks, or when security is essential.

More expensive but less obtrusive personal interfaces use eyewear or contact lenses to overlay the display on the user's vision of the real world, and take instructions subvocally, by air-typing or -signing, or by writing or typing on a sleeve or somesuch. Despite their virtues, such systems are strictly single-person, so tend to not be used much in ordinary life.

Nevertheless, although computers are used to advise and assist humans, and extend their capabilities, they are very seldom given the power to act autonomously. The reasons for this are manifold, and so rooted in centuries of history as to be unquestioned.

To an engineer, the primary reason is that the "computational ecosystem" is too complex to be fully trustworthy. Software has been learning, adapting, and evolving for centuries: the day when an engineer could be familiar with all the software involved in any nontrivial operation was over before the Republic, perhaps even before the Diaspora. Despite the general good reliability of modern software, there is no telling what inherited defects, both accidental and deliberate, might be waiting to bite. To the problems lurking in otherwise desirable programs must be added an equal number of centuries of proliferation of virii, worms, datamids, and their endlessly malcreative ilk. Computers have been routinely networked for so long that only the most isolated and expensively-maintained systems (such as military vehicle and weapon automation) are anywhere near free of such malign creatures, and even those are not perfectly immune.

The man in the street is unwilling to rely on computers because they aren't "as smart as real people". The failure of even the best and brightest of the Republic to make artificial intelligences has permanently stigmatized automation as inferior to human intelligence, period. Anyone will admit that a machine can have much faster reflexes, better senses, and greater precision of action than any human, but almost no one will accept that a machine's judgement can be as good as a human's.

Computers are also commonly viewed as being unacceptably vulnerable to criminal or malicious action. That they are vulnerable cannot be argued; given the ubiquity of computers, the incentive to figure out ways to break their security is strong enough to attract some very bright minds. As a result, it is true that no datastore or communications channel is perfectly secure: any algorithmic encryption can be broken in a sane amount of time with enough cleverness and computing power, and any precomputed-key encryption is vulnerable to key theft.

That humans are also vulnerable, and to even older and more highly-evolved tricks, is not usually taken into account, possibly because computers normally lack the ability to cover up or downplay their failures of security.

For some combination of all these reasons, computers that do anything potentially dangerous do it under human supervision. Though vehicles normally fly under autopilot, there must be a human pilot able to take control if necessary. Autofactories have human supervisors to shut them down if they go awry. Construction and loading robots operate only under human direction. Automated security systems can take no action more aggressive than locking doors and sounding alarms; even in the most secure facilities, intruders must be beat up and tossed out by human guards.

Obviously, this applies less to more constrained systems: while every aircar needs a pilot, a single remote supervisor can watch over several trains, and elevators can run unattended. It also applies less to military systems: although weapons are among the most dangerous devices to automate, missiles, mines, and point defense systems simply cannot be controlled directly by humans and still be effective.

This file was last modified at 1635 on 22Jun99 by