How to Create Artificial Intelligence - The Art of World Building
Jun 072021

An Artificial Intelligence (A.I.) has become a staple of SF, particularly in the form of a ship that can be controlled via the A.I., which typically gets a name for ease of reference. Let’s call ours Surance, which sounds female, and that is a decision to make. Is the gender fixed or can users change it? This arguably matters more in TV/film because every change of voice or appearance means another actor, plus a brief explanation, which can be as short as someone answering raised eyebrows from crewmates with a shrug and saying, “I changed her.” This change matters less in books because the reader won’t notice, which begs the question of why we’d do it.

And the answer is that Surance may have the ability to impersonate others, both visually and vocally, because it can be useful for the crew (and us). We live in a world where genders are not viewed and treated equally despite attempts to change this. What if our crew welcomes aboard a race, hostile to females, which takes offense at Surance? It’s both a storytelling decision and a character one for whether the crew changes the A.I. gender to be sensitive, takes a “too bad for you” attitude and leaves it as is, or splits the difference (leaving gender unchanged and explaining to and “educating” the visitors on their own cultural views). We can think of other scenarios for gender changing, with the caveat being that gender is theoretically irrelevant with a non-living being, but we all know that most of us look male or female. We can choose a neutral one to bypass the whole question.

We must also decide where the A.I. “lives.” These are often shown as part of a vessel or structure like a house, but there’s no reason Surance can’t be portable and as small as a pendant. Think of how we use Siri or Alexa devices and the possibility of having it everywhere if desired. Surance can manifest as a full body projection standing near a character (decide on the range of this), or maybe she’s a floating head, or just a voice, one that maybe only the wearer can hear (via a matching earpiece). If Surance appears, is it apparent that she’s an A.I. or can she seem real? Can she appear as more than one person at once? Imagine needing to give the impression you have more fighters by your side so that someone pursuing you backs off.

We should decide if there’s the equivalent of an internet or if all of the A.I.’s data store is local to them. How much info can they take with them? With the advances in memory storage on Earth, we can give them everything but the latest changes. But this may not be true in a universe unconnected to us. Then again, a society capable of producing some A.I.s must surely be able to create portable memory storage? They can likely connect to other systems unless we decide to restrict them.


Personality is another major area to invent, but we’ll do so in much like any other character. While we can make them human-like, we can also assign the A.I. another race and culture, which makes sense if an alien species designed the ship. This can make them less cooperative unless hacked and altered (with one degree of success or another). Unless Surance has been programmed to respect something like modesty or secrets, this can result in awkward situations, including being watched or listened to without permission. Decide how much of a conscience has been included.

An A.I.’s personality can be used as a counterpoint to those on the crew. Being more serious is likely, even a default. Think about it from a manufacturer’s point of view. You might like a sarcastic A.I., but plenty of people don’t. If I’m in the market for a ship and you’re a spaceship dealer, you could lose a sale over an obnoxious A.I. This certainly suggests that Surance is modifiable, possibly with aftermarket parts or programming. Consider this and the age of an item (a ship for example), not to mention its origin, to determine the status of the A.I.


An all-powerful A.I. is as problematic as anything else, maybe even more so. They’re designed to control an entire ship, house, or wherever they reside, deeply impacting living beings, even sustaining or killing them via environmental controls. It’s important to determine a way they can be turned off while leaving the item (let’s say a ship) still controllable, albeit with minimal systems? This has been done before but it’s an obvious plot device to have someone hack the A.I. or for it to suffer other problems. This is as realistic as a computer bug or car maintenance problems, so don’t avoid it. Just use it when it permits a new spin on the story. In our modern world, virus attacks are constant; they will be in a world with A.I.s, too.

Decide what tasks the A.I. is designed to do and to what degree. What can’t it do? What does it need living beings (or at least physical machines) to do for it? How many backups are there and how robust? A spaceship would likely have more than one, given its importance, but of course doing so is less fun for storytellers because restoration can happen quickly enough to ruin the tension introduced. An A.I. with less important functions is more likely to have one or even no backup.

Can Surance’s abilities be augmented? It’s a program after all. Maybe there’s a crew member tasked with maintenance and enhancements. As a professional software developer, I see tons of poor work and processes resulting in bugs. Cocky coders make mistakes, and the geniuses often shown in SF are likely worse. Prudent design requires teams of people, with a project manager and others in oversight roles, involved in any such effort. This would likely be true on a large vessel, so keep this in mind. Imagine living in a ship where everything about our life is on the line if someone codes a mistake into the A.I. The idea of a single coder is not believable in many situations.


Leave a Reply

%d bloggers like this: