Robot Brains: The Skill Budget Model

In Gurps, perhaps the simplest way to give AI's skills to give them skills for being functioning minds. Which is to say, just like any other character. The skills are an intrinsic part of who they are. They can learn new skills the same way any other mind does, by studying, practicing, and through hands on experience. They can no more download a new skill than a human can purchase one in the form of a book. 

 

This is probably one of the simplest ways to think about AI skills, but there are a few important implications. One is that we have to think explicitly about how many skill points we are giving our NPC's. Most of the time, GM's just wing that or balance the points spent on skills with a budget for the entire character. Skills as a point budget can also say something deep about the nature of intelligence in our settings. Its makes robot minds similar to human minds in many ways, and it sticks limits on technological singularities. Most of the time, this is good: singularities are by definition hard to come to grips with, and thus hard to run a game in. But we should look at some of the implications.

Implications

In the Skill Budget Model, skills can't be separated from the AI itself. They are an integral part of it, much like its ability to speak or to reason. They can't download new skills, and they can't upload their hard-earned skills to others. This makes minds very "Human" and treats intelligence as a sort of "black box" that is difficult to manipulate.

AI's using this scheme can still participate in "skill for cash" schemes, but this looks like recruiting rather than equipment upgrade. The skill comes attached to a character with their own motivation algorithms, operating hardware, and independence. And that has been available to civilization for a while.

 

While I have been emphasizing the just like "humans", aspect of these situations, the ability to back-up and copy Artificial Intelligence can still result in some fairly heavy flirting with technological singularity.

Setting the Budget

Three things should determine an AI's point budget: Its complexity, its AI Type, and its effective experience and training. I say "effective" because with Artificial intelligence you may be able to copy skills from one AI to another, simulate training, and use other ultra tech tricks.

 

Gurps has different types of AI, and generally, the more independence an AI shows, the more points it should have available for skills. Oddly, while more independent AI's are more likely to have a lot of skill points, more narrow AI's are likely to dump them all in just a few skills, so the amount spent in each skill may be roughly equal.

 

AI's with higher complexity should be able to learn more skills. This probably shouldn't scale with the number of processors or amount of memory though. I suggest having it scale linearly with complexity, or perhaps with the square of complexity if we're feeling generous to high complexity machines.

 

If we want a basic suggestion on the number of skill points to work with, we can use:

 

Dedicated AI:	complexity skill points
Non-Volitional AI:	2 * complexity skill points
Volitional AI:	4 * complexity skill points

 

These numbers, of course, are just a suggestion. Everyone plays GURPS differently, and Different games and groups may have more or less points for NPC's in general. And NPC skill levels are the relevant comparison. Even if we have decided that AI is universally more talented than humans, we still need to think about how much better we want them to be, rather trying to calculate numbers.

 

All members of a given model of robot may all start out with the same factory-created set of skills, or they may be somewhat configurable. Shape the budget to be what the game requires: these numbers are a tool to help figure out what a root looks like, not statement of what robot minds must have. It may even be that different makes of robot with the same complexity have differing point budgets.

A core part of the holy grail of "true AI" is the ability to it to learn, so experience (or its equivalent) should matter for artificial intelligence. Robots with more experience should have more skills. This doesn't always mean age. It also can include training and the sorts of high impact events Player Characters thrive in. The type of AI should effect learning rate as well. Some hidebound AI's may get a budget of skills when they are first created and make do with just those. Others may learn a lot early on, then sort of peter out. Don't feel the need to calculate a robot's age to figure out its skills. Plateaus are common not only in human learning but in many other phenomena, and its fairly likely that artificial minds will run into limiting factors, just like most of the universe.

When To Use The Skill Budget Model

The skill budget model works well when considering an AI as a Player Character: it advances and learns just like everyone else does. Of course, in that case the budget is likely to be implicit in character building rather than explicit. The skill budget model works well for AI's that are made for a specific purpose, or "grown" by running them through a training regimen. It also works well when you want to explore the human aspects of AI rather than the Inhuman aspects. I hope you enjoy using this method!