The Distant Future

In the TV show Star Trek: The Next Generation , Captain Picard relaxes by playing a computerized role-playing game in the holodeck, a place capable of temporarily creating (somewhat) solid matter, up to and including living things. While it's anyone 's guess whether we'll ever develop the holodeck, Picard's "holonovels" have a lot in common with games that already exist: simulated characters , a fantasy setting, and a plot whose progress is influenced by the player. They do contain a few improvements over current games , however. Picard can interact with the people in the story in all the same ways that he could with real people, he isn't perpetually stopping to look at his statistics, and he doesn't have to stuff everything he sees into his pockets in case it might be useful later.

In this section, we'll talk about some of the things that might make Picard's experience come true.

Automated Programming

Most game machines are single-CPU computers with dedicated graphics and sound-processing units. Their CPUs follow the traditional one-instruction-at-a-time model that, as we said earlier, was originally intended for computing ballistics tables for artillery guns. In the more distant future, it is bound to be replaced , although with what isn't at all clear. Neural networks hold a certain amount of promise for pattern recognition; so does massively parallel computing. Molecular transistors are a certainty ; quantum computing is probably not far behind. We might begin to see programming by evolution, using the principles of natural selection to create programs without human intervention. This has already been done to create simple electronic circuits, with successful but startlingly peculiar results.

It seems certain that eventually computers will learn to program themselves , although by that time they might bear no more resemblance to the machines we know today than the aircraft carrier does to the trireme ”and probably less. For that to happen, computers will need a fundamental understanding of the nature of information and the way that it can be stored and manipulated. In particular, they'll have to be able to make the abstractions that human programmers do all the time. Teaching a computer to program itself to calculate mathematical formulas probably won't be that hard because most mathematical formulas are calculated using fairly simple algorithms anyway. The bigger challenge is to teach a computer to model imaginary situations mathematically, to solve the "story problems" that we remember from algebra class. The hardest part of a "story problem" was not solving the algebraic equations, but determining how the situation described could be modeled mathematically in the first place.

Intelligent Design Tools

Automated programming doesn't mean, however, that game designers are doomed to extinction . Captain Picard's Dixon Hill stories are written by real people; they're just executed by a computer. People will always want to design games, and that's not going to change. What will change is the way we do the work. Rather than collaborating with a large team of programmers, artists , writers, and audio engineers , the game designer of the far future could be collaborating with a game-design tool, a program specifically constructed to assist in creating games. We can easily imagine a conversation that goes like this:

Designer: "We're going to want three kinds of aerial units: fighters, bombers, and transports."

Computer: "OK. Have you figured out their operational parameters and combat capabilities?"

Designer: "Not yet. But fighters are designed to attack anything in the air; they won't be able to attack ground targets. Bombers can attack both air and ground units. Transports have no weapons."

Computer: "Why would you ever want fighters if bombers have more functionality?"

Designer: "Because bombers will be slower and less maneuverable than fighters, and also more expensive. Compare the performance characteristics of the World War II B-17 with the P-51 Mustang. Also notice the manufacturing time and cost of each."

Computer: [ reads history for a nanosecond ] "Okay, I understand the principle. Do you want me to use those numbers as a baseline?"

Designer: "Sure, what the heck? It'll do for now."

Self-Adjusting Plots

One of the many weaknesses that computerized role-playing games suffer in comparison to tabletop RPGs is that their plots are fixed, designed in advance by the developer. As a player, you can ignore the plot if you like, but the game won't be very interesting and you won't go far. To see everything that the game has to offer, you have to meet the challenges that it presents . If you don't, nothing happens. Larger games typically offer several subplots to choose from, so you don't have to do everything in a strictly linear order; nevertheless, what you bought is what you get.

On the other hand, live role playing with a human dungeon master frees players to ignore the dungeon master's intended plot, wander around on their own, and explore areas of particular interest to them. Although this might be frustrating for the dungeon master ”particularly if he has spent a lot of time devising an adventure for the party, only to have them completely ignore it ”a good DM is capable of adjusting the game to suit the circumstances. The players are much more the masters of their own fates in live role playing. The DM also has the ability to adjust the nature of the challenge to suit the nature of the party ”if it is badly damaged, he can surreptitiously see to it that they don't meet anything capable of wiping them out in an instant.

This ability to devise new adventures on the fly and to adjust the difficulty of the game to match the abilities of the players is bound to appear in games in the future. At the moment, we have randomly generated adventures in games such as Diablo , but this really applies only to the layout of the rooms and the number of creatures in them. It doesn't change the personalities or dialogue of the nonplaying characters.

Artificial People

In 1950, the English mathematician Alan Turing proposed a famous test for determining whether computers are really thinking. Let a computer and a human each chat with another person, the Interrogator, in another room via teletype (today we would use instant messages). The Interrogator must ask them both questions and, on the basis of their responses, try to decide which is the human and which the computer. However, both are trying to persuade the Interrogator that they are human. If in a series of five-minute conversations the Interrogator correctly identifies the real human less than 70% of the time, the computer can be said to be thinking. ( Turing 's definition was slightly more complicated than this, but this is the generally accepted formulation.) Although it was largely abandoned as a serious goal of AI research, the Turing Test remains a popular informal standard for artificial intelligence.

In game development, we have a similar challenge, but the bar is far higher. Turing's test required only that the computer converse in typed text. We hope someday to be able to simulate credible artificial humans , computer-generated characters who look, act, and speak just like real people. Doing this successfully requires vast improvements in many different areas: graphics, animation, physics, simulation, and, of course, many kinds of artificial intelligence. A computer-generated person should be able not only to converse in natural language, but also take part in a political debate or interpret another person's mood just as well as an ordinary human would.

Computer games don't need to have artificial people, of course, any more than the telephone company has to have artificial operators. We can let human players play against one another, just as the phone company can hire real people to answer the phone. But artificial people are a key part of the "holodeck" fantasy and are probably its biggest challenge ”perhaps even bigger than creating the mechanics of the holodeck itself. It's still a worthwhile goal for game development, even if the solution is centuries off.