By Mark G. McLaughlin
The gaming industry was the first to make use of artificial intelligence (AI). Early arcade games from “Missile Command” to “Space Invaders” used a rudimentary and fairly predictable AI, but from such humble beginnings evolved into NPCs (non-player characters) in role-playing games and an AI that could run entire empires in “Civilization” and other strategy games. Of course, the strongest AI of any game is still that which is capable of competing with grandmasters in chess. While the AI in games has progressed exponentially over the last three or four decades, it is advancing even more rapidly now, with AI characters and programs that learn from the people they play, and who now no longer merely follow a pattern or a decision tree, but which can truly think for themselves. Here are just a few of the cool ways that AI is transforming gaming.
Evolving from barely intelligent to almost intelligent
In even the most advanced and complex video, computer an arcade games, the AI is still, at best, barely intelligent. It can be programmed to evaluate the possibility of success or failure, or make a move or hit a target that most hurts the human player or best enhances its own status, power or other attributes, but in most games the AI is still essentially just a mathematical program playing the odds. That is changing, as programmers are working to allow AI opponents to adapt to and learn from each game, and from players who do not play in predictable, logical patterns, but assess and manage risk. Most game companies, however, are still comfortable with an AI that is good enough to give a good game, rather than one that can truly challenge and even decisively defeat human opponents.
Convergence: From mimicking players to making decisions.
Most AI programs seek to mimic what players do, in order to give human players the feel that they are competing with someone like themselves. These programs seek to behave like humans, but do not really think for themselves or even learn. They give a good game, but often because the scales have been weighted to give them an advantage in some areas to make up for what they lack in the ability to make creative and intelligent decisions. They are, thus, more brawn than brain. Giving an AI a bigger and more advanced brain, and one that is capable of taking a creative approach rather than following or choosing from a scripted path or choice of paths, is the goal of many AI programmers – and they are getting closer to that goal. Those taking that path work with a theory of convergence that they believe will produce the same kind of AI in video and computer games that chess players can now confront in AI chess opponents – if they dare.
Hierarchical Task Network: if it’s good enough for the Army, it’s good enough for gamers
Hierarchical Task Network, or HTN, is a system that starts with simple tasks and builds upon them layer by layer until the AI is making truly complex decisions. It breaks down decisions into many and more simpler components and works from the bottom up to reach a decision. Reaching one goal is a precondition for reaching the next. This type of programming is used by Department of Defense computers to solve complex logistical problems. Some game programmers believe this path, rather than the convergence path, is the way to creating a gaming opponent who can not only make better decisions, but also autonomously create new challenges beyond what the game designers have programmed.
Voice recognition and interaction
Voice recognition and response software is already a mainstay in many industries, especially when it comes to directing callers to the right department or in paying their bills. Computer games have been slow to implement voice recognition and response techniques, other than to provide a computer opponent with a selection of stock phrases with which to answer a narrow list of questions. Computers speak, but they do not create speech – they only parrot what they have been told to say. It will take a great leap in computer learning technology for AI opponents to do more than just choose from a modular menu to create a sentence. Games that understand speech, use pattern recognition and reinforcement learning algorithms are not on the horizon – but are perhaps just over it, on the far side of the hill. The day will come – and soon – when that computer AI will be something with which you can have a two-way conversation, either as an opponent or as a partner who understands what you need it to do to win.
Virtual Reality, Augmented Reality and AI – the Trifecta challenge, or making the holodeck real
Game companies are vying with one another to make the best use of virtual reality technology, augmented reality technology and artificial intelligence. The history of VR in gaming goes back to the clunky helmets and goggles of the ‘80s and has advanced to where lighter, wireless headgear that allows more freedom of movement as well as a more physically comfortable gaming experience. Augmented reality seeks to provide a similar experience, but while still using a screen or handheld device rather than a headset. Marrying up either of these two types of advanced realism with an AI that can operate in and alter that environment is a long-term goal of game designers, designers who dream someday of making the holodeck of the starship Enterprise come to life.