We hear a lot of talk about artificial intelligence today, thanks to high profile companies like Tesla who are using the technology in their products to automate certain actions. Gamers are probably more familiar with the term than most as it’s often used to refer to computer-controlled characters that we play against in single-player modes.
This hasn’t always truly been artificial intelligence though, even if it often appeared to be. It also isn’t the only way that the technology is used within the gaming industry. We’ve seen it before with Virtual Reality, which is also very present today. Here are some of the ways that AI has evolved in the way we create and enjoy games.
Creating AI to Showcase Supercomputers
Artificial intelligence and computers that are comprehensively fast have often been paired together by manufacturers and data scientists to perform all kinds of useful tasks, like modelling the weather and climate change and finding cures for terminal diseases.
No matter how valuable these endeavours are, they don’t make great news headlines. That’s one of the reasons that several of the most powerful supercomputers ever made have been tasked with beating the best players of games like chess.
In 1996, Deep Blue was an IBM supercomputer that beat Garry Kasparov in one of six games the two competed in. However, the machine was specially designed for the task, with 512 specially designed chess chips.
Since then, some manufacturers have attempted to create systems that can beat humans at other strategy games like poker. This makes sense as there are many similarities between the two, including the fact that there are a number of standard opening “moves” a player can make and that it’s difficult to keep score during a game. In both games, it’s also vital that players recognise the value of each piece or card, as they are not all created equal. In poker, different combinations of cards are ranked higher than others, while in chess, pawns are considerably less important than the rook or queen.
However, it’s a little more difficult in this case because computers can’t read body language and facial expressions in the same way and, therefore, can’t account for the bluffing that’s a big part of the game.
While supercomputers are yet to master poker, they have mastered chess to the point where it’s no longer a fair fight between them and humans.
AI in Non-Playable Characters
From the earliest video games like Spacewar!, Pong, and Pac-Man, we’ve played against computer-controlled characters or objects. Usually, their task is to attack us or beat us in a sporting simulation.
To make this interesting and not just have these non-playable characters (NPCs) running randomly around the screen or map, they have been programmed to behave in certain ways. This is often referred to as the “game’s AI”, though it isn’t strictly true.
In older games, the method for programming behaviours used a “Finite State Machine” algorithm. This essentially was a list of pre-programmed responses to all the possible movements that the programmer anticipated a player could make.
The 1999 game Driver, which was released on PlayStation and PC is a good example of this. When Tanner is being chased by the police, the cop cars will attempt to ram his car and follow the path it travels with almost no deviation. This means that it’s easy to evade them by simply weaving between traffic and objects, causing the police cars to crash.
More recent games make use of other techniques such as pathfinding and decision trees, with some beginning to use AI to make more realistic NPCs. One way that this has been done is to introduce “hunting” where the NPC will use visual and audio clues to search for a player rather than instinctively know where they are.
AI for Map Generation
Another area of gaming where AI is used is in the creation of maps. In games like Grand Theft Auto, the map is finite and has been carefully crafted by the designers. While maps have been getting bigger and bigger, they can still feel like a sandbox.
To get around this, developers have begun turning to artificial intelligence to create maps that are infinitely large and are developed on the fly by the computer or console as the player moves around.
This process is called procedural generation and is actually not a new concept. In 1978, the game Maze Craze used this technique to create new mazes every time the player started playing. However, modern hardware has allowed developers to finally make use of it in larger titles.
Today, Minecraft, Elite: Dangerous, Star Citizen, and No Man’s Sky all use this AI-based process to create maps for players. This allows for an entirely new experience each time you load the game and prevents new players from being disadvantaged when playing against veterans with map knowledge built up over hundreds of hours of gameplay.