AI in Games Other Than Player
I define myself as a core gamer, a gamer who enjoys wide range of games and haves deep interest on various mechanics of games. I have played several hundred in my whole life so it might be very natural that I connect many things on my life with similar aspects of games. When talking about AI, of course, I would confidently bring out topics related to games.
Wide meaning of AI includes mechanically coded action pattern of machine. For example, a famous classic video game called Pacman is a game to control Pacman, or Mr. Pac in some series, to eat all cookies on stage without being caught by ghosts. There are four different ghosts in stage and each of them has different rule to catch Pacman. Blinky, the red ghost directly chases Pacman in shortest possible way. The pink ghost Pinky rushes toward 4 cells in front of Pacman. The blue one, Inky tries to cooperate with Blinky to surround Pacman. The slowest orange ghost, Clyde tries to keep distance with Pacman, causing random mistake when avoding other ghosts. These patterns work based on hardcoded reaction depend on Pacman’s movement. According to wide definition of AI, those movement of ghosts are exact AI itself.
With several decades of conceptual and technological advancement, definition of AI became narrow. In fact, the recognition of public about AI has changed into narrower field. People nowadays first think about machine learning when they are told about AI. Especially many AI figures we can see on movies or novels, which talk, think, and act like human and hardly distinguishable from mankind. This is little bit too far to achieve at this time.
Combining game and AI would give us wider consequenses then we expected. Most widely known AI in position of player might be AlphaGo of Google DeepMind. In March 2016, Google challenged Se-dol Lee, the strongest Go(Baduk) player in the world with their trained AI. Go was challenged many times by many AI developers but the simpleness of the rule and complicacy of conjunctures of Go frustrated all of them. However, contrary to many people’s expectation, AlphaGo defeated Se-dol Lee by score of 4 to 1 and Go was considered as another game conquered by AI from at time.
Also, AI named OpenAI from company openAI had once defeated pro players team in strategy game called DOTA 2, by limiting number of actions in unit time similar to human players. Also, AlphaStar of Google DeepMind is developed to play another strategy game Starcraft 2 like human. However, AlphaStar project has been stopped since it could not show new type of strategy of build order, but focused on only microcontrol of individual units.
One might ask a question “So AI can only replace human players’ roles?” The answer is, that is apparently not true. Only portion of AI technologies are used in replacing human game players. Majority of AI is used on processing data in and out of game itself.
There is a project called Enhancing Photorealism Enhancement. This project is aiming on processing image and change it like photorealistic image by using machine learning. The engine has image model which is trained with real images. When this engine is added to game, it will convert every frame rendered into photorealistic image, making game fall in deep realism.
I once created an AI-supported game as a class group project using Google’s Teachable Machine.
This simple boxing game is developed with image classifying model created by Teachable Machine and some bunch of javascript. The trained model would distinguish position of arms and fists, and when the player’s pose and instruction matches, the player would get score by decreasing enemy’s health. In this game, it seems like AI has no place in the game. However, the core mechanic of the game — image classification — is built on trained machine-learning model.
Also I have tried to create new design of fighter planes with GAN-generated images. By using over 150 images of fighters appearing in game called R-Type series, I let AI to create new designs of fighter. The portion of viable result was relatively small, but it is important that this project has shown potential of AI to support, or even replace, human work on creativity perspective.
Many people might think about AI being used in medical field or manufacturing. Machine can do many things with deeper concentration for longer time than human. Yes, there are various fields that deserve initiative on applying AI technology to make mankind’s life rich and fruitful. Natural language interpretation will tear down the wall between different language users, and crime predicting AI will lower crime rate globally. It seems like entertainment like games has very low preference to apply AI technology. However, I believe that amusement is very huge part of modern human’s life. Being tired of repeating everyday life can be relaxed by playing game, listening music, or watching movies. Game, the only interactive mean of entertainment among many media contents, is of course important component of our lives. I anticipate that one day a game fully developed by AI will be available for public users.
