There are plenty of articles out on the internet talking about how 2017 will be the year of AI and how it will change the way in which we drive our cars, organize our homes with the internet of things (IoT), and live our lives. But much of the progress made in developing AIs could never have been made had it not been for video games spurring developers on.
Video game AIs have come a long way from when Nintendo sent Goombas marching at Mario in a set path and are certainly now capable of carrying out complex commands. But AI developers are now trying to create AIs that can actually think, learn and develop their own personalities as opposed to following a script with branching paths. This can not only affect how we interact and play video games, but help create better AI to change how we live our day to day lives.
The power of machine learning
Of course, we have already managed to create AI which can learn. The famous Google AlphaGo program which defeated the best Go player in the world last year and recently demolished additional top players playing on online, became so good thanks to machine learning which allowed it to analyze millions of played games and develop strategies and moves which no humans had considered. Japan, which has been leading the way in AI development, launched an initiative this week which “aims to set the foundation for a smart society by March 2021” in part through machine learning and big data.
It is one thing to learn how to play Go or understand data, and another to learn and understand human relations. But such a field could be incredibly useful in designing a video game world. Instead of making a city where the non-player characters (NPCs) stand around or spout the same few lines upon prodding by the player character, there could be a city where those characters can meaningfully interact with one another and function like actual human beings.
Of course, the question becomes how an AI would learn how to communicate and interact through machine learning. But just as AlphaGo learned to play Go through watching millions of games, another AI can learn how to communicate by watching examples of human communication. Think about the infamous example of Microsoft’s Tay chat bot. While Tay ended up learning to spout racist and sexist drivel, it still ended up learning something about human communication.
But it is in the field of video games where we could learn to diversify and expand how AIs can learn. A company like Amazon will want a chat bot for a few, limited purposes such as customer service. But if we think about a video game city where AIs are interacting with one another after learning how to interact through machine learning, the result can be a much more varied experience. Even a racist AI like Tay would have value in this sort of context.
More than video games
While machine learning can help produce chat bots which can enhance the quality of games, AI game developers and gaming schools have been taking other paths to create AI that can have emotions and try to create something resembling human relationships for years. The Guardian notes an AI engine developed in 2013 named Versu where instead of coding a certain number of limited interactions for each NPC, each NPC “had their own beliefs, abilities and parameters which, in theory, lead to an infinite range of habits and quirky behaviors.” Versu’s developer noted a scenario where one NPC grew angry with the player character because the player character had been rude to a second NPC who then told the first NPC. This sort of realistic interaction is different from the standard scripts most players are used to, and would provide a new, better experience.
Helping the real world
Video game developers may be working to create new emotional AI but its usefulness is limited if we only consider games. Take the above city example. While it may be more interesting to have a city where the NPCs interact with one another, a game developer may decide that it does not appreciably enhance the player’s experience and conclude that it is not worth the time or extra memory. But an AI like that could still be used in fields beyond video games where an emotional AI would be valuable.
The result is a continued progression of more varied and elaborate AI. Video games can be used as a testing ground to develop new AI, as the field of games can use more varied AI which can in turn lead to different kinds of games as we can create new emotional bonds with a deeper AI. At the same time, other businesses can take advantage of the work which video game developers have done to enhance their own AIs and create a new smart society.
This article was written by Gary Eastwood from CIO and was legally licensed through the NewsCred publisher network.