This Is ANGELINA, And This AI Doesn't Play Games. Instead, It Creates Them

We use the term "AI" to describe non-player characters (NPCs) on video games. But the fact is that the "real AI" is something that learns to do something, not coded to.

Mike Cook, an AI researcher at the University of Falmouth, has spent his time creating a simple video game that involved a basic concept: the player is an adventurer around a dungeon, killing skeletons to reach the exit. Cook didn't finish developing the game, but made an AI to do the job instead.

That system is called ANGELINA, a tongue-in-cheek acronym for “A Novel Game-Evolving Labrat I’ve Named ANGELINA”.

What ANGELINA did, was taking Cook's initial idea, and processed it. But throughout the development, the AI started designing its own levels for the game, with the result of changing the game's purpose.

Instead of the player controlling the adventurer to kill skeletons to reach the exit, ANGELINA designed the game so the player needs to sacrifice some of the multiple adventurers in each level in order to allow the rest to escape safely.

It was a really clever piece of level design," said Cook.

He has been working on ANGELINA and watching it create games for a few years. From about 30 games ANGELINA built, some were fun games, few were pretty bad, and some were in between.

Most of the games were deleted and never saw the light of further development. Cook said that they don’t really represent the full scope of the system’s capabilities, just how an artist throwing away old sketches, Cook explained.

"The most common technique in ANGELINA is computational evolution. So let’s say it’s designing a level for a game it’s making. It’ll begin with a big bunch of random levels, most of them really bad, and it’ll play each level and try and finish it. And almost all of them will be terrible, but maybe one or two are slightly less terrible than the others. So it picks the best ones, and mixes them together to make new levels. Maybe it takes the top half of one level and the bottom half of another level and glues them together. Then it starts the process again with these new hybrid levels."

"I’ve broken down the game design process into different tasks, like designing levels, inventing rules, or testing difficulty. ANGELINA selects tasks based on what it’s current game is missing, and then uses techniques like computational evolution to perform each task. Right now it has to rely on my opinions for some things – for example, I tell it that levels which take longer to solve must be better. Hopefully in the future I can make it more independent so it can throw away these ideas and develop new opinions for what makes a level good or a game fun."

Mike Cook has been interested in software creation, how software can be creative, and how it can help humans in becoming more creative when designing games. When Cook learned about the field of computational creativity, he started building games on his own.

And then an idea popped in his mind. Instead of building games on his own, he could build an engine to do just that.

ANGELINA is an AI that can be made to develop games. While it has evolved throughout the years, it's a very simple AI and it can't do much. But it indeed allows us to ask bigger questions about the fields of AI. Like for example: "if an AI creates something, who legally owns it? Should we pay for something that is created by an AI?"

As for researchers, they can learn the limits of how AI can be made to "think" and become creative. They can learn how AI makes its own decision to solve problems on its own.

"The best artificial intelligence research isn’t about really creating technology for now, it’s about asking questions about the future," Cook said.

Related: Unity's Machine Learning Agents Want To Change The Perception About "AI" In Games