Developing An AI That Can Build Video Games


Michael Cook, a PhD student with the Computational Creativity Group at Imperial College in London, aims to bring his PhD project “Games By Angelina” to the masses! He wants to answer one simple question that no one seems to have answered yet: Can we evolve entire arcade games from nothing? The basic concept is that the project starts with nothing at all, except for a few ideas about what a game contains. The computer is then asked to create characters, design levels, and create a simple set of rules that is both fun and challenging!

Designing the Designer

For the next two years, “Games By Angelina” will be worked on along with updated games that Angelina (the program) creates herself for everyone to enjoy. This is such a fun and unique concept that it will be interesting to see how this project takes off in the future. Will this program be able to design games for everyone to play or will it fail miserably? We’ll all have to wait and see.

Progress to Date

So far a total of two games have been created to play by Angelina herself. While they are not that advanced as of it, it’s a very interesting concept to look at how the games were designed. Although there has been some criticism concerning this project. Some users have asked why this project was even created, as it seems that Michael is just trying to automate the least important tasks of making a game. Michael however, answers by saying that “Our remit really is looking at building software that performs creative tasks – I wanted to work in videogames because I pretty much live and breathe them, which is really useful when you’re doing research because you need to be able to talk around these topics and know all the edge cases and tiny projects and what’s been tried before where. There’s something appealing about aiming for something stupidly difficult, too. Think how hard it’s going to be to build software that can express meaning through gameplay mechanics, score original music and create interesting, clear graphics for players to recognize”

His answer represents something that most computer scientists love to mess with. That is, taking an absurdly difficult problem or task and trying to make it work. Just think about how many programs have been coded in the past to solve the silly problems: statistical sports calculators, the amount of pennies it would take to fill up the Empire State Building, etc. The list goes on and on.

“Co-operative Co-education”

Upon working on this project, Michael has stated that he is using “co-operative co-education” to essentially work backwards in relation to certain problems. Michael says: “With co-evolution, multiple processes are being run at once to improve the search for certain problems. The different evolutionary processes evolve separate bits of a solution and co-operate in their search for a finished whole (the game itself).”

As far as releasing of the source code, Michael says that this is in the works, but he just needs to get everything together. With a release of the source code, other computer scientists could possibly collaborate together to build even better and more intelligent versions of Angelina to create even better games. Over the next two years, we should be seeing some interesting projects and development between both Michael and Angelina! Check out “Games By Angelina” for more information and current progress on the project.

This article was written by Rachel Oda for the team at who are facinated by the human mind and any attempts to replicate it using technology.

Choosing The Right Image

Previous article

Outsourcing Solutions: Wave Of The Present?

Next article

You may also like


Comments are closed.

More in News