Generating Game Content
Problem being addressed
How can machine learning be used to generate content during a computer game?
Machine learning and gaming have always gone hand in hand. This ranges from the development of programs to play checkers, to fully fledged AI adversaries in strategy games. In this research paper, generative machine learning techniques are applied to create game world content. The focus is on level creation, map creation and game world story creation and the basic idea is to use build a generative AI by training it with "good" content. Multiple techniques for content creation are considered and these range from neural networks to deep learning and hidden Markov approaches.
Advantages of this solution
Game content creation is usually something that is a task for human story writers and developers. This is an interesting approach, and may lead to novel game worlds that are created on the spot as the game is being played.
Solution originally applied in these industries
Possible New Application of the Work
In the education sector generative content such as multiple choice questions could be useful as a tool for teaching and learning. This could lead to customized exams that are tailored to where the student is currently, rather than the current one size fits all approach.
Generative media could be an interesting side application of this research. What about an AI that writes articles, creates music, and does the work that "human" creatives normally pride themselves in doing.
Generative AI content could be useful for building space training programs. We could think of simulated VR environments for potential astronauts to navigate. It's not too far away from the Star Trek holodeck.
Source DOI: #############
Source URL: #############