AI designs Super Mario levels by watching videos

Researchers develop computer program capable of building new levels of video games by watching videos being played

Photo by: Game ScreenShot
Photo by: Game ScreenShot

Updated Jul 28, 2015

Researchers from Georgia Institute of Technology have developed Artificial Intelligence (AI) software which can build new game levels by watching gameplay videos on streaming services such as YouTube.

The team decided to try their new computing system on Super Mario, making his quest to find the Princess much harder.

“The system focuses on the gaming terrain and the positioning between elements on-screen – be it pipes, blocks, coins or Goombas – and it determines the required relationship or level design rule,” according to a blogpost on the university’s website.

The AI, simply by watching the gameplay videos on YouTube, successfully created 151 unique levels from 17 game samples.

“An initial evaluation of our approach indicates an ability to produce level sections that are both playable and close to the original without hand coding any design criteria,” says Matthew Guzdial, lead researcher and PhD student at Georgia Tech.

While watching the game being played the AI focuses on the areas where gamers spend most of their game time to collect bonuses or overcome a challenge.

“Our system creates a model or template, and it’s able to produce level sections that have never been seen before, do not appear random and can be traversed by the player,” says Mark Riedl, associate professor of Interactive Computing.

“One could say that the system ‘studies’ the design of Mario levels until it is able to create new playable areas.”

Riedl and Guzdial presented their research at the International Conference on the Foundations of Digital Games which began on June 22 in California.