A team of researchers from the University of Copenhagen has developed an AI to generate Super Mario Bros levels from a series of words entered by the user. A more scientific tool than a video game.
Even Mario video games can be generated indefinitely! After ChatGPT popularized the use of artificial intelligence to generate complex responses to user requests, MarioGPT proves that the recipe can be applied to video games as well.
This is demonstrated by six researchers from the University of Information Technology in Copenhagen in an academic article pre-published online on February 12. Based on the first two Mario games (Super Mario Bros et Super Mario Bros: The Lost Levelsreleased in 1985 and 1986), his tool allows to generate random but playable 2D levels of the Nintendo hero, while adapting to the wishes of the users.
« Many pipes, many enemies, small blocks »
If creating new levels for the mustachioed plumber was already possible with the series of games super mario creator or through other unofficial level generators, MarioGPT is the first to “generate multiple levels, but can also rely on textual suggestions to control level generation“, explain the researchers in the introduction to their article.
In fact, the content produced by MarioGPT is linked to ” indications“, those more or less thought-out combinations of words that we type into the search bar of an image-generating AI like Dall-E to obtain more or less precise results.
In the case of MarioGPT, the AI allows you to play at the frequency of certain game elements, enemies, or even the overall height of the level. The researchers’ article gives some examples: many pipes, many enemies, few blocks, low altitude».
A different method of ChatGPT
To achieve this, the Danish researchers relied on the GPT-2 artificial intelligence model, the online medium specifies. TechCrunch: an older version than the one used by ChatGPT and less good at imitating human language.
But this is not a problem in the case of MarioGPT, since the group of academics used it to generate high-speed computer code: MarioGPT gradually learned to identify similar code patterns (game elements, enemies, etc.) before repeat them according to the logic of the game.
Available for free… for insiders
If you want to try the program, it’s available for free on Github… but it will be a bit more complex than using ChatGPT. MarioGPT requires some basic concepts in game development and knowing how to run code in Python 3.8, details the information sitecutting gear.
Similarly, game levels spawned are limited to one type of enemy spawned at a time, rather than varying goombas and koopas. Because the objective of MarioGPT continues to be more scientific than playful: to demonstrate that the procedural generation of video games through prompts is possible.
But Danish academics are not closing the door on improving their tool. As the researcher Shyam Sudhakaran explains in his interview forTechCrunch«in the future, with larger data sets and more complicated messages, we may need to use a more sophisticated model like GPT-3.»
To follow us, we invite you to Download our app for Android and iOS. You can read our articles, archives and watch our latest YouTube videos.