Researchers at the Georgia Institute of Technology have developed a system capable of creating sections of a video game after analyzing a video game.
On the Internet there are quite a lot of videos that show how a character is playing a video game. From first-person to strategy, to console games from the 1990s, YouTube and other platforms are packed with recordings that pretend to be tutorials or simply demonstrate a player’s skills. This material is enough for a new computer to create imitations of these video games.
A team of researchers from the Georgia Institute of Technology has developed a computer system that can design sections of video games just by watching a video where the original title is represented. Computer it analyzes the video and autonomously develops a similar scenario the one that the pictures show.
In a first approach, the system creates sections of scenarios ready to play, imitating the originals. Researcher they don’t have to program code or set design criteriathe computer takes care of everything. Of course, to be able to do so you have to see the characters advance on the stage.
If there is no one playing the system has no reference. After all, video games are meant for someone to play and move forward. With this information the algorithm discovers which parts the character spends the most time, where he earns the most bonuses or in which places it costs him the most effort to earn them. Apart from being used to configure the scenario, this type of data is useful so that the computer does not make errors. For example he will never create jumps so big that the character can’t save them.
The system has started with simple video games, but researchers hope it will soon be able to mimic more complex titles
The first video game that researchers at the Georgia Institute of Technology have experimented with was the Super Mario Bros, a two-dimensional video game that was released in 1985. The new system has been able to mimic the scenarios of bricks, pipes,blocks and a friendly blue sky with a few clouds.
It should be noted that it is only the scenarios that are designed and programmed automatically. However, to be able to play them you have to export them to a video game engine, an aspect that the system cannot handle simply by analyzing a video.
On the other hand, researchers trust that their technique can be applied to different video games in order for the computer to accumulate knowledge in order to create original games of a certain genre. There’s still time for this. The team goes step by step and the next is to develop full-scale levels and evaluate how players interact with them compared to the original games.
Images: Bloomua and Valua Vitaly