Video games generated only with artificial intelligence, evolution or involution?

Artificial intelligence has led to the automation of numerous activities. One of them has been the development of video games. Are there noticeable differences between these and those created by humans? Are they equally exciting? A new study has sought answers to these questions.

The study has been carried out by a team from the Valencian University Institute for Research in Artificial Intelligence (VRAIN) of the Polytechnic University of Valencia (UPV) in Spain, the University College of London in the United Kingdom and the University of San Jorge in Zaragoza in Spain.

Video games generated only with artificial intelligence are worse than those that combine artificial intelligence with the participation of human developers; the latter have more quality and naturalness. This is what the study reveals.

The authors of the study reached this conclusion after a rigorous evaluation of automatic video game generation techniques with artificial intelligence.

According to Carlos Cetina, researcher at the VRAIN Institute of the UPV, companies like Meta, Apple or Microsoft have the vision that in the future we will not only understand each other, but we will also work, study and socialize in digital environments where video game technologies are a fundamental pillar, what the European Union (EU) calls Virtual Worlds.

“However, the most urgent challenge to continue building video games or achieve that vision of Virtual Worlds is the generation of video game content. It is precisely the generation of content that makes developing video games like Cyberpunk 2077 or GTA VI require development of around 10 years with teams of thousands of people who have heterogeneous profiles,” says Cetina.

Carlos Cetina. (Photo: UPV)

Currently, there are different techniques to accelerate the generation of content, although, because research in video game development is still very incipient, a comparison of these techniques had not yet been made through a rigorous controlled experiment. “Also because it is necessary to consider specific aspects such as design, difficulty, fun or immersion that are not considered in traditional software,” adds Carlos Cetina.

Along with Carlos Cetina, Mar Zamorano and Federica Sarro, from University College London, and África Domingo, from the University of San Jorge, participated in the study.

The work has won the award for the best technical academic article in the 18th edition of the ACM/IEE International Symposium on Empirical Software Engineering an Measurement (ESEM), the world congress where results related to empirical software engineering are presented. (Source: UPV)