Artificial intelligence (AI) has reached the ninth art and by Microsoft. The technological giant has been investigating the application of this tool in video games for years, but its approach intensified in 2023, after the purchase of Blizzard Activision and the excessive success of Chatgpt. In small aspects the AI It is already a useful instrument that is used in the industrybut he was not yet able to understand the needs of video game developers. Until now.
Microsoft Research presents this Wednesday In the magazine ‘Nature’ To Muse, the first model of human and world action (WHAM), a tool of AI that not only creates image, video or voice, but also 3D environments with which one can interact and modify, with the precision of a developer that has been in the industry for years. You can create maps, characters and their physical ones with the ease of one more professional, provided you have an extensive database to nourish. As we said, the AI so far was able to create designs for cosmetics and outfits (see the so -called ‘skins’ in video games) or dubbing for characters, but always under marked guidelines, ‘instructions’ that the developer indicated of principle In order to execute. And if you tried to retouch something, the safest thing is that it became chaos.
Playable mechanics such as those of a more developer
Muse, on the other hand, generates new playable mechanics that resemble the existing ones, as if behind them were a person with years of project experience and not a machine. The tool has trained using around 500,000 games (7 years of experience) of the video game ‘Bleeding Edge’a multiplayer battles simulator. The researchers discovered that Muse could create complex 3D video game sequences that were consistent with the pre -existing mechanics of ‘Bleeding Edge’, with great diversity in the design of levels and that allowed the creatives to adjust and polish the details at will.
This is thanks to the fact that it has Divergente thought, one of the greatest shortcomings in the video game sector, according to Katja Hofmann (author of the WHAM model) and his colleagues, which interviewed 27 video game developers of creative equipment To understand your needs. «IA usually has a limited context, with few data. This means that it is difficult for her to generate content that follow specific rules and mechanics, ”a vice president of an independent video game study told the authors. Muse Sol for this problem, when generating ideas with the style of the game that feeds on information.
“It has been surprising to see the variety of ways in which Microsoft Research has used the environment and data of ‘Bleeding edge’ to explore novel techniques in a rapid movement industry,” explains Gavin Costello, technical director of Ninja Theory, in A Microsoft press release. «From the Hackathon that began everything, where we first integrate the AI in ‘Bleeding Edge’, until the creation of AI agents that could behave more as human players, to the model of human and world action capable of dreaming completely sequences New ‘Bleeding Edge’ game under the human guide, it has been revealing to see the potential that this type of technology has ».
In addition, Hofmann and his team have also developed the WHAM DEMONSTRATOR, An open source visual interface for users to interact with the results of the WHAM model and customize them, as if it were the program that is used in the graphic engine of a video game or in those of video editing. WHAM DEMONSTRATOR In addition, it will be publicly shared so that every interested developer can struggle with him. «I am incredibly proud of our teams and the mile The creative uses of generative models, ”explains Hofmann.
Consistency, diversity and persistence
The bases behind Muse are based on three concepts: consistency, diversity and persistence. About the consistencyIt is the capacity of the model to generate game sequences that respect the dynamics of the game. For example, the character moves consistently with the controller’s actions, does not cross walls and, in general, reflects the physics of the underlying game. As for the diversityMuse is able to create different sequences of the same creation. That is, it can generate, for example, three different routes or paths on a map, changing the angle of the camera or the objects that appear.
Finally, the persistence It demonstrates the model of the model to “endure” user modifications in sequences generated by AI. For example, based on what is generated by AI, developers proved to include new elements (such as a jump platform, interactable objects or even a new character), and the sequence remained stable and adopted the additions as if they were their own, generated by AI.
Muse’s advantages
The authors consider that, since Muse learned to generate sequences with just training on the game and without prior knowledge, this tool could easily transfer to generate levels of other video games. It can be useful for games as a service, such as ‘Fortnite’ or ‘League of Legends’, with the aim of alleviating the growing demand and expectations of players to count, constantly, new content.
However, the authors point out that it is unlikely that the tools of AI as the WHAM model can serve as the final point of the design process and that, however, they should be used as a tool to help game designers. WHAM models They do not replace the worker, but they allow expanding the ideas of the creative design process of a video game, in addition to reducing development times.
About its implementation in future titles, Xbox has already begun to explore AI in some projectsand only time will say if the rest studies of the sector adopt this type of tool, but the road leads, little by little, to that. «I do not see the time to see all the ways in which these models and subsequent research will help shape and increase our understanding of how generative models of human play can support the ideation of games and pave the way for future experiences of novel play based on AI, ”concludes Hofmann.
#Microsoft #presents #muse #capable #developing #video #games
