Xpert
@xperts
Microsoft Unveils World's First AI to Generate Video Games
0 reply
1 recast
6 reactions
Xpert
@xperts
Microsoft has unveiled Muse, a generative AI model designed to analyze and predict gameplay scenarios. Unlike traditional development tools, Muse doesn't create new game mechanics, but uses learned data to predict how gameplay will change when certain changes are made. The basic idea behind Muse is to train AI based on existing game data, so that it can then predict how changes to a level or mechanic will affect gameplay. For example, if a developer adds a new ramp to a location, Muse can visualize how players will use it and what consequences it will have.
0 reply
0 recast
1 reaction
Xpert
@xperts
For training, Microsoft used gameplay footage from the multiplayer shooter Bleeding Edge. The training process itself required a cluster of 100 Nvidia H100 graphics cards and about a million training iterations to extend 1 second of real gameplay to 9 seconds of simulated gameplay. The main advantage of Muse is its ability to preserve the logic of object interactions. For example, if a model duplicates an enemy or an item in the game, it will behave the same as the original. This can be useful when testing level balance or developing new mechanics.
0 reply
0 recast
0 reaction
Xpert
@xperts
While Muse cannot currently replace traditional development tools, the technology demonstrates an interesting approach to gameplay analysis. In the future, if it is possible to reduce computational costs and improve the quality of the simulated data, such systems may become useful assistants for developers, especially during the testing phase. Microsoft also mentioned the possible use of Muse in game archiving, suggesting that the model could help with the adaptation of classic games to new devices. However, it is still unclear how accurate and effective such a method would be.
0 reply
0 recast
0 reaction