Music AI was trained on video games To bleed
Microsoft
An artificial intelligence model of Microsoft can rebuild realistic video game footage that the company says that can help designers to create games, but experts are unrelated that this tool will be useful for most game developers .
Nerve networks that can produce consistent and accurate footage from video games are not new. Recently Google-manufactured AI created a fully playing version of the classic computer game Apocalypse Without access to the underlying game engine. Original Apocalypse, However, it was released in 1993; More modern sports are far more complex, with sophisticated physics and computationally intensive graphics, which have proved to be difficult for AIS to rebuild honestly.
Now, Katja Hoffman Microsoft Research and its colleagues have developed an AI model called Muse, which can re -create the full sequences of the multiplayer online battle game is bleeding. Hoffman says that these sequences follow the underlying physics of the game and keep players and in-game objects compatible with time, meaning that the model has understood the deep understanding of the game.
Muse is trained on a seven -year human gameplay data, provided by both controller and video footage, To bleedMicrosoft -owned developer, Ninja Studio. It works similar to big language models such as chatgip; When an input is given, as a video game frame and its affiliated controller tasks, it is assigned to predict the gameplay that may come forward. Says Hoffman, “It is really quite desired, even now, that purely training model to guess what is going to appear further … it is the one of this complex 3D atmosphere A sophisticated, deep understanding learns, “says Hoffman.
To understand how people can use AI tools such as muse, the team also surveyed game developers to know what facilities they would find useful. As a result, researchers added ability to recur the changes on the fly, such as changing the character of a player or new items entering a scene. It can be useful, Hoffman says that it is useful for coming with new ideas and what is the landscape for developers.
But the muse is still limited to generate sequence within the range of origin. To bleed Game – It cannot come up with new concepts or design. And it is not clear whether it is an inherent range of model, or something that can be removed with more training data than other sports. Mike Cook King’s College at London. “This is a long, long way from the idea that AI systems can design games on their own.”
Cook says, while the ability to generate frequent gameplay sequences is impressive, developers may prefer to control more. “If you create a device that is actually testing your game, then the game code is running yourself, you do not need to worry about firmness or stability, as it is running a real game . Therefore, they are solving problems that have been introduced by the generative AI itself. ,
It promises that the model is designed keeping in mind the developers, says Georgios Yanakis At the Institute of Digital Games at Malta University, but it may not be possible for most developers who do not have so much training data. “It comes down to this question whether it is worth doing?” Yanakakis is called. “Microsoft spent seven years and collected data and trained to train these models that you can actually do so. But will a real game be borne by studio (to do so)? ,
Even Microsoft itself is similar to whether the AI-designed games can be on the horizon: when asked if they can use the developers tool in its Xbox gaming division, the company can comment Refused
While Hoffman and his team hope that the future versions of the future museum will be able to normalize beyond their training data – coming with new scenarios and levels for the games on which they are trained, as well as also Working for various games together – this will be an important. Challenge, Cook says, because modern games are very complex.
“A game is one of the ways to separate a game separately and introduce new ideological level ideas. This machine inventions out of its training data for learning systems and invents what you have seen, ”he says.
Subject:
(Tagstotransite) video game
