[ad_1]
Nvidia introduced Monday (3) the first demonstration of a "game" created with graphics generated by artificial intelligence. The game, a driving simulator, is not really a fun gaming experience, but rather a proof of concept of a new way to render graphics that are still far from reaching the biggest titles in the industry.
Pokémon Tamagotchi is the virtual pet that you might want enough to stay alive
The impressive advances in artificial intelligence now offer a new area in which to show success: the generation of game graphics. The demonstration released Monday by Nvidia is a hybrid graphics system that combines a traditional game engine with images generated by artificial intelligence.
The company's engineers explain in a study that the creation was not totally unprecedented: it was based on a number of existing methods, such as the open source pix2pix system. Nvidia 's work utilizes a neural network known as the Generative Gator Network, or GAN, very common for imaging, observed recently in action as part of the program. artificial intelligence auctioned by Christie's.
"Nvidia has created new ways to generate interactive graphics 25 years ago – and it's the first time we've been able to do that with a neural network." Especially the generative models – will change the way the graphics are created, "said Bryan Catanzaro, vice president of in-depth learning application of the Nvidia website. For Catanzaro, this method will allow artists and developers to "create at a much lower cost, using artificial intelligence that draws lessons from the real world."
The graphics of the video below were generated with the following steps: first, the collection of data to form the badysis of impact, taken, in this case, to from data sets used for autonomous management research. The shoot is then segmented and each image is separated into different categories: cars, trees, roads, buildings, sky, etc. The GAN is then formed with this segmented data to generate new versions of these objects. Engineers then create the basic topology of the virtual environment with a traditional game engine – Unreal Engine 4 in this case – and this environment serves as a framework for deep learning algorithms to generate graphs for each category of object in real time, pasting them into models of the game engine.
Although it is not 39, act as a simulation of direction and use of existing methods, the creation of Nvidia is important because it is the first game demonstration with graphics generated by AI. As you can see, there is nothing ultra-realistic, but the goal is not yet that, and the application of this type of method is still far away.
In an interview with Verge Bryan Catanzaro said the technology is in its infancy and should take "decades before AI-generated graphs reach headlines for consumers"
. The demonstration was done with a single GPU, which is remarkable for such work. However, it should be emphasized that it was a GPU "A": Titan V, of Nvidia same, presented by the company as "the most powerful PC GPU ever created". Verge points out that even though we are far from seeing technology in the main games of the industry, it would have possible applications in other areas, such as robot research or autonomous cars, which can generate in the latter, for example, training environments for the systems. Bringing novelty back to the playing field, the site is also considering a possible application of technology badociated with traditional methods, using artificial intelligence to create more characters and similar objects. This would allow, for example, that your avatar in a 2K NBA life is even more similar.
Another important point to be addressed is that advances in such technologies pose challenges to humanity as a whole. The use of artificial intelligence has already posed a problem recently, with deep fakes, more and more advanced and hard to distinguish from real images. Catanzaro admits the dangerous potential: "If this technology can be used to create deceptive content? Yes, any rendering technology can be used for that." However, he prefers to avoid liability. "Do you feel that the energy company is responsible because it has created electricity that feeds the computer that makes the fake video?"
Whatever the possible perverse uses, work with artificial intelligence in its various fields will continue – with even more impressive results, even in projects that are still living their initial stages, such as that. is the case of the one directed by Nvidia
[Nvidia e The Verge]
Source link
