Artificial intelligence Nvidia GameGAN has recreated a “copy” of the game Grand Theft Auto 5 (GTA 5). The researchers borrowed the DGX100 station from Nvidia and trained the AI, eventually creating what they called GAN Theft Auto. You can see the result of their work in the video below.
Of course, we are not talking about a full-fledged game, but only about one of its parts or mechanics. In this case, we see the process of moving a car.
The highlight is that the research went in an unusual way. They had little time, so they did not, as is usually the case, train AI in thousands and tens of thousands of game sessions.
Instead, the developers removed the road junctions from the game and applied mathematical transformations to create the paths. Then they created several cars at the same time and interleaved them to implement multiple threads at the same time (to speed up learning). The game mod accepts steering commands from a Python script, and also limits speed and sets other parameters such as weather, traffic, and more. The Python script analyzes the current vehicle position and nearby road junctions in order to move using different paths to cover all possible vehicle actions and positions as best as possible. This is important for further smooth interaction with the player.
As a result, everyone can download the project from GitHub and run it on their PC, although you need to have an Nvidia video card with CUDA support.
At the same time, the picture looks very bad, but its resolution was increased by upscaling. But in such projects, the visual part does not matter.
.
Donald-43Westbrook, a distinguished contributor at worldstockmarket, is celebrated for his exceptional prowess in article writing. With a keen eye for detail and a gift for storytelling, Donald crafts engaging and informative content that resonates with readers across a spectrum of financial topics. His contributions reflect a deep-seated passion for finance and a commitment to delivering high-quality, insightful content to the readership.