This is Grand Theft Auto as you have never seen before, named GAN Theft Auto.
Please join us on our new Facebook page here: The Best in Gaming
Artificial Intelligence can be used for far more than we are used to seeing in games, recently we have seen a couple of Al researchers look into Nvidia’s GameGan neural network and have created something pretty incredible.
They have created GAN Theft Auto, a version of Grand Theft Auto that is completely Al-generated, a playable demo which includes driving along the highway in GTA 5.
It could be said the demo looks as if you were playing GTA 5 in a dream as the graphics are hazy and slightly pixelated.
Despite this, there is a still a number of incredible details to take note of such as the generated shadow from the car, it gives an accurate shadow depending on where the sun is located.
There is also an impressive reflection in the rear window which moves accurately as the vehicle changes directions.
The GAN was trained to react by being fed multiple scenes from highways in order for it to learn how he car moves and responds.
“GANs can create images that look like photographs of human faces, even though the faces don’t belong to any real person,” Nvidia explains. “GANs achieve this level of realism by pairing a generator, which learns to produce the target output, with a discriminator, which learns to distinguish true data from the output of the generator. The generator tries to fool the discriminator, and the discriminator tries to keep from being fooled”.
We first saw GameGAN revealed by Nvidia last year around this time, it gave us a look into a clone of a fully functional game of Pac-Man and since has been used to create this impressive version of GTA 5.
Of course none of this has been coded by humans leaving the whole gaming demo created entirely by GameGAN.
This opens a door to huge possibilities in the future, but in the meantime GAN Theft Auto it currently available from GitHub.
Be sure to check back with us at Gamerficial for the latest news and updates.
Featured Image: GitHub/sentdex