Back in the day, games didn’t have the ability to adapt to players, or adjust the difficulty according to a variety of factors. Could you imagine? Mario? Platformed levels creating itself according to readings on the player, and likely ways to defeat them. Well now, thanks to GANs, or general adversarial networks, AI researchers have brought DOOM and Super Mario Bros new life.
General adversarial networks are comprised of a generator and a discriminator. The generator will create fake samples of training data, while the discriminator will try to recognize if the samples are real or fake. The two will continue to communicate back and forth, and the generator keeps learning more realistic ways to trick the discriminator. All of this information and AI development, will hopefully be used to create better games, and AI behavior in the future.