Artificial Intelligence: dueling neural networks or Generative Adversarial Networks GANs

The generator takes simple random variables as inputs and generates new data. The discriminator takes “true” and “generated” data and tries to discriminate them, building a classifier. The goal of the generator is to fool the discriminator (increase the classification error by mixing up generated data with true data as much as possible) and the goal of the discriminator is to distinguish between true and generated data. 

GAN is an already mature & usable technology, but more progress is expected in the coming years. The main uses are: fake data generation (images, text, video,...), fake data detection, automatic data classification. This technology could have an important impact on society, being used to generate "deep fakes“ for example.

It can either be used as a "base brick“ in classification solutions (e.g., detect asset defects from photos) or to generate dummy data to avoid transferring "real data" protected by GDPR for example. Ethical and policy issues still need to be resolved.