GANs a tool for image production utilized to produce Text in a automated chat bot system
GANs uses the concept of adversarial training with 2 neural networks i.e generator and producers where they both enter a min-max game such that the genertor tries to minimize the difference between original and produced images and the dicriminator tries to maximize the differentiating factors between the real and produced fake image.
GANs produce highly realistic images where the discriminator networks cant discriminate whether the produced images are real or fake (produced by the generator network)
This concept work nicely with Images but fails with text data as we need to produce human like text responses from the machine where we cant differentiate between machine produced or human produced text responses.
thus enters the concept of reinforcement learning , which helps machine to decide human like responses.
more details will be updated shortly.