In the near future will not be able to find the differences between a real and a computer-generated image.
In this paper the authors of the Cornell University describe how, the recent advancement in the field of Machine Learning, created a set of algorithms capable to generate images with some desired characteristics.
More in details, the authors show how, starting from an image, it is possible to use neural network to modify such image in order to change some properties. One example is given by the possibility to modify a photo taken in the morning in order to look like a photo taken at night.
Those results open new perspective in the image processing field. Indeed, those algorithms allow to create realistic scenario minimizing the human effort needed for the post-production.
More info on https://arxiv.org/pdf/1703.07511v1.pdf