Scientists, mathematicians and engineers continue to improve the algorithms for processing and generating data – machines learn faster and more efficiently, from time to time demonstrating their capabilities. For example, a group of researchers from Imperial College London and Samsung’s AI Research Center demonstrated the work of technology to turn a photo into video.
For clarity, a photograph of Grigori Rasputin was used, which suddenly sang in Beyonce’s voice. About 100% realism of speech does not go, but this fact does not make what is happening on the screen less impressive.
A technique called GAN (Generative Adversarial Network) is used to animate photos. Speaking very simplified, in the interaction of several AIs, one of them tries to “fool” the other, by finishing drawing the pictures until the second AI “believes”. The result depends on the volume of the “fed” information machine — obviously, the larger it is, the more natural the clone will come out.
For this reason, Albert Einstein was more “alive”:
And Rasputin sang Halo with “poker face”: