OpenAI's improvements to contemporary diffusion models see them beat the state-of-the-art generative adversarial networks (GANs) in both conditional and unconditional image generation tasks.
Generative adversarial network RSS
Selim Seferbekov won Facebook's Deepfake Detection Challenge. His model achieved a 65% average precision on the test dataset. None of the participants hit 70% despite good training accuracy.
In collaboration with the University of California, Nvidia researchers have introduced a deep learning-based model, the AI Choreographer, that generates choreographed dance moves based on input music.