In 2014 Ian Goodfellow, a PhD student at Université de Montréal under Yoshua Bengio, presents at NeurIPS "Generative Adversarial Nets": two competing neural networks — a generator producing synthetic samples and a discriminator distinguishing real from fake — trained simultaneously to improve continuously. Born from a bar discussion in Montreal, GANs revolutionize generative AI, enabling photorealistic image synthesis, style transfer, and synthetic faces, but also opening the deepfake problem. Direct precursor to diffusion models (Stable Diffusion, DALL-E).