What Happened
In June 2014, the GAN framework was proposed: a generator network learns to produce samples, while a discriminator learns to distinguish generated samples from real data.
Why It Matters
GANs became one of the most influential families of generative models, shaping research on image synthesis, representation learning, and later—competitive, high-fidelity generation methods used in creative tooling.
Technical Details
GAN training is posed as a minimax objective between two models. Practical success depends on stabilizing training dynamics, architecture choices, and loss design.