Generative vs. Discriminative Models

Everything we have built in this course so far—from the simplest Multi-Layer Perceptron to the deepest ResNet CNN—shares one defining characteristic: they are all Discriminative Models.

If we want to build an AI capable of generating deepfakes, writing poetry, or creating realistic art, we need a completely different approach. We need to shift our focus to Generative Models.

To understand Generative AI, we must clearly define the difference between these two paradigms.