Diffusion Models - The Battle Between Generating and Decoding

There was an entire class of models that were actually built on this premise. They were called GANS, or I used to call them G-A-N. And it's fascinating because the premise of those models was to take the mechanism you just described and sort of capitalize on it. The thing that is magical is that I can type pretty much anything into a state of diffusion model and get it out.

Transcript

Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.