> "Stable Diffusion was trained primarily on LAION-Aesthetics, a collection of subsets of the LAION 5B dataset, which is itself a subset of LAION-400M" Err, no

> "Stable Diffusion was trained primarily on LAION-Aesthetics, a collection of subsets of the LAION 5B dataset, which is itself a subset of LAION-400M" Err, no? As you can tell by the name itself, the 5B dataset is larger. > "Unlike autoencoder-based deepfake content, or the human recreations that can be achieved by Neural Radiance Fields (NeRF) and Generative Adversarial Networks (GANs), diffusion-based systems learn to generate images by adding noise" This is confused. Diffusion is orthogonal to NeRF. For example, here is a paper that uses both: https://arxiv.org/abs/2112.12390 https://arxiv.org/abs/2112.12390 > "Within days of release, the open sourced Stable Diffusion code and weights were packaged into a free Windows executable" That's not how it became popular. > "Additionally, at the time of writing, Google Research has just released a similar system called DreamBooth, which likewise ‘tokenizes’ a desired element into a distinct ‘object’." The approaches are actually very different. DreamBooth uses fine-tuning, unlike textual inversion.