AI + Machine Learning Articles

Diffusion

• The Annotated Diffusion Model - Some of the deep nuts and bolts behind Annotated Diffusion Models, aka Denoising Diffusion Probabilistic Models, aka DDPMs, aka autoencoders.
• A Recipe for Training Neural Networks - some good takeaways on patience, why training neural nets fast doesn’t work.
• dreambooth - an influential paper demonstrating how to create a model which is capable of synthesizing a personal subject (such as an animal or item) in different contexts. An expansion on existing generative techniques allowing us to generate images of a specific subject.
• Illustrated Stable Diffusion

Misc.

• Prompt Injection
• Scaling Laws - a good overview of the scaling laws behind LLMs, like for example, when can a model learn to morally self correct? (the answer is 22 billion).
• DragGAN - crazy drag image editing.