Abstract: Hot on the heels of popular text-to-image Generative Art models like OpenAI's DALLE 2 and Midjourney, Stable Diffusion is an open source Latent Diffusion model that can be used to create images from just a few words. This talk will go through high level details of how the model was built and trained, and some of the precursors models that it built upon. Then it will move to some comments on the ethical implications of AI art models generally and ethical considerations for those using stable diffusion and models like it. The main portion of the talk will be covering with how to get started using Stable Diffusion. This will help those even with limited background in ML or deep learning techniques to get an understanding of practically how to work with the model. This section will also focus on tips, tricks and examples to improve your output images as well as hardware considerations when running this powerful model locally. Finally we will finish with a couple of examples of how this model can be used in a few specific workflows or applications and some examples of what future generative art models may be able to do.
Bio: Hunter Kempf is a Data Scientist working in the cybersecurity industry and a Z by HP Global Data Science Ambassador. In his free time he works on various side projects relating to Data Science and some of those projects end up as articles for his Medium blog. Previously Hunter worked as a Data Scientist at AT&T working on preventing Fraud and Security incidents and graduated from the Georgia Institute of Technology (Georgia Tech) with a masters in Cybersecurity and the University of Notre Dame with a masters in Applied and Computational Mathematics and Statistics.