Mar 22, 2024
Join host Craig Smith on episode #177 of Eye on AI as he explores the cutting-edge world of generative models in artificial intelligence with Björn Ommer, a visionary AI researcher and Head of Computer Vision & Learning Group at Ludwig Maximilian University of Munic.
In this episode, Björn delves into the fascinating inner workings of diffusion models, shedding light on the pivotal role these models play in advancing technology and society. Discover the groundbreaking development of stable diffusion, a key innovation by Bjorn's team that is setting new standards in the democratization of AI technology.
Learn about the challenges and solutions in making AI accessible on consumer hardware, ensuring that the power of generative AI is not confined to well-funded organizations but available to all. Bjorn shares his expert insights on the balance between open-source innovation and proprietary development, emphasizing the societal implications of generative AI.
This episode is an essential listen for anyone intrigued by the potential of AI to transform our understanding of the visual world, the importance of accessibility in technology, and the future of AI development.
If you're captivated by the evolution of AI and the profound impact of generative models like stable diffusion, don't forget to rate us on Apple Podcast and Spotify.
Stay Updated:
Craig Smith Twitter: https://twitter.com/craigss
Eye on A.I. Twitter: https://twitter.com/EyeOn_AI
(00:00) Preview and Introduction
(01:38) Bjorn Omer's Journey into AI
(03:03) The Evolution of AI Before Deep Learning
(06:38) Democratizing AI with Stable Diffusion
(09:45) Explaining Diffusion Models
(13:40) Challenges of AI on Consumer Hardware
(17:05) The Binding Problem in Vision Research
(22:27) Mechanisms in Stable Diffusion: Attention and Diffusion Processes
(26:09) Open Source vs. Proprietary AI Models
(30:16) Making Compute Resources More Accessible
(34:30) Reducing Compute Requirements in AI Models
(38:52) The Concept of World Models in AI Research
(41:50) The Future of AI and the Fallacy of Scaling