Aditya Ramesh (OpenAI) | Manipulating Images with DALL-E 2

Virtual London Machine Learning Meetup - 12.10.2022 @ 18:30 (BST)

Agenda:

• 18:25: Virtual doors open
• 18:30: Talk
• 19:10: Q&A session
• 19:30: Close

Sponsor: Evolution AI - Intelligent data extraction from corporate and financial documents.

Title: Manipulating Images with DALL-E 2

Speaker: Aditya Ramesh, Researcher at OpenAI

Papers:

• https://arxiv.org/abs/2204.06125

Abstract: Aditya will begin by motivating the hierarchical modeling approach called unCLIP that underlies DALL-E 2. Then, there will be a demo of three of the capabilities that it enables: inpainting, variations, and text diffs. These capabilities rely on the bipartite latent representation that unCLIP provides for any given image.

Aditya will discuss how images are mapped to this representation, and how it can be used to implement the capabilities shown. Finally, there will be a demonstration of how inverting the representations of a discriminative model provides insight behind its "stupid mistakes", as well as limitations of the unCLIP approach.

Bio: Aditya Ramesh is a researcher at OpenAI who works on generative models.