Will Large-scale Generative Models Corrupt Future Datasets?

Speaker: Hiromi ARAI
RIKEN AIP

DEIB - Beta Room (Bld. 24)
September 27th, 2024 | 10.00 am

Contact:  Prof.  Stefano Zanero

Abstract

On September 27th, 2024 at 10.00 am the seminar titled "Will Large-scale Generative Models Corrupt Future Datasets?" will take place at DEIB Beta Room (Building 24).

Recently proposed large-scale text-to-image generative models can generate high-quality and realistic images from users’ prompts. Not limited to the research community, ordinary Internet users enjoy these generative models, and consequently, a tremendous amount of generated images have been shared on the Internet. Meanwhile, today’s success of deep learning in the computer vision field owes a lot to images collected from the Internet.
In this talk, we will explore the potential impact of large-scale text-to-image generative models on the quality of future datasets and the performance of computer vision models, sharing our recent work.

Short Bio

Hiromi Arai is a Unit Leader of AI safety and reliability Unit, RIKEN Center for Advanced Intelligence Project (AIP) in Tokyo. She received PhD from the Tokyo Institute of Technology. Before joining RIKEN, she was a posdoc at University of Tsukuba, a researcher at RIKEN, an assistant professor at the University of Tokyo, and a Senior Researcher at NICT. Her research interests lie in privacy enhancing technologies, trustworthy machine learning including security, explainability, fairness, and human factors.