In-person Talk: Siobhan Mackenzie Hall and Fernanda Gonçalves Abrantes
Speaker(s):
Associated with:
We will be joined by Siobhan Hall and Fernand Abrantes, who will be presenting their research on 'Navigating data concerns in the age of large scale AI'.
Abstract:
Vision-language models (VLMs) are advancing rapidly and reaching ever-wider audience across numerous applications, such as image search queries, automatic captioning, and text-to-image generation. While there is reason to be excited about these advancements, there is reason to pause and question if these advancements are the same for everyone. In this talk, we'll explore underlying social biases that lead to representational and allocational harms, discuss where these biases arise, what can be done to counteract them and where the limitations in these methods lie.