Fixing the bAIs.

AI image generators tend to picture professionals as men: Aurora50 cites DALL-E 2 producing images of white men 97% of the time when asked for a CEO or director. The National reported the campaign's point that women are under-represented in the data used to train AI, which leaves the systems biased against them.

The team used Midjourney, DALL-E 2 and Stable Diffusion to create a large databank of diverse images of women in many professions, removed gender from their file names, tags and descriptions, and published them royalty-free on fixingthebias.com for anyone to use; the agency's award entry says the site fed them to stock image banks, social media and the open datasets used to train AI. Workshops for young people, summits and a Discord community added more images, according to the agency's award entry. Aurora50 presented the campaign at its Inclusion Summit in Abu Dhabi in November 2023 (The National).