Confronting AI Bias: Joy Buolamwini's Fight to Overcome the 'Coded Gaze
Published on: March 10, 2024
While working on facial recognition software at MIT, Joy Buolamwini discovered that the program failed to detect her dark skin unless she wore a white mask. This incident introduced her to the concept of the 'coded gaze,' a term she coined to describe the embedded biases in technology shaped by those who develop it.
Buolamwini's research revealed significant biases in AI systems, particularly in facial analysis. She found that generative AI like Stable Diffusion often associates high-paying jobs with lighter-skinned men and criminal stereotypes with darker-skinned men, perpetuating harmful biases.
In her book, 'Unmasking AI: My Mission to Protect What Is Human in a World of Machines,' Buolamwini warns of the potential harms these biases can cause, especially if they reinforce existing stereotypes. She argues that AI systems often reflect a distorted view of progress, taking society backwards rather than forwards.
Buolamwini initially entered computer science to create futuristic technology but soon realized the need to address the misuse of AI. She emphasizes the importance of everyone's involvement in AI discussions, encouraging people to question the adoption and implications of AI systems in various settings.
The book also delves into the reasons behind AI's misidentification and misgendering issues, attributing them to non-representative training data and biases in AI development. Buolamwini highlights the impact of AI on real people, citing cases of wrongful arrests due to facial recognition errors.
As a 'poet of code,' Buolamwini uses art to humanize the impacts of AI biases, like in her piece 'AI, Ain't I a Woman?' Her work has influenced discussions on lethal autonomous weapons and AI ethics at high levels, including defense ministries across the EU.
Buolamwini urges President Biden and other leaders to focus on preventing AI harms and promoting biometric rights, emphasizing the need to protect individuals' likenesses from misuse and uphold civil and human rights in the age of AI.
She concludes by addressing the urgency of AI's current harms, stressing that the danger of AI is not just a future concern but a present reality affecting lives through structural violence and lack of access to essential services.