AI systems, like language models, are based on historical data coming from databases that often contain biases. The model will then reproduce these biases as it learns, which can create huge problems.
One person who’s done great work here is Timnit Gebru, an AI researcher specializing in algorithmic bias and data mining. She is the founder of the Distributed Artificial Intelligence Research Institute (DAIR), a community-based space for AI research.
She advocates for greater diversity in technology, and has shared how the underrepresentation of Black people in the industry leads to discriminatory technology with gender and racial biases. In this TED Talk, she offers insight into how AI technology can discriminate against communities and what it will take for Black people to be better represented.
Dr. Joy Buolamwini
It started as an art project and a dream of becoming Serena Williams.
As a grad student at MIT, Joy Buolamwini wanted to make a mirror that would project a digital image of one of her heroes onto her face. But for some reason, it didn’t work when Joy – who is Black – looked in the mirror…not unless she wore a white mask.
This was the experience of a young Dr. Joy Buolamwini, whose research has started a movement that is now challenging the tech world. Joy discovered the bias baked into artificial intelligence systems because they are trained on datasets of mostly white men’s faces: it can recognize caucasian males, but not so easily faces like hers.
Her research and discoveries led to a greater awareness of bias in AI technology and a demand to address it. In this short TED Radio Hour Comic, you can find out more about her fight against algorithm bias.
Realizing that the problem went beyond art projects, Joy founded the organization Algorithmic Justice League, which is dedicated to educating people about AI bias and how to reduce it. In the documentary Coded Biases, she exposes the biases in AI technology, particularly facial recognition technology.
“I’m Jordan, and I like to look at all the different ways that we interact with artificial intelligence on a daily basis.” This is the message you’re met with visiting Jordan Harrod’s YouTube channel.
Jordan Harrod is a Ph.D. candidate in medical engineering and medical physics in the Harvard-MIT Health Sciences and Technology Program, where she studies machine learning and neuro-engineering.
She uses her YouTube channel to talk about emerging technologies and, among other things, issues relating to fairness, bias, or racism in artificial intelligence or algorithms more generally.
In this video, she discusses the ethics of AI art using Lensa’s viral AI as an example.