From facial recognition software that fails to recognize women and people of colour, to Twitter algorithms which seem to prefer white faces in photo previews, there’s been no shortage of negative headlines about artificial intelligence over the past few months. But whether we like it or not, AI is at work all around us, all the time, and there’s a growing movement to make sure it’s being used in an ethical way.
A new survey from RBC’s world-class artificial intelligence research center, Borealis AI, reveals that most businesses in Canada believe it’s important to implement AI responsibly, but 93% experience barriers in doing so, like cost, time, or lack of understanding. And barely over half of them have someone on the payroll who’s responsible for ethical data and AI practices.
On this episode of RBC Disruptors, host John Stackhouse re-connects with the head of Borealis AI and RBC’s Chief Science Officer, Dr. Foteini Agrafioti, to learn about a new online hub called RESPECT AI, aimed at making open source resource code, tutorials, academic research, and lectures available to the entire AI community. He also sits down for a frank and challenging conversation about the risks and rewards of AI with Ruha Benjamin, a sociologist and an Associate Professor of African American Studies at Princeton, and Saadia Muzaffar, a Canadian entrepreneur, author, and the founder of TechGirls Canada.
To learn more about Borealis AI, the work it’s doing, and its new RESPECT AI campaign, you can visit www.borealisai.com. RBC’s Thought Leadership Group has also published a new paper exploring ethical challenges in AI, which can be found at rbc.com/thoughtleadership. Ruha Benjamin’s website features a wide range of research and resources related to her studies on race, justice, and technology. For details on how Saadia Muzafffar’s non-profit organization is breaking down barriers for diversity and equity in science and technology, go to techgirls.ca.
You can hear previous episodes of Disruptors on the subject of artificial intelligence here: