Is it possible for AI to be racist? When Google's AI chatbot 'Gemini' launched a new text-to-image feature on February 1st, it received a barrage of accusations on social media for possessing an inherent reluctance towards generating images of light-skinned, Caucasian people. These accusations came after instances where the Founding Fathers were shown to be Native American and East Asian men, while the Pope was represented by an Indian woman.
The Senior Vice President of Google responded soon after that the bias was an unintended consequence of programming the chatbot to represent a variety of ethnicities. Should text-to-image models accurately reflect the existing prejudices in society based on training data? Or should they, as in the case of Gemini, artificially promote inclusion and diversity as per the prevailing value judgment at the time?
Join Rohan Pai and Ashwin Reddy in this episode of All Things Policy as they tackle these difficult, controversial questions.
A new course on Unpacking Urban Governance on OpenTakshashila is now accepting applications. Click here for more information.