New research suggests robots could turn racist and sexist with flawed AI
2022-06-22
The Indian Express
The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Researchers from Johns Hopkins, Georgia Tech, and University of Washington found that robots using the popular CLIP AI model consistently displayed gender and racial biases, favoring men over women and white people over people of color. The study highlights real-world risks of deploying AI systems trained on biased internet data.[AI generated]
Why's our monitor labelling this an incident or hazard?
The robot operates using an AI system (CLIP neural network) that processes images and commands to perform tasks. The study shows that the AI system's learned biases cause the robot to act out toxic stereotypes, such as favoring men over women and white people over people of color, and making inappropriate associations (e.g., criminal or doctor) based on appearance. This is a direct manifestation of harm caused by the AI system's outputs influencing robot behavior, which can lead to discriminatory outcomes in real-world applications. The harm is realized in the robot's biased actions, not merely a theoretical risk. Therefore, this qualifies as an AI Incident due to the direct link between AI system use and discriminatory harm to communities and individuals.[AI generated]