Latest / AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence
How AI Is Learning Bias from Satellite Images
Lucas and Luna explore a surprising frontier of AI bias: satellite imagery. When computer vision models trained on overhead photos predict everything from property values to crop yields to crime hotspots, they absorb and amplify the human patterns embedded in those pixels — including redlining, segregation, and economic inequality. Lucas walks through a concrete case from a 2025 Stanford study that found AI models predicting neighborhood income from roof color and tree canopy, and how that creates feedback loops in insurance pricing and municipal services. Luna presses on whether satellite…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.