Latest / AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence
When Your AI Medical Imaging Misses Darker Skin Tones
In this episode of AI Ethics with Fexingo, Lucas and Luna explore a troubling blind spot in medical AI: diagnostic imaging systems that perform significantly worse on patients with darker skin. They examine a 2022 study from MIT and Harvard showing that three leading commercial AI systems for detecting lung disease had error rates up to 34% higher for Black patients than for white patients, largely because the training data was overwhelmingly sourced from lighter-skinned populations. The hosts discuss the root cause — dataset imbalance and lack of diverse representation in training images —…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.