Self-driving cars are something many people are looking forward to but a machines ability to decipher light skin from a dark skinned could be deadly issue.
According to a new study “Predictive Inequity in Object Detection,” by the Georgia Institute of Technology, the skin color of a person is somewhat of an issue when it comes to self-driving cars, Vox reports.
So that means, if a dark-skinned person isn’t easily identified by an automated vehicle, as quickly as a white person, then people of color are at a great risk to get hit.
Automated vehicles, the study determined, may be better at detecting pedestrians with lighter skin tones.
The authors of the study wanted to study: “How accurately do state-of-the-art object-detection models, like those used by self-driving cars, detect people from different demographic groups?”
Detection by the automated system was five percentage points less accurate, on average, for the dark-skinned group, according to the report.
Kate Crawford, a co-director of the AI Now Research Institute said info like this is valuable since these companies don’t make this kind of research work readily available.
“In an ideal world, academics would be testing the actual models and training sets used by autonomous car manufacturers. But given those are never made available (a problem in itself), papers like these offer strong insights into very real risks,” she said on Twitter.
Fatal accidents have already hit the courts.
Prosecutors have ruled that Uber is not criminally liable for the death of Elaine Herzberg, a woman who was struck by a Volvo XC90 that the ride share company had been using to test its self-driving technology as she crossed a road in Tempe, Arizona in 2018.
The car’s back-up driver could still face criminal charges.
The National Transportation Safety Board is also investigating the crash.