Autonomous cars are taking over the roads worldwide, with most newer models and top-notch vehicles being equipped with algorithms-based tech, offering humans greater convenience. But the new study concluded these self-driving cars are a danger for people with a darker complexion.
AI systems are replacing human drivers, resulting in canceling out most of the road rage and accidents caused by human error. But not to forget, these algorithms are also designed by humans, which leaves them with the margin of error of their own.
While autonomous driving systems are a tool of great convenience, they still bag their share of flaws. And it would be a total nightmare if your smart car kills a person with a darker tone walking on the road.
Having a vehicle that drives you to places while you use your phone for entertainment or meet a deadline using your laptop, with zero effort and attention required to control the car, sounds like a fair deal. However, the algorithms that they are backed with have a margin for improvement. Algorithms learn from the instances they are given, and surely they haven’t been made familiar with diverse populations.
UK’s Law Commission is working to build a framework for autonomous vehicles, stating it is suspected that self-driving cars might fail in recognizing dark-skinned faces in the dark, alongside posing risks to the disabled persons in the absence of ways to deal with the full variety of wheelchairs and mobility scooters.
Researchers found that AI systems use two different categories on the Fitzpatrick scale to classify human skin tone. It was revealed that the system’s efficacy was decreased by 5 percent when exposed to people with a darker complexion.
The study calls for developing unbiased algorithms to back the AI systems. To attain complete control on detecting all skin tones, the algorithms need to be trained with more images of dark-toned individuals. It is hoped that in the future, developers of autonomous cars and AI systems would include data of diverse populations while designing the systems that will eliminate the bias and decrease the levels of risk for darker tone individuals crossing the road.
