In the increasingly near future, the promise of autonomous vehicles dominating the roads leads us to consider the safety of this technology. A recent study from Kings College London has highlighted a worrying issue of bias in self-driving cars.
The study, which tested up to eight artificial intelligence (AI)-powered pedestrian detection systems used in autonomous driving research, produced disturbing results, according to gizmodo.
The researcherss analyzed more than 8,000 images using software and found that autonomous vehicle systems were almost 20% more accurate in detecting adult pedestrians than children. Furthermore, the effectiveness for detecting light-skinned pedestrians was more than 7.5% higher than dark-skinned pedestrians.
The study did not stop there. He also revealed that the AI had even more difficulty detecting dark-skinned people in low-light or no-light conditionswhich made it less reliable at night.
What do the experts think?
Put more bluntly, the study suggests that in the near future, crossing the street could become more dangerous for children and people of color due to the inaccuracy of detection systems in self-driving cars.
Jie Zhang, one of the study authors, detailed: “True justice in AI is achieved when a system treats privileged and disadvantaged groups equally, and this is not the case with autonomous vehicles.. Automakers don’t divulge details of the software they use for pedestrian detection, but since they’re generally based on open source systems, it’s pretty safe to assume they face similar issues of bias.”.
Sandy Karp, a spokesperson for Waymo, told Gizmodo: “The detection systems in our autonomous vehicles are designed to work safely in a variety of conditions and scenarios, and we are committed to continually improving technology to address the challenges it presents”.
This study highlights the importance of recognizing that algorithms can reflect inherent biases in data sets and in the minds of those who create them. In this context, even small biases could have life-threatening consequences, underscoring the need to address this issue before autonomous vehicles become the norm on the roads.