This question is one of the most important decisions while driving and, in the case of humans, this is taken intuitively, since it depends on social interactions learned from an early age.
All over the world, the main automotive companies have focused their efforts on creating vehicles with autonomous driving systems. However, this technology has not yet been developed enough for it to replace the capabilities of a person behind the wheel.
One of the main limitations of this technology, according to recent research by scientists from the University of Copenhagen (Denmark), is that the artificial intelligence on which it is based is not capable of understanding the most basic social codes when it comes to driving, which allow drivers, among other things, to decide whether to yield or proceed.
This question, which might seem like a minor problem, is one of the most important decisions while driving and, in the case of humans, this is taken intuitively, since it depends on social interactions learned from an early age.
“Sorry, it’s a self-driving car!”
As detailed by the scientists in an article published in the Association for Computing Machinery journal, during their study they analyzed 70 YouTube videos, a total of 18 hours of images, of self-driving vehicles in different traffic situations.
In one of the clips reviewed by the team, a family in a residential area in the US is shown trying to cross the street in a place where there is no zebra crossing. When the vehicle operated by the autonomous driving system detected the people, it slowed down; however, passers-by waved their hands to signal the driver to continue on his way, a signal that was not understood by artificial intelligence.
After stopping for 11 seconds, when the family finally decided to cross the street, the car started up again, causing pedestrians to jump onto the sidewalk to avoid being hit. “Sorry, autonomous car!” apologized one of the passengers of the vehicle from the back seat.
According to Barry Brown, co-author of the publication, this situation highlights “the inability of self-driving cars to understand social interactions in traffic.” As he explained, although the car stopped so as not to run over people, it was close to hitting them, since it did not understand their signals.
For the expert, the main challenge faced by developers of autonomous driving systems is programming the ‘software’ to decipher socio-cultural codes. In his opinion, “the understanding of the social interactions that are part of the traffic should be used to design the [sistemas] of self-driving cars [que regulan la interaccion] with other road users.
If you liked it, share it with your friends!