Self-driving cars are around the corner. Literally. Last month during a visit to Pittsburgh, PA, we had the opportunity to see an Uber robotic car cruising the streets of downtown and transporting humans as passengers. Companies such as Google, Tesla, Uber, and Volkwagen have built and are testing their autonomous vehicles prototypes in a few cities around the world.
However, it would take some years until robotic cars populate streets and highways and become a form of massive transportation. Despite the development of GPS, computer vision, and artificial intelligence, there are several problems for robotic cars that keep scientists and engineers busy. Going over bridges with few environmental cues, outdated maps, navigating roads without clear lane markings, dealing with inaccurate GPS signal, and unpredictable humans, for instance, are important limitations. The latter, perhaps, is one of the most complicated barriers that artificial intelligence would have to overcome.
Human action that is unpredictable can break the flow of a networked human-machine system. Bugs and errors can appear as humans enter inputs that the system does not expect. Minimal errors in an a system composed of networked computers, sensors, robots, humans, and things, have the potential of affecting multitudes. Other times, these errors can become just ephemeral disruptions in the everyday life of a couple of individuals. In both cases the unpredictable human behavior and its disruption of interconnected systems can generate comic and absurd situations.