How to Deceive a Robot
How do you fool a robot? It’s not difficult—machines are only good at what we tell them to do. But even still, parts of our world may not be “readable” to some robots. The fatal Tesla car crash underscored that point when its Autopilot system was unable to see a white tractor-trailer driving across the road, causing the car to crash.
A white tractor-trailer may seem obvious to us, but what humans might perceive as a visible world, sometimes confound robots.
Writer Geoff Manaugh interviewed robotics expert John Rogers for New Scientist. In the interview, Rogers points out how robot’s perception differs. Robots do not have eyes, they are equipped with lasers and cameras to sense their environment. “Rubber absorbs light and prevents laser-based navigational systems from relaying spatial information back to the robot.”
For Tesla’s Autopilot system, the MobilEye sensor was unable to detect the truck. “It is also worth noting that the camera they use sees only red and gray intensity, it does not see all the colors, making it have an even harder time with the white truck and bright sky,” Brad Templeton wrote in his blog.
“One has to be very careful in creating machines to not assume they’re more capable than they are,” says theoretical physicist Lawrence Maxwell Krauss. “That’s true in cars. That’s true in vehicles that we make. That’s true in weapons we create. That’s true in defensive mechanisms we create. And so to me the dangers of AI are mostly due to the fact that people may assume the devices they create are more capable than they are and don’t need more control and monitoring.”
Reflective materials, like mirrors also pose a challenge for robots. “It actually appeared that there was a virtual world beyond the mirror,” Rogers told Manaugh.
There’s this idea that the very nature of our homes may have to be rethought in order to make them more accessible to robots. It’s not out of the question to even think our cities will be reshaped by robots in order to make them more efficient, making us more efficient.
Photo Credit: Keystone/Getty Images