Google’s Self-Driving Cars Now Trained To Avoid Collisions With Pedestrians Who Matter More
DETROIT, MI — Critics of Google’s self-driving car project Waymo often like to pose philosophical questions: How can a driverless car make a moral decision? How would it respond to an unavoidable collision between one of two pedestrians? How can a computer decide who lives and who dies?
Waymo’s answer: With advanced facial recognition technology and deep integration with Google’s vast stores of user data.
“Look, we’re never going to be able to eliminate fatal accidents 100%,” said Waymo Chief Technical Officer Greg Pierce at the company’s annual tech preview in Detroit yesterday. “But thanks to powerful machine-learning techniques, we can at least be sure those fatalities will be people we won’t miss all that much.”
Pierce says the key is a highly accurate facial recognition algorithm that matches a pedestrian’s face to one of Google’s millions of user profiles. Within nanoseconds, another algorithm scans the pedestrian’s browsing history, bank statements, and social media to compute an Life Value Index (LVI) based on meaningful contributions to society. If the LVI is above a certain threshold, that will signal the onboard navigation to begin swerving safely away, hopefully into a less valuable individual or, if the passenger’s own LVI is low enough, a tree.
“In 2016, the number of healthy, attractive, productive Americans killed needlessly by automobile accidents was 4,836,” said Pierce. “If self-driving cars can help us turn even one of those tragedies into an acceptable loss of life, I think we will have done our job.”