MIT professor and IT expert Erik Brynjolfsson says that self-driving car technology has developed faster than he expected.
“It caught me off guard how quickly this technology caught on,” Dr. Brynjolfsson said Thursday during a panel discussion at Council on Foreign Relations headquarters in New York, where he and two other experts—Intel Corp. scientist Jennifer Healey and consultant Chunka Mui—engaged in an often spirited debate about how the technology will evolve.
Not so long ago, Dr. Brynjolfsson said that self-driving cars were beyond the capability of computer science.
“About 10 years ago, I was teaching a class at MIT. One of the topics of discussion was what machines can do and what machines can’t do. One of my examples of things that machines can’t do was drive a car,” he recalled. His reasoning, he said, was that computers were good at structured tasks, but that driving involved too much unstructured information, such as visual data and the rules of the road, for computers to adequately handle. A 2004 book, The New Division of Labor, by Frank Levy and Richard J. Murnane, explored some of those themes, too.
In 2012, however, Dr. Brynjolfsson found himself in a fully automated Google test car, making the trip from Google’s campus to San Francisco and back. “It was humbling,” he recalled. “When I got to ride in a driverless car, I was surprised at how effective it was.”
At some point during the trip along Route 101, a vehicle in front of the Google test car came to a full and unexpected stop, “It was a full stop right in the middle of the highway. I hoped that the Google car noticed. Fortunately, it did,” Dr. Brynjolfsson said.
During the ride, he said, he experienced three stages of reaction, including a few minutes of fear, five or 10 minutes of exuberance, and a long period of boredom. “It was a bit like watching a dishwasher. It drives very carefully, and of course it obeys the rules. I think it is a microcosm of how society will react to it.”
“I think you all agree this will happen,” a member of the audience said.
“We disagree on what this is,” Dr. Brynjolfsson replied. “I think we agree something big will happen. We just don’t agree on what will happen.”
“This is our primary disagreement, the timeline,” Dr. Healey said. “I don’t think policy changes that rapidly. I think it is going to be slow adoption. I think it is going to be gradual adoption.” Some of those policy issues include new ideas about safety and liability, as well as the government’s encouragement of connected car technology, which links vehicles with mobile communication technology and allows them to exchange information with one another and with transportation infrastructure. Dr. Healey said she thinks vehicle to vehicle technology “could be very important.”
She said that the IT in cars “is fabulous technology, but it can’t see ahead forever. But information from forever ahead can be sent back to the car or cars, through the cloud, or they can talk to one another.”
Mr. Mui, head of the Devil’s Advocate Group, suggested that the policy debate would catch up with the technology, instead of the other way around. He said a long debate over vehicle to vehicle communications, followed by an even longer deployment, would slow down the introduction of self-driving cars, a technology that he said had the potential to save thousands of lives a year by reducing traffic accidents.
Mr. Mui said that if deployment of self-driving vehicles is dependent upon vehicle to vehicle communications, “you will never see it happen.”
