Look, Ma, no hands Melissa Cefkin on self-driving cars
You have just arrived at a four-way stop, and you notice that the car to your left has no driver. Suddenly, it accelerates through the intersection and after a jolt, you realize that, yes, it was that car’s turn. As it turns out, the vehicle “knew” exactly when it should go. The challenge of the four-way stop is an example cited by Melissa Cefkin, who speaks about driverless or “autonomous” cars for the School for Advanced Research on Thursday, Sept. 14. “I come to a stop, and I’m looking to the left to see if the car that got there ahead of me is starting into the intersection,” she said. “At that point, without even thinking about it, I might let my foot off the brake, and that slow rolling of the car would be a signal to people that my car is about to go. An autonomous vehicle would have to decide or learn to take the action, or be told to remove the brake.” That situation is one of many that anthropologist Sefkin and Nissan Motor Corporation engineers are trying to anticipate in developing software for such vehicles. These cars will actually arrive at the intersection with an advantage — they will be equipped with systems that will have calculated, to a very precise degree, when everyone arrived at the stop.
But that doesn’t mean the car should just plow through, regardless of what the other drivers are doing. The autonomous vehicle needs to know when to go and turn and stop and not run into anything, or over anything — altogether, a fairly intimidating proposition. “It’s pretty awe-inspiring. There are the engineers who are very good at understanding and working with things like roadways, lines on the roads, and the visual images of regular kinds of things on the street. But the group I lead really focuses on the aspects of moving around on the road that have a social dimension to them, and they’re very subtle.”
She gave another example, one from a sociologist colleague. “If you’re driving along and there are cars parked on the right, and you see somebody on the sidewalk stepping out toward the street, an autonomous vehicle would see a moving object entering the road and possibly crossing its path. And that’s good because then maybe it wants to slow down or be careful. But if you saw someone standing and waiting on the other side of the car, you’d think, no, the person stepping between cars is going around to the driver’s side, which in the United States is on the left side of the car, so they’re probably just getting in their car.”
A human would make instant, almost unconscious judgments about that situation and, while exercising caution, would proceed with no need for drastic swerving or braking. Can machines handle such circumstances? That’s one of the tasks facing the Nissan Research Center in Sunnyvale, California, where Cefkin leads the human-centered systems practice. A Fulbright award grantee, she has a Ph.D. in cultural anthropology from Rice University. Nissan hired her a little over two years ago. What an idea: a car manufacturer employing an anthropologist. “The director of the lab here, whose area of expertise is artificial intelligence, works side by side with social scientists and strongly embraced the idea that autonomous vehicles need to really be developed through the lens of a very user-centered, socially-centered point of view,” she explained.
“The main thing that goes on here is the development of the software for driving the vehicle: how it will move, how it will operate, how it will make decisions. The autonomous-vehicle software is making decisions: It knows where it is, it knows where it should be, in this lane or that lane, it knows what it sees around it, and it then can make decisions and can command the car.”
The timing of Cefkin’s talk, “Walk, Don’t Walk: Everyday Interactions With Self-Driving Cars,” is
interesting. Domino’s Pizza planned to begin testing pizza deliveries in Ann Arbor, Michigan, in the last week of August, using a driverless Ford Fusion sedan bedecked with sensors, cameras, radar, and lidar. According to an Aug. 29 story in The New York Times, spotting a driverless vehicle is not uncommon in that city, where the University of Michigan has a pilot project in “connected-car technologies.” The Domino’s car will have humans aboard to monitor not only the automobile but pizza customers’ reactions. It’s an exploratory foray into the realm of product delivery; eventually a retailer like Home Depot might use an autonomous truck to transport building materials directly to a job site.
Even after the development and testing phases, self-driving cars using Nissan’s Seamless Autonomous Mobility (SAM) technology may operate with some kind of remote monitoring, and perhaps emergency control. “There are likely to be times when the vehicle would need help or would be able to do more if it had a remote monitor,” Cefkin said. “We worked with NASA, using the same technology as what’s used with interplanetary robots when they have to assist a robot that’s somewhere else to get over a boulder or whatever. The idea is an assisted tool. It’s not what we consider a drastic option, because nobody else ever fully controls the vehicle. It’s always operating autonomously, but what the SAM operator can do is provide it with more information and give it some new instruction to say, ‘This is what we think you should do in this case.’ ”
In the example of a pizza-delivery fleet, monitoring could function similarly to air-traffic control keeping track of airplanes. The vehicle software would incorporate real-time learning. “All the vehicles would be sending information to the Cloud and the SAM capability would get information from the Cloud, updated on an ongoing basis,” she said.
Questions of ownership issues arise — for example, whether autonomous cars will be shared instead of having one in your garage. “As an anthropologist, that’s the kind of question that really drew me in. One thing I really want to focus on at SAR is the way in which, as people move about, we move about through identification of notions of place. One of the things I’ll be addressing is place-making and experiences and the creation of meaning, how our experience of place is very experientially significant to our humanity and our way of being in the world. So I’m going to explore how autonomous vehicles might or might not sort of play into that, and disrupt or participate in the meaning of place or places.”
How long will it be before there are driverless vehicles all around us? “Not right around the corner,” Cefkin said, but she added that in some specified, defined areas and under certain conditions, it will be sooner. “One example would be what is referred to as the ‘last mile’ issue. Studies have shown that one of the problems for mass transit use is not that people don’t want to use it, but that they first have to get to the point of departure, or get back home from the point of departure, from the bus stop, for example. If you had a more flexible, viable means to transport people to the points of departure, that could be a great application.”