By Brice Wallace
The likelihood of having a driverless car as a commonplace thing is, according to a recent discussion in Salt Lake City, not around the proverbial corner.
Autonomous systems will continue to advance but widespread use in such things as driverless vehicles will occur incrementally, according to a panel at the Utah Technology Innovation Summit, presented by the Utah Science Technology and Research Initiative (USTAR).
Dan Patt, CEO of Vena Robotics, said we are on the verge of a new revolution. “At the same time, on our day to day, our continuing time scale, this change is going to be gradual,” he said. “I don’t believe that suddenly in 2019 you’re going to go to all the car dealerships and you’re going to find these driverless cars.”
Panelists mentioned several obstacles in the way of widespread use, including humans’ comfort level with the technology, the technology needing to improve and the possible job losses or changes resulting from implementing the new technologies.
Devin Stewart, corporate general manager at Autonomous Solutions Inc. in Logan, said the implementation will be subtle. While autonomous systems can be used in certain industrial situations — he mentioned mining specifically — for more-general use, people will need to attain a certain confidence in the technology.
“Compare it to the cruise control in your car,” he said. “We’ve had that for a good while. It’s going to take time for people to start getting comfort in some of these driverless features you’re starting to see.”
John Hedengren, associate professor at Brigham Young University’s Chemical Engineering Department, said advancements in computers, algorithms and codes now allow technical systems to solve problems about 2 billion times faster than they did 30 years ago. As a result, several industries are being disrupted or are on the verge of being disrupted: cars; medical procedures; the purchase, shipment and delivery of packages; and long-haul trucking, he said.
“It’s a process for a variety of reasons, Hedengren said. “[One is] Regulatory. We have people factors — how much they trust the autonomy and whether they want to use it, and there’s that kind of confidence in the technology that gives the players in that area a license to operate.”
However, technology has yet to advance to the point of making appropriate decisions when something unusual happens, or to detect such a situation and know when to give control back to a person who could easily handle the situation, he said. As Google has developed an autonomous vehicle, it has classified 2 billion objects that the vehicle can identify. During testing, one it could not was a woman in a wheelchair crossing a street while being chased by a duck, he said.
Patt calls these situations “edge cases.” Many autonomous vehicle tests occur in ideal conditions but the systems have trouble adjusting to a deer, bicycles, bad weather or other unexpected-but-common situations.
“The whole community around self-driving is struggling with this right now. … There aren’t really good solutions to the edge case problem right now which are mature. There are ideas out there that are bubbling and that are promising, but this edge case problem is going to hold us back,” he said.
Such systems are great for automating highly repeatable actions but haven’t yet replaced human reasoning, he said. “There are problems which are trivial to people which are incredibly complex for automation,” he said.
He cited the example of a box that falls inside a warehouse. An employee on his or her first day on the job, even before being trained, could figure out what to do by checking out the box, determining if its contents are dangerous and putting it back where it belongs.
“This problem is a five-decade problem for automation,” Patt said. “There’s so much contextual reasoning about what to do … that it’s impractical to think about trying to automate that.”
Patt added that the human role will shift with the implementation of autonomous systems. A human can supervise the work of machines but can struggle to pay attention and quickly address the 1 percent of the time when things go wrong. “This is something that people just aren’t really good at,” he said.
That why, he said, machines will not simply replace people in the workplace. “It’s not as simple as just a machine doing what a person did. It’s a more-subtle change in that relationship and it’s going to take a while to sort this out,” he said. Despite all the “remarkable” advancements made in autonomous systems, “we are at a tiny fraction of the progress needed to replace humans,” Patt said.
Patt joined other panelists in predicting that autonomous systems will cause some jobs to be lost but also create new employment opportunities. Stewart said miners, for example, could be trained to help run or do repair and maintenance work on autonomous systems.
“The jobs are going to be changing,” Hedengren said. “It’s not going to necessarily be a taxi driver. Their new job might be something different, which is help train and classify ‘edge’ conditions. Or in retail automation, it’s going to be less about putting things in boxes and taping them shut and delivering them. It’s going to be about helping people make decisions about what they buy. Some of these things that need a personal touch, computers are not going to be very good at these things.”
Asked what the autonomous system world will look like in 25 years, Hedengren predicted it would improve the quality of life and present new opportunities for people throughout the world. Patt predicted “we’ll still be debating these questions” about automation and humans years from now.
“I don’t know if we’ll see that we’re at full autonomy as far as commercial vehicles and passenger vehicles and so forth, but we will definitely see that we’ve figured out some policies and regulations at that point and it’s prevalent,” Stewart said.
“And my kids will probably not fully understand how to drive a car, and maybe the ownership of a car is not even really necessary. You just have a pod running around and we just say, ‘You know what? I need to get from Point A to Point B. Come to my house and pick me up and take me there.’ That’s kind of what I see.”