Sunday, June 23, 2024

A GM Cruise self-driving vehicle in use as a taxi in San Francisco. (Photo: Cruise)

Over the past half-dozen years, we have been hearing stories about how robotic cars will be coming soon. They will be completely automated, with no steering wheel and no role for the human driver. During those years more than $300 billion has been spent worldwide trying to make the idea work. Now, according to a recent Reuters article, “industry executives and experts say remote human supervisors may be needed permanently to help robot drivers in trouble.” 

The original idea was that computers would take over the driving and result in fewer accidents, saving lives. Computers were assumed to be inherently smarter than humans. Now we are finding that even the engineers are saying a safe automated car will be impossible. We still need human operators who can take over in emergencies and make better and safer decisions. 

Experts have admitted that they are stymied by what they call “edge cases” – sometimes rare incidents in which robot cars get confused and do not know what to do. Humans have proven themselves capable of handling those situations better. So despite the $300 billion spent, we still need the special skills of human beings to bail out the computers. 

A key embarrassment occurs when robot cars get confused: “I don’t know what to do.” They simply slow down and stop – usually in the middle of the road. This is the “safety” strategy for all automated cars baffled by those confusing “edge cases.”

General Motors has invested more than $10 billion in its Cruise unit, and this year it plunged into providing robot taxi service in San Francisco between 10 p.m. and 6 a.m. every night. The first breakdown came within a week – the wireless communications link between the central computer and the fleet of 30 taxis quit. All the Cruise taxis came to a stop wherever they were at around midnight. GM service personnel (humans) had to come out and “rescue” each vehicle, sometimes with tow trucks. The chief executive of Cruise admitted: “I can provide my customers peace of mind knowing there is always a human there to help if needed. I don’t know why I’d ever want to get rid of that.” Reuters quotes another chief executive of a firm doing research into robot car problems that there are “tens of billions of potential edge cases” covering all roads and driving conditions. Thus there can be tens of billions of possible human judgment interferences. The message is that the task of designing a robot car is impossible.

A big problem occurs when weather interferes, especially heavy rain, hail storms, sleet and heavy snow. Despite all the R&D money invested, no one anywhere in the world has designed a robot car that can operate in a blizzard. They cannot handle heavy rain or fog either. The cars cannot “see” and thus they stop because it is a situation in which they cannot figure out what to do. Today’s robot cars are a flop in bad weather. 

What happens with human drivers? For decades they have been dealing with blizzards, usually with good results. In a snowstorm, drivers will slow down, peer past their windshield wipers and find a way to make it home. They may be tense but alert. Safety data shows that in snowy weather there are fewer injury accidents and a few more fender-benders. There should be special awards for all those drivers who magnificently achieve getting home safely, without abandoning their vehicles in the middle of the street.

There is more than just a good safety record for human drivers in snow. The important thing is that – unlike robots – those human drivers do not throw up their hands and quit. They stubbornly stick it out and find a way to get where they are going. Millions of them do it, every year. 

Meanwhile, we are designing robot cars that are a bunch of quitters. Human beings can be quitters too, but a snowstorm must bring out the best grit in them.

Back in February 1978, we had the great blizzard. A jackknifed truck blocked an exit ramp, and the traffic stalled southbound on Route 128. People were trapped in their cars all night. One truck driver had several boxes of oranges in his rig. He got out and went from car to car, giving out oranges. What computer would ever have thought of doing that? 

Investment in robot car technology has nosedived this year, and if the money dries up, the hope for robot cars dies with it. An experienced engineer in the field has admitted that robot cars are not as skilled as people because their “perception and prediction algorithms are not as good as how a human brain processes and decides … I am concerned that companies will rush to market without proving the safety is better than human-driven vehicles.” If the money dries up, there will be no market to rush into. Our cars – so often made in automated factories – will still be driven by human beings.

Some projects are well funded, such as Cruise. Its research and testing costs close to $2 billion a year, and the available funds could be burned up in two years. Ambitious forecasters have estimated the market for robot cars at $50 billion a year, but if they can’t come up with a solution to snow and heavy rain, the robot car could simply come to a dead stop.

The continued predominance of ordinary cars still leaves us with tens of thousands of people killed in traffic accidents every year. We still have problems of worsening congestion on our highways, especially as transit systems fail. Our auto-dependance contributes to problems of climate change, noise and air pollution. Think how different things might be if instead of automated cars we had invested $300 billion in better transit systems worldwide.

Welcome the victory of human beings over those “experts” who would try to use automation to denigrate human skills and spirit.


Stephen Kaiser attended the Massachusetts Institute of Technology in mechanical engineering and was a traffic and transit engineer for the Commonwealth of Massachusetts for 15 years. His doctoral thesis was on the topic of automated transportation.