The NHTSA website’s section on self-driving vehicles starts off with some compelling statistics: 36,560 people killed in motor vehicle accidents in the US in 2018 90% of motor vehicles accidents are caused by driver error Moreover, according to the website, today’s driver assistance technologies such as emergency braking and pedestrian detection are already saving lives. This is a fact that I believe. However, I disagree with NHTSA and believe that it is time to let go of caution and permit fully autonomous vehicles (i.e. Driverless self-driving cars should not be allowed on our roads and highways, without proper testing. Data Fundamentalism Kate Crawford from MIT, who studies AI’s impact on society, coined the term data fundamentalism. Data fundamentalism is the belief that computers can tell the truth. One of Yale’s professors took me to Belmont Park as a postdoc. He gave a lecture about handicapping before the race. All the New York pundits were squinting and scowling at each other. The professor then pulled out some green and white notes on computer printer paper. This was back in 1979),. It is a common misconception that computers always are right. This could prove fatal in the case of self driving vehicles. Apart from the fact that we tend to assume that computers always are right, there are good reasons for questioning whether self-driving vehicles will avoid accidents. Edge Cases in 2009, Captain Sully Sullenberger was just about to fly his aircraft into the sky when Canadian geese pounced on the engines. Sullenberger was just 2 ,900 feet from the ground and had only minutes left to maneuver the aircraft before it crashed into the ground. The pilots had no experience with this particular scenario and could not apply basic rules or common sense. They considered the probability of the passengers being able to survive various crashes, injures on the ground and where the rescue vehicles could be reached quickly, as well as other considerations. They landed heroically in the Hudson River. All 155 passengers were safe and sound. Although pilots are given extensive training, it’s impossible to prepare them for all situations. They must apply their common sense knowledge and reasoning abilities to deal with edge situations, which are similar but not identical to their training. Automobile drivers are the same. Clearwater high school student saw a woman have a seizure as her car moved. She pulled her car up in front of the female’s vehicle and stopped it. There were no serious injuries to the car. Many of us have experienced unexpected events while driving. Floods can make it difficult, or even impossible, to navigate the roads. Trees can fall and block the roads. A car is approaching the accident scene or construction area. The boulder that fell onto the mountain road was a rock. One section of asphalt is completely unlined. Black ice is something you might notice. When you attempt to climb an icy hill, your car may fishtail. Everyone has their own stories. Driving school does not teach us about every possible outcome. We use commonsense reasoning to anticipate actions and their outcomes. We know that children will run toward ice cream trucks when we see them in the neighborhood. If the temperature drops below 32 degrees and the hill is steep, it is likely that there has been precipitation. When we notice the vehicle in front of us turning, it is a sign that they are drunk or using text messaging. We then change our driving habits. We will be on the alert for other deer if a deer crosses our path. Our common sense knowledge says that deer travel together as families. When we see trucks with “extra large load” signs on their backs, it is important to maintain a safe distance from them and pass any vehicle. We slow down if we see a ball bounce onto the streets. A child may run in the street to pursue it. We can see large pieces of paper along the roads, so we are able to drive right over them. However, if there is a huge shredded tire on the road we will know whether we should stop it or not. Nobody knows how to put commonsense reasoning in cars, or computers. Their manufacturers only have two options because autonomous vehicles don’t possess the ability to use commonsense reasoning in these situations. The first is to use machine learning and data collected from human interactions with rare phenomena to create systems capable of handling each one individually. They can also try to predict every scenario possible and develop a program using input from the vision system to tell the car what it should do. Manufacturers will not be able to predict every possible edge situation. While it may work for slower-moving shuttles at corporate campuses, this is not possible for self-driving consumer cars. How can cars be expected to use commonsense reasoning in these situations? This will cause many more accidents. Autonomous Driving Cars Don’t See Like People. Computer vision systems can make incorrect classifications, and this could lead to them being fooled in ways people cannot see. Researchers found that even minor modifications to speed limits could make a machine-learning system believe the sign was 85mph. Some Chinese hackers also tricked Tesla’s autopilot to change lanes. These minor modifications fooled the cars, but not people. A bad actor could also devise ways to confuse cars and trucks so that they drive off the roads or over obstacles. There are many concerns beyond hackers due to the differences in how self driving cars see the world. Many Tesla owners reported that their cars often treat shadows (e.g., tree branches) as real objects in real-world driving. The Uber driver who killed the pedestrian was first identified by the vehicle’s object recognition software as an unidentified object. Next, it classified him as a vehicle and then as a pedestrian. I don’t know about others, but I wouldn’t want to be a pedestrian and driver on roads if cars can’t recognize pedestrians with 100% accuracy. Safety and Traffic Jams Unintentional accidents are one of the concerns. A driverless car competition was held in early 2020, Moscow. A vehicle stopped at a red light shortly after the competition began. The human drivers might have thought about the situation and decided to go around it. But none of these driverless cars could do that.
Home Innovation Can Self-Driving Cars Reduce Unintentional Accidents?
THE FOREFRONT OF TECHNOLOGY
We monitors and writes about new technologies in areas such as technology, innovation, digitization, space, Earth, IT and AI.