The concept of a self-driving car reminds me of the concept of a space elevator: easy to think, impossible to realize. Despite that, both concepts have enjoyed huge publicity. The reasons for that are both simple and complex. But above all, the ultimate reason for such concepts is a whole new generation of engineers with an education founded on simulation. In simulation, everything is possible, including anti-gravity.
The idea of a self-driving car is appealing as it was Leibniz’s idea of an automated system that could make legal decisions free of human intervention and associated corruption. More than 300 years after the death of one of the greatest scientists ever, the concept of an automated legal decision system is still not realizable. The legal system is anthropocentric and its conversion to digital would require a simultaneous conversion of legal foundations to remove human judgment. However, a foundation for such transformation does not and cannot exist because of fundamental issues with logic. One example is the paradox of material implication and the fact that a proposition, regardless of its truth or falsity, always implies a true proportion:
True implication: If the moon is made of cheese, then Jack is mortal
Resolving the above paradox requires constructing a strict logic in which non-relevance between the antecedent and the consequence is absent. However, relevance is subjective as long as the truth nature of reality is not known. We do not even know what reality means. Quantum mechanics, the most accurate description of the world at the microscopic level, has led to more than a dozen descriptions of reality, including anti-realism, multi-universes, non-locality, randomness, uncertainty, you name it.
Automated systems can only be constructed by applying logic to a finite number of inputs from sensors. As the number of inputs gets very large, relevancy is lost because identification of the dynamics leads to high redundancy. We are stuck with the problem of the under-determination of theory from data: the inputs we receive are not enough to provide a complete framework for making concrete decisions. This is true in everyday life, in science, in the legal field, in medicine and also in economics, where the problem is more pronounced. Every decision made has a probability attached to it for being correct.
Therefore and due to the fact that drive-less cars will be subjected to many inputs from other cars, pedestrians, the environment, including the morphology of space and conditions, like rain for example, all decisions made will be of probabilistic nature other than under very simple conditions where driving takes place at constant speed along a straight road with just a few competing objectives trying to negotiate for resources, like the empty space in front of the car or a parking space. These are easier to handle during simulations or during actual tests when there is a human supervisor that can override decisions and correct for wrong actions. But what about cars driving in a crowded city under rain conditions with people running all over the place holding umbrellas and trying to catch a cab?
There is a whole generation of engineers that after the 1980s was educated through simulation. That was an extension of the early video game generation of the 1970s. It gave birth to concepts like the space elevator and the drive-less car. But such concepts lack a touch with reality. Reality, even if it is some kind of functional virtual reality, is still more complicated that the computer simulated reality that we show in the Matrix movie. There is no possible escape if something goes wrong and there is no anti-gravity to allow defeating laws of nature. Every mistake in reality is necessarily penalized. One could argue that human drivers also make mistakes. But until we can prove under actual driving conditions on a massive scale that drive-less cars can make fewer mistakes, the concept will remain a simulation dream of some geeks.
There is also a much more dangerous aspect of the computer simulation generation: They may attempt to force reality to conform to the requirements of computer simulations by placing everywhere sensors and turning this world into a chaotic network of data transmission. Then they can claim that their idea was sound. But that would be true only in the sense that they managed to change the state of reality. Yes, changing the state of reality is a sound idea. But it pays to do that only if the new state will be better than the previous, not only economically but also from the perspective of facilitating human evolution. If the objective is to change the state of reality so that drive-less cars can become possible but at the same time less functions of everyday life will be available to humans to deal with, then this is a goal that may be accomplished at the expense of human qualities, evolution and future prospects.