Last week TechCrunch Disrupt SF 2016 took place. It was by no means a surprise that the biggest announcements somehow all came from the car industry - the new autonomous car industry. The most cited announcement was the cooperation between Udacity and Daimler. It is understandable that Udacity wants to fully utilize this wave to feed its growth. The origin of the learning platform Udacity are in the AI / robots sector, which is a perfect fit for offering specialized courses. The cooperation with Daimler now gives them attention and more momentum.
Besides, of course, offering courses and learning material for everything related to autonomous cars they also announced to build a fully open-source version of a (truly auto) car software. This already asks more questions then answers. While I believe that we are shortly before a breakthrough in the technology for self-driving cars I also believe that the major issues are still to be tackled. We need a global set of rules that need to be followed by such cars. These rules need to be followed by any software that is allowed to be used in a car.
The rules should enable a common standard in difficult situations. Such a situation involves a potential accident where fatalities cannot be avoided. If the car still has the ability to decide "who to damage" it should be pre-determined to avoid blaming the company or even the individual programmers. Otherwise, this will become a society problem where programmers are regarded as the new military generals - ordering who should be killed. This should not be our goal.
Another thing is about logging and connectivity. What information should be logged? What should only be available (physically) on the car, what should be transferred? And who should the information be transferred to? We need some regulations here as well. Otherwise, most cars will log too much - send it to anyone who pays money for the information and hides / deletes sensitive data from the car. As an effect in case of accidents the information to reconstruct the scenario are lost or hidden behind the protected data centers of the manufacturer.
Udacity is at its core a technical MOOCs supplier. No one there (especially the students) will take care of these questions - and they are not at the heart of the current discussions. Therefore, I doubt that we have seen the peak already and the current hype is not justified and the crucial questions are still unanswered. Still, the self-driving vehicles will come and baby steps are already made. By 2020 we will have self-driving cars and taxis with reduced human supervision and control. By 2030 we will partially rely on this and by 2040 most people won't even bother to learn driving a car.