I generally agree with most points about AI(artificial intelligence) in the Stanford One Hundred Year Study on Artificial Intelligence. However, I can not accept the authors view about the future application and good performance of AI in transportation or self-driving cars. More specifically, the authors think that when self-driving cars have a higher speed and better performance, these cars will cause great changes to the whole society. I think the authors are a little bit over optimism in the application of self-driving car, and they did not consider the problems brought by human drivers, emergency to AI system, and uncertainty on the road.

Driving is not only a skill to human beings, driving can also becomes a hobby to them. Even with the arrival of self-driving car age, in which AI takes the place of most human drivers, it is impossible to extinct those peoples interest in driving, or even forbid those to drive. There will still be a considerable amount of human drivers in the future. The existence of these human drivers will limit the performance of self-driving cars, especially their speed. Although the maximum speed of self-driving car could be higher than that of cars driven by humans, self-driving cars should not reach that speed when driving on a road with other cars driven by humans, because that speed could be too fast for human to react and could cause traffic accident. Therefore, the speed will not increase and people will not save time on the road, not like what the authors said in the essay.

In addition to the limitation brought by human drivers, long term self-driving by AI will humiliate humans driving skills. This kind of change will not cause safety problem when the AI system keep working properly. However, there is never a perfect program. When the AI system in the self-driving car get something wrong, and cannot fix it by itself or get connection to get remote control, it is not a good idea to just stop the car, an action that is even more dangerous, so it make sense to switch its control to the human driver in this kind of situation to keep running. Then a more serious problem comes, with a long duration of AI-driven trip, the human driver in the car is more likely to be a passenger rather than a driver. That means he or she gets unfamiliar with driving car by hisself or herself. In this way, making human driver to take control of the car losing AI control increases the safety risk to both the driver in the car and other cars on the road significantly. This could be a hard problem for AI system to choice: whether it should switch the control to human drivers, when it is sure that it has something wrong in its problem.

Besides the problems related to human drivers, self-driving cars also have problems to deal with uncertainty on the road. It is true that machine learning gives intelligence to the AI system (numerous data and information are send to AI and make it know how to deal with most of situations on the road), but such kind of intelligence is still very limited. There are still lots of situation are not included in these data and information, which can be called uncertainty on the road. When meeting uncertainty, AI will face the previous problem, whether it should give control to human drivers.

I think all these questions will challenge the future AI application in transportation.
