Which method is more likely to have implementation errors?
However in the real world where the car/sensors gets dirty/wet, and the air is filled with snowflakes, raindrops, or mist it gets much more complicated.
Even if statistically better there's value to acting more like a human driver. After all the roads are filled with human controlled cars and any deviation from human norms because of weather(like wind, snow, rain, fog, and blown sand) cause problems.
Machine learning is great at complex algorithms if done well.
Humans are very good at estimating distance from a combination parallax visual cues and experience. We don't need to have seen a specific model of car before to judge how far away it is with a high level of accuracy.
And humans aren't good at estimating distance. They're good at estimating relative distance which is different. Some people claim we need LIDAR because cameras can't give us accurate depth information but humans don't use accurate depth information.
Are you saying these systems need to have seen a specific model of car in the past before they can determine its distance? Certainly I've seen systems that do not require that information.
Personally, I view each additional layer (network, someone else's data center, machine learning) to be something that can fail and put people in danger. Assurance via local brute force is much more reassuring for me.
I should add a disclaimer: I'm waaaaay out of date. The last time I worked on this stuff was back in 2005 during the DARPA Grand Challenge. And while I'm fairly confident that the basics are still the same (the fact that we're still discussing monocular vs. stereo vs. LIDAR is testament to that...), I'm going to defer to someone more up-to-date to provide more recent research/documentation.
These vision systems do OK on things that look like typical road obstacles - cars and people. Other stuff, not so much. Only Urmson from Google talks about that much. The videos from the vision systems don't seem to be doing much with road edge clutter. They're not marking every post, fireplug, and low-hanging branch. Train a deep learning system on cars, trucks, pedestrians, and bicycles, and your system sees only cars, trucks, pedestrians, and bicycles.
These vision systems do a lot less than many people think they do. Watch the videos.
[1] https://www.youtube.com/watch?v=GZa9SlMHhQc [2] https://www.youtube.com/watch?v=2NGnvGS0AtQ [3] https://www.youtube.com/watch?v=Uj-rK8V-rik
As it turned out, our mechanical engineering was great. What got us was software: we failed to free memory for passed obstacles, ran into memory exhaustion issues as a result and crashed out at mile 9. When we fixed the leak and reran the course, the truck finished in just over 7 hours.
It was an object lesson that garbage collection won't save you from memory-related issues. In retrospect, it was the also first time I ever encountered a million dollar bug. D'oh.
[0] http://www.cs.toronto.edu/~urtasun/courses/CV/lecture01.pdf