Thank you for posting this. This is very good information.
I note that his actual words cited in the article sound less certain of his claim than what the headline implies.
The only way for me to simply describe the differences between the approaches to autonomy is GM believes brute Force hardware is required and Tesla believes brute Force software will work. GM is right on that their approach will take 10 years or now because hardware only doubles in terms of power per dollar every 18 months. Where as software really had no limits in how fast it can actually improve. Elon often points to Google and how it was thought it could never beat a good go player in 20 years and now after just a few years it can beat the 18 best go players at the same time. This level of acceleration can't happen in hardware like it can on software.
I don't know who is right, but I can conceive of a solution in my imagination of radar + gps + vision that would work given enough advances in machine learning and neural networks. Usually one of the hardest things is a good variety of driving data, but Tesla is gathering more then it can handle. If anything they need more and more processing power to handle the massive amounts of driving data.
Another thing to note, GM is handling redundancy in a way that may or may not be necessary. My point is that every system does not need 3 backups. If one system fails, the car just stops, you know like a car today would if it ran out of gas or the motor seized. Things on cars fail and the world does not end and rarely people get injured. From my most amateur assessment, Tesla is using overlapping redundancy from different systems to have good enough redundancy to have the car safely move to the side of the road safely in case something fails. For example if the vision system fails, the high def 3d maps and sonar + radar can get the car to the side of the road. Not ideal, but similar to what might happen if you have a tire blow out. If a system fails, the car is disabled to drive in autonomous mode and a driver has to fetch the car. It's not the end of the world even if the car has to come to a dead stop in the middle of a road and roadside assistance, which could be remote and have access to move the car to safety.
To be Frank this is not rocket science, and everything does not have to be perfect. If you think about a system that is good enough, not perfect, and a system that doesn't need to be redundant enough to work on Mars where you can't get there easily to fix it. If on rare occasion a system fails, which happens all the time for cars today, the car just needs to safely shut down. The system can be held to higher standard and monitored more closely to guard against failure. If your check engine light is on, you can't engage the system. If you have not had a break job in years, you cant use the system.
I am confident that no one knows how long it will take, but I'm also confident Tesla is taking a path that will work and if I'm right, it should take less time then other methods because software > hardware in terms of adapting quickly.