The issue with the statements from Tesla is that they seem to prioritize PR and obfuscation before actual relevant factual statements. The fact that the driver

The issue with the statements from Tesla is that they seem to prioritize PR and obfuscation before actual relevant factual statements. The fact that the driver received disengagement warnings prior in the drive is not very relevant to the accident per se. Might as well say the driver has received multiple disengagement warnings during his ownership of the Tesla. It is not relevant to the actual circumstances of the accident. Note that they do not say "he received disengagement warnings right before the accident and chose not to respond". Tesla instead says that his hands were not detected on the wheel 6 seconds prior to the accident. Okay, so thats the case, but obviously the Autopilot system allows that to happen, as disengagement warnings were not active at that time, therefore THE SYSTEM WAS OPERATING WITHIN ITS ESTABLISHED BOUNDS. This leaves it down to, the autopilot system drove into the barrier while the driver wasn't paying attention, but was operating the vehicle within the defined bounds of the semi-autonomous system, but violated the parameter of "pay attention to the road and be ready to take control". He did not receive any system warning before the accident, he just simply chose not to correct the system behavior before the accident happened. Everything else is just trying to kick up dust and obfuscate the facts of the situation it seems. The issue with Tesla's semi-autonomous system and to a certain extent, other systems, is that there is a poor boundary between system responsibilities and user responsibilities. This causes issues around the boundary where sometimes the autonomous system will accomplish something, but sometimes it can't and the user must. Because of this fuzzy boundary, I think its almost more important to invest into disengagement logic, equipment, and programming than pure self-driving technology. A good example of investment into this arena is actually GMs Super Cruise, which seems to have MUCH better user engagement detection and much sharper boundaries than systems like Tesla's