This is all too easy. On Friday, I was taking the off-ramp to join a different highway, on cruise control (because I knew the road layout ahead), at the speed limit (70MPH). I glanced down to adjust the heater and when I looked up, a BMW had been sitting on the ghost islands separating the highway and the off-ramp (apparently missed the turn), and the driver had just decided to join the lane I was currently steaming down. Fortunately, the excellent brakes (as well as the loud horn...) in my Supra made up for my lapse in concentration.
I'm in two minds about vehicle automation. Most of it is to compensate for human idiocy (such as the above example). The same gains could be made by humans not doing illegal maneuverers like that. Yes, there are occasions where the electronics can react appropriately to a rapidly developing situation much faster than a human could (e.g. child/animal running into the road), but I agree with the article's caution on complacency. All it needs is for a driver used to auto-braking to get into a car without it and cause an accident, and the technology will suddenly become mandatory.
At the other end of the spectrum, as I've often said, it's technology developed by humans, based on our own understandings, so there will be flaws. Having experienced poor ABS systems that would activate at low speed to the point that I feared the car would not stop at the junction I was approaching, I am not confident auto-brakes will react appropriately in every situation they're called upon.
A lot of vehicular automation is compared to aircraft systems (especially fully autonomous technology) but there is a critical distinction - you never get tailgaters in the air (or if you do, something has gone terribly wrong). Aircraft can rely on huge amounts of automation because in normal use, there are huge amounts of open air in every direction to allow the pilot time to take back control should the systems fail, and still have space and time to correct the immediate problem. Vehicles on the ground are tightly grouped together. Not only does this leave precious little time for humans to react to a bad situation, it also means that any automation systems are similarly constrained. They have to work correctly in a span of seconds or milliseconds. There is no time to 'nope' out and pass control back to the human when they exceed their programming, as is frequently the case with Tesla. Failure is thus very difficult to analyse and account for.
Automation is generally a good thing, but I can't help thinking we might end up automating ourselves into a hole on this one.