What is critical is to consider further context and its unclear if the interpretation of the image by itself would cause the AV to make an incorrect decision. It's part of a large amount of information that is consumed in the decision-making process I assume - again similar to humans.
That's exactly what was achieved.
Disclosure: I work at Cruise (a self-driving car company)
https://www.wired.com/story/cadillac-super-cruise-ct6-lidar-...
https://www.cadillac.com/world-of-cadillac/innovation/super-...
Meanwhile there's ~10 other car manufacturers out there making lane-keeping systems that use cameras, and yet people only care about bashing on Tesla.
Tesla is the one making the big claims and drawing attention to themselves. How many others are claiming full autonomy is just around the corner? They bring the publicly on themselves and can't have it both ways.
https://electrek.co/2019/03/29/tesla-full-self-driving-compu...
When combined with the big neural link announcement that was supposed to happen weeks ago, you may be able to monitor autopilot while you sleep by having it tap into your dreams. They are also going to give us point-to-point earth rocket travel for less than a first class ticket within 8 years.
I'm joking about neural link, though it did miss an announcement, but I'm not joking about the rocket one.