I'll be picking up my P100D AP2 car in a couple of months. I'm hoping that EAP/AP2 will be pretty much the same as AP1 by then, and I can just forget about the details about AP2 vs AP1... and just use it the same way as I've been doing for the past year or so on AP1 cars.
Time will tell - though I have to admit, from reading the anecdotes of people using AP2, I'm starting to wonder if Tesla is suffering from the
Second System Effect. Of course, I know that sometimes you have to take a step backwards in order to leap forward.
However, it's hard to feel super confident right now... they fall out with Mobileye, take a different approach entirely, then fall out with the guy running the Autopilot program. Mobileye and Sterling Anderson are extremely smart guys... luminaries in the self-driving industry. The Chris Lattner + Nvidia + Tesla Vision combo is so new that it's hard to know if they can achieve the same result as AP1 at all, since it's such a radically different technical approach, with a complexity level several orders of magnitude above tightly tuned lane finding and headlight recognition.
I guess they surprised themselves with how well they got AP1 to work... It's difficult to sort of 'forget' about what they did with AP1 and reset the 'wow that's cool' barometer with AP2's advancement so far. I suppose following the road slowly, some of the time, under extremely careful human supervision is still remarkable for a car - it's just seemingly not much of an advancement considering how well AP1 currently performs.
Interesting times; exciting times; risky times! In fairness, they're also usually the best times