AP2 may have forged ahead, possibly, in some driving scenarios, but the jury is still out whether or not it is as reliable. See e.g. here:
As for feature parity, the test is pretty easy:
Where is AP1's speed-sign recognition in AP2?
Where is AP1's vehicle type identification/display in AP2?
And, arguably related, where are auto-wipers in AP2?
Does AP2 have some feature AP1 does not?
Appreciate your continuing to keep those at bay who would spread the falsehood that AP2 is at collective parity with AP1.
Just because both may manage a vanilla use case most of the time does not parity make.
Lost in the debate are also a couple of salient points from 50,000 feet. Here’s just one:
*Neither* AP1 *nor* AP2 handle lane-shifting at all. Specifically, if I’m on a 2-lane state route with oncoming traffic, I’d prefer AP (AS, actually) to shift to the right (within the lane) to both increase safety margin and to decrease rock chips - especially with the refresh which seems to be a windshield magnet for rocks. As well, if I’m on some generic freeway overtaking a large truck (me on the left, them on the right), I’d prefer the car to shift to the left (within the lane) to do likewise (and to reduce truck lust).
So while you’ve got a small group ignoring the reality of all the things AP2 doesn’t have relative to AP1, no matter the countless examples given by those who have actually owned/driven both AP1 and AP2 (versus all of the armchair jockeys), I’d suggest the larger issue is that there’s room for improvement for AP in general.
Now pretend you’re Tesla. You’ve got your own board (SoC) under development, neural networks (plural) learning every day, and a need for many more miles driven to satisfy regulators (solved by having many more cars on the road, which is to say by selling more cars).
Well, they’re doing all of these things. And in time, AP3 will exceed AP1. And it won’t be a big deal to retrofit the relatively few AP2/FSD cars with a new board, exactly as Elon has said.
Doesn’t change the fact that *today*, AP2 is not at parity with AP1, nor the fact that it would still, without immediate intervention, drive into medians in Sepulveda and over lane markers on the 405 and 101.
The fact that it will drive 100 miles up the 5 without wandering off the road isn’t particularly compelling. It did that over 2 years ago. Well, after a couple of patches.
Anyone want to take a stab at when we’ll have stop sign reaction (versus recognition) as seen in the late 2016 video? I didn’t think so.
Also, here’s one for the apologists: a case could be made for AP2 cars never recognizing and reacting to speed limit signs. AP1 owners may relate - everyone else will have to play along. Imagine you’re on a freeway not in the right lane adjacent a speed limit sign. As you pass the sign, a big truck is tweenst your car and the sign. So no reading of the sign and no associated adjustment to speed/TACC.
Now pretend you’re a Tesla engineer and think, “Hey, let’s load a current database of geotagged speed limit signs and use that instead. Why, some people won’t notice the difference because it’ll just work.”
By and large, it’s a great idea *except* when there are errors. And boy, are there errors.
Back to Tesla Engineering: “We’ll just update the database, or better yet farm it out to a 3rd party, and periodically update it going forward. Still better than blocked cameras on the freeway.”
And there you have it. Now, rumor has it they’ll get back to reading speed limit signs, but perhaps with database backup for those times when there’s a big truck in the way.
In the end, it’s a better solution than *either* AP1 *or* AP2 offers today.
And that’s my point. I and others have proven that AP2 isn’t at the level of AP1 over and over again (see above). But would you trust EITHER in the hands of a younger or older family member driver?
Bring on AP3, I say. And if that includes FSD, great. If not, just deliver the stop sign reaction first referenced at the AP1 event 10/2014 and we’ll call it even.
Until then, my next car will be an AP1 car because AP2 simply has not delivered.
Notwithstanding the above, let the record show that I am still rooting for csnow’s 3 bets (all for charity) concerning AP2 functionality. And remain (greatly) thankful that not a one of them involves a tattoo upon a certain glaringly bright (nether) region.
Edit: And for anyone who doesn’t think a sense of urgency is helpful (Tesla’s global infrastructure lead and unmatched scalability notwithstanding), have a look at this, and the number of cameras and sensors and such:
http://www.cnn.com/style/article/japan-nissan-driverless-car/
