Autopilot review Update


Of course, just after releasing my review of Tesla Autopilot they announced new pricing and features, with some explanation of what "full self drive" is.

For now, it turns out it's still driver assist, but on city streets. It's an interesting question if that's a good idea. I offer some additional analysis and updates.

Read my Update to Tesla Autopilot Review


Note that the "Full Self-Driving" package must be purchased in addition to Autopilot. So it's really $8,000 priced before delivery, not $5,000.

Some analysts are saying that there's little room for profit (if any) at $35,000, and therefore think that Tesla is making a mistake. I think it's a sign that Tesla trusts that their software add-on features are going to be very popular, both at initial purchase and in the future when they become more advanced. Give 'em the razor and sell 'em the blades -- sort of.

Any word on what buyers of the old packages are getting, and how much they have to pay to upgrade, if they don't get everything?

Thought I said that. Clarified it. I have asked about old buyers. One rumour suggests 2019 buyers get "free" full self driving, older ones must pay the $3K price difference. Older ones already have basic summon, nav on autopilot and auto park (sort of) so it's confusing to be sure.

As I wrote, $3K basic autopilot is a more reasonable price. I have not found much use for simple summon or nav on autopilot and auto-park has never worked for me, so it would be hard to pay $5K more for those. If summon really works I could see it being useful in some rare narrow parking locations. Or for the disabled.

Nav on autopilot has never done it for me. In tends to make me more nervous, not more relaxed. It's very little work to handle on and off ramps yourself.

I guess I never even considered buying Tesla with basic autopilot on its own. That might be why the pricing confused me but seems obvious to others.

What got me seriously considering buying one was "automatic driving from highway on-ramp to off-ramp including interchanges and overtaking slower cars" and autoparking and "your parked car will come find you anywhere in a parking lot" and "recognize and respond to traffic lights and stop signs" and "automatic driving on city streets," it sounds pretty awesome. Unfortunately, the reality seems to be much less exciting than those descriptions, and the fine print (*) might be enough to get them out of a lawsuit if they fail to deliver what they've promised.

Maybe not, though. I'm still tempted to buy one and sue for benefit of the bargain if they fail to deliver on their promises.

(*) "The currently enabled features require active driver supervision and do not make the vehicle autonomous. The activation and use of these features are dependent on achieving reliability far in excess of human drivers as demonstrated by billions of miles of experience, as well as regulatory approval, which may take longer in some jurisdictions. As these self-driving features evolve, your car will be continuously upgraded through over-the-air software updates."

I find Nav on Autopilot to be not that exciting. With both, you still hold the wheel at all times and watch the road at all times. With nav on autopilot, as you approach your exit, it gives you reminders to get in the right lane on the screen -- you must trigger lane changes manually, but it does them after you check your blind spots. And then it actually makes the exit into the off-ramp, and then nav on autopilot turns off.

Without nav on autopilot, the regular nav system tells you that the exit is coming up and you do pretty much the same thing to get to the right lane. Finally, as you approach the off-ramp, and it tells you to exit, you tweak the wheel to go into the off ramp and you are steering.

It's not a big difference. It's a slightly bigger difference on highway to highway interchanges, as it will steer them. But frankly, it's scary. I don't like how it does merges at high speed.

Yes, partially from your review update, and partially from my own research, I'm less excited now than I was when I first read the description. Still, I expect this to get better and better over time. Hopefully "hold the wheel at all times" will go away. I don't even always do that in my non-self-driving car. I thought you only had to torque the wheel once in a while. (A bit of googling suggests that this is correct, though the "nag time" may have been cut down as low as 15 seconds. Can you confirm this?)

In any case, navigate on autopilot is only one of the five features under FSD. The two features slated for "later this year" are also interesting. Maybe not at first, if the quality is poor, but some time within the lifetime of the car I expect them to be very nice. My commute to work doesn't use the highways at all, and most of the route (apart from a couple narrow streets in my development that are shared with law-breaking pedestrians/cyclists/parked-cars/animals and a pair of too-short left-hand turn lanes where oncoming traffic frequently crosses the double-yellow line) is probably something that Tesla's current early access software can handle (with me watching). And if I'm going to pay for Full Self Driving eventually, why not have it from the beginning?

I guess one reason is because they might cut the price later.

All this said, $43,000 ($35,000 base fee + $3,000 autopilot + $5,000 FSD) is a lot of money. Especially when I can probably get at least three more years out of my 40-50 mpg car without any significant repair costs. It'd definitely be a splurge on something unnecessary. But it'd be a lot of fun.

Yes, you can take your hands off for a while. However, Tesla recommends that you are supposed to keep them on, and so do I. It will make mistakes, and you want to be ready to fix them. Even when it's really good and you only need to fix the mistake once a month, you want to be quick at it, unless you want to crash once a month.

I am not sure autopilot on city streets will be that pleasant a product to use. But I have yet to try it. My own robocar safety driving was limited to highways, we could ask the more experienced safety drivers how they feel.

As for the cost -- don't forget the tax breaks and the gasoline savings when comparing to other cars.

Surely autopilot on city streets will be much more pleasant than safety driving on city streets. With autopilot, the absolute worst it can be is equal to regular driving (if it's any worse, I just don't use it at all). With safety driving, it's guaranteed to be more stressful than regular driving. The biggest stress that I would have in safety driving would be during times that I would want to take over, but am not supposed to because interventions are supposed to be avoided. With autopilot, I'll be happy to take over if I see something out of the ordinary, or something that the autopilot doesn't handle well.

That said, I'm not sure how often that will be. There's a school zone on my commute to work. Will Tesla properly handle that? Eventually, I'm sure they will, but maybe not initially. Will they handle the two turns (one right, one left) from a stop sign onto a 45 mph road, with limited visibility? That one's going to be tough. You have to pull out past the stop line, into the crosswalk, just to see if you can make it. But that will be the most rewarding when they do handle it, especially the left hand turn, because it requires checking traffic in both directions plus checking for pedestrians. Once they are able to handle that, it'll be much safer than driving it manually.

Most of the commute should be fine. It's mostly just stop and go traffic with some traffic lights. Easy stuff. Just stay in your lane and don't hit anything. Those few days when there's something weird going on I can just take over. Building a good level 2 (or even level 3) system is so much easier than building a robocar.

In terms of cost, yeah, there's the federal tax credit. I think that's it for me in terms of tax breaks. And in terms of gasoline savings, with my current car getting me 40-50 mpg, it won't be anywhere near the amount on Tesla's website. One unknown is how much more I'll be driving. Once "fall asleep in the car on the highway" becomes safe it'll be a lot more than I'm driving now, as I'll be much more willing to take weekend road trips that require me driving late at night (or getting a hotel room). But they're not there yet, and it's not clear when they will be there.

Are instructed to intervene if they have any doubt the vehicle will perform its actions safely. However, doing it all day, they get a better sense of that than the ordinary driver will.

Presumably you know from your odometer how much you drive currently. People tend to drive a little more in a new car. If you drive 10,000 miles/year at40mpg, you are spending $600 to $800 per year on it depending where you live. My electricity bill went down here in California after getting the electric car, but generally you might spend $300 on the electricity if you charge at home. For those who commute over a bridge in this town there is the extra saving of $3/workday for 4 years.

However, if this is too expensive, wait a bit and cheaper cars will be available.

I'm sure all safety drivers are officially told to intervene if they have any doubt the vehicle will perform its actions safely. Whether or not that's what they're really expected to do probably differs from company to company (and may very well have changed around March 18, 2018).

But that's neither here nor there. Any ADAS I use is going to be more pleasant than not using it. If it wasn't, I wouldn't use it.

Do I know whether or not I'd actually use any of the Tesla autopilot or FSD features? No, not really. I just saw a video this morning where the car would start a lane change and then abandon it halfway through it. It did this I think three times in a row. The speculation is that it's because it was driving in the direction of the sun. That's pretty awful. It's also something I expect will be fixed, in time, though I'm not sure exactly how. Good maps, deduction from other information, and a little bit of guessing? It's a place where LIDAR can definitely help, if not in detecting the lines themselves (can high res LIDAR see the paint bumps in the road?) then at least in detecting the exact positions of the other cars.

(Driving in the direction of the sun is actually one of those things I've wondered about during some of my regular driving. I sometimes just stick my hand up to cover the sun. It's also generally on roads I already know well. And there's probably a little bit of just getting lucky.)

This is another area where ADAS will be much safer than human driving, once it's perfected.

Add new comment