Tesla "Full Self-Driving": Good, Bad, and Scary
I Just Pressed the CANCEL Button On My FSD Subscription
Q: Richard, how long have you been driving?
A: Almost 60 Years!
Q:You must be very tired.
Another old and stupid joke, but to answer the implied question, I'm not tired of driving, which is fundamentally relaxing, I'm just tired of paying attention.
How I Happened To Try FSD In The First Place
In my First Impressions of the Tesla S, I noted that I eschewed purchasing the $7000 self-driving option because I knew at the time that it didn't make the car truly autonomous. Now that it's a $10,000 option, I didn't suddenly change my mind. Rather, Tesla offered the opportunity of renting the software. One could buy a "subscription" for $199 per month. I admit it: I was $199 curious, if not $10k curious. Because of the pandemic, I had been doing very little driving in the two years since I got the car since there were no in-person trade shows to attend. When CES de-virtualized itself, a trip to Las Vegas, a few hundred miles from Sedona, was in prospect. I "subscribed" to the software to see how it improved the driving experience over my 2020 trip using the "Autopilot" software that comes with the car at no additional cost.
I give credit to Tesla for trying the increasingly pervasive "subscription" model for software. Whether it was a whim of Elon Musk or implemented after months of debate within the Tesla management structure, assuming they have one, I have no idea. "Income streams" are now a thing, and, as with streaming entertainment, you can end up paying forever for something you used once or only sporadically. But $199 per month gets your attention more than does the monthly Netflix bill. My finger didn't linger; when the month was over, so was my subscription.
What is Full Self-Driving?
Who do you want to believe? Tesla or my own observations? FSD Features currently include:
- Auto Lane Change
- Smart Summon
- Traffic Light and Stop Sign Control (Beta)
In my First Impressions blog, I covered the "Autopilot" features, which I found valuable but to some extent problematic. Let me extend my comments to cover these additional features, which are potentially valuable but also to some extent problematic. First let me dispose of the parking and "summoning" features, which some may covet and some may feel, as do I, that they serve little purpose beyond proving they exist. Even with a big Tesla S in an unexceptional and less than capacious garage, I have no trouble parking or extricating the vehicle. I suppose one could consider these features as permitting a garage to hold even more junk since the door would never need to be opened while the car is inside the garage.*
Which brings me to the two driving-related features. The surprisingly wonderful Auto Lane Change and the surprisingly irritating and dangerous Traffic Light and Stop Sign Control.
Auto Lane Change (Good!)
My recent trip to Las Vegas involved two lengthy segments, one on I-40 from Flagstaff to Kingman in Arizona and Rt, 93 from Kingman to Hoover Dam, at which it starts to gentrify. I-40 in particular involves a lot of road-sharing with an unending procession of 18-wheelers which are always going just a little slower than you want to drive and exhibiting very truck-like behavior. Ideally, at least for me, would be setting the Tesla cruise control slightly above the speed limit and taking a nap after leaving Flagstaff. This doesn't work for obvious reasons; there are any number of occasions on which lanes must be changed. FSD software makes it almost fun:
- Tap on the turn signal indicator for a fraction of a second. Your blinker turns on.
- The Tesla watches for conflicting traffic and changes to the appropriate lane when safe.
- The blinker turns off and the car continues on its way in the new lane while increasing to the speed commanded by cruise control.
No craning of the neck, no guessing whether it's a good time, no need to pay attention, or at least there wasn't during my trial of the feature. Neat-O!
Traffic Light and Stop Sign Control (Bad and Scary)
Sedona actually has more than one traffic light! Big City, eh? And the Tesla under FSD tried to slow down and stop for all of them, even if they're green! I sort of understand why the software is programmed this way; it's not smart enough to know when they might become red**. But any human driver develops an intuition as to whether and when it's necessary to slow down, and, unlike the Tesla, doesn't s-l-o-w d-o-w-n for every light, to the discomfiture of driver and passenger alike. To prevent this behavior, the driver must gently tap the accelerator to reassure the car that it can continue. And, if the light is genuinely red, the car will stop, properly, but then won't start again when it turns green without driver encouragement. Compare this with behavior of the cruise control without the engagement of FSD. The car follows the car in front and behaves exactly as it should! (Of course, if there is no car in front, it won't stop for a red light at all. There is that.)
This feature performs exactly as Tesla claims. But, at least in my opinion, it's more effort to use it than simply driving down the boulevard as always. If it were smarter and didn't require so much driver interaction it would be far more valuable. Far worse, it's scary in the same manner if not to the same degree as the cruise control issue I discussed in my First Impressions. Because the car will slow down for what a human would perceive as "no reason," there is danger of being rear-ended. I've driven the Tesla enough to know when to disengage the cruise control when I see a situation developing that will generate a spurious braking event.
So Why Press The CANCEL Button?
I love the lane change. I find the summon/parking feature worthless. I believe the traffic signal feature is worse than useless; it requires more driver effort and attention than if it didn't exist. It would be nice if the lane change feature were incorporated in cruise control as a feature of the standard version of the car, but it isn't, and I'm not willing to pay $199 per month (or $10,000 one time) so I don't have to think too hard about passing trucks. Or, in marketing talk, there's no "value proposition" for me.
And, as much as I love the car, admire Tesla the company, and enjoy Elon Musk the progenitor, I do believe that the fanciful cost of what is essentially a single byte switch in software that is already present in the vehicle will plummet. The billions of dollars other manufacturers are spending on autonomous vehicle research (and inter-vehicle communications) will create enough competition to assure that.
Using the FSD has, sadly, discouraged me from my previously optimistic take on autonomous driving. Although I trust the Tesla enough that I can use its features on an interstate with no cross-traffic and no immediate interchanges, I have to obey both Tesla's and the safety crowd's injunction to pay full attention under almost all other circumstances. As I stated at the very beginning, it's the paying attention that vitiates the benefits of autonomy.
Hey! Good Idea!
This isn't quite up there with saving the world, but I had a notion how to solve the problem with electrc cars overloading the electric grid. If you have a "time of use" electric plan, of course you will set your Tesla or other EV to charge when the rates go down. Let's say that's 7PM. Presumably your neighbors with the same power company will do the same, resulting in a big power demand at that time. But, if you're a commuter you've only used a percentage of your battery range during the day. While you can fully charge a depleted battery "overnight" in most cases, you can get enough energy into that battery from a day's commutation in only an hour or three. Assuming you work normal hours, e.g., 9ish to 5ish, that allows you to charge for those hours any time between 7PM, when the rate goes down, to 8AM, when you want to drive. Instead of programming the car to start charging at 7PM, program it to start at a random time that will satisfy your need for energy before 8AM, but NOT necessarily at 7PM. That will spread out the power drain on the grid by perhaps a factor of four or more.
Of course you're not going to do this yourself, but the vehicle manufacturer can easily set the start-charging timer randomly with a few extra bytes of software. If the battery is truly shagged out after a long squawk, the charging algorithm will know it and start as early as necessary.
Two Good Ideas!
Although you won't like this second one: A way to enforce this for the greedy utilities will be to allow only so many kWh at the low price every day so that the owner is encouraged to spread his charging activities throughout the week. Nothing will prevent him from fully charging on any given night but it will be more expensive if he doesn't get with the program. (Don't tell anyone about this part.)
* Quantifying this, let's assume you can get an extra 60 square feet of junk storage by squeezing the car among the bags of pet food, old magazines, gardening implements, and that ancient copying machine that you may someday use. That comes out to an annual cost of $40/sqft. Perhaps a bargain for a Manhattan apartment, assuming you can park in one. A suburban garage? Not so much!
** My second blog ever, about using your GPS to save gas, might be a valid solution to this problem in some cases! I was very proud of that one, which subsequently has been universally ignored.