It looks like the market has rendered a verdict on Elon Musk's big cybertaxi announcement last night:
Musk told the audience that anyone who didn't believe in Tesla's driverless future shouldn't be an investor in the company. It looks like they took him at his word. It's just the usual Musk vaporware, same as always.
As usual, if you really want to judge where autonomous car technology is, look to Waymo. They've (finally!) started broad testing on freeways, and you can legit set a Waymo car on its way and then take a nap while it drives you around. I don't think they'll quite make my forecast of a real driverless car on sale to the public by 2025, but it will be close.
It is one thing to say "Our company's goal is to sell a cheap software product which will entirely eliminate human labor costs in a huge market, and will transform that market's cost structure, in our favor",
and an entirely different thing to say "If you think we are lying to you again, then you shouldn't be an investor in our company."
Elon seems to be entirely tone-deaf when it comes to talking to...well, anyone.
In his favor, he brought electric cars to the market when the existing automakers were blowing smoke about their intentions to do anything real about that.
To his detriment, Elon is an idiot.
Putting aside Musk, and you have to do so, its not "vaporware" as I have had it since 2019. The latest version, running on a Model 3 with HW3 is almost perfect driving around the 90042, and the 90042 is full of badly marked streets, blind intersections, almost one way roads in the hills, etc.
What it does not do yet is pull into parking lots or driveways and find spaces by itself, although this last version (since last week) does find a spot to what I would call "pull over" in front of the address you have selected.
The difference is geofencing vs. use anywhere. The other difference is no one can buy a Waymo and anyone can own FSD. The final difference is Tesla FSD runs on basic hardware and is essentially a software program.
Its an open question of how it will be monetized. I am not a fan of taxi companies, but that's me.
I grew up in 90042. FHS, Burbank, and Garvanza. My mom still lives there.
That pesky last twenty percent taking eighty percent of the time…
Maybe this has been discussed somewhere and I've missed it, but how do you insure a driverless car? If something goes wrong while the vehicle's in driverless mode and you're in an accident, who's at fault? The owner, or the company's software? The liability issues seem really intractable .
Liability issues are not intractable at all. They don't have answers under the law as it stands, but legislatures can, and I certainly hope will, pass laws answering those questions.
I believe you will find that state legislatures generally, and especially the California legislature, are well aware that a statute dealing with autonomous vehicles is going to be needed.
No legal issues needed. Who's the driver? They're liable.
In level 4 and 5 autonomy, the driver is the company. In level 3, it swaps between the driver in the car (who is responsible most of the time, and responsible for making sure the vehicle is in autonomous mode safely) and the company (who is responsible only while in the mode).
And level 2, where the driver is responsible always, no matter the assistance features.
It's not that simple. If you own the car, you may or may not own the software. If you own the software and it glitches, you are responsible at first instance. You may be able to bring the car company in as a third party; if your car company bought the software, they may be able to bring in their supplier as a fourth party. Product liability cases can get very complicated that way.
A legislated answer is by far the best way to deal with such things.
This doesn't seem like a very difficult or unusual problem. Our civilization is full of machinery that runs without human operators always present, and which has some ability to hurt/kill people if it malfunctions, is used in ways its engineers didn't anticipate, or if it meets conditions for which it's poorly suited. Just off the top of my head, there's escalators, elevators, and airport people movers. I expect that for all of these, there's a well-settled body of law about insurance, inspections, and product liability.
It's not obvious that autonomous cars present any new intractable problems in this area.
Since elevators, escalators, people movers, carousels, roller coasters, etc all run on fixed and delineated and restricted rights of way, maybe aircraft autopilot would be a closer analogy. Flying is dimensionally more complex but I think probably a lot less subject to urgent situations involving unexpected behavior by living beings with legal rights and unanticipated appearances of objects whose owners have legal rights.
I don't know what legal framework aircraft autopilots function within. But self-driving now is rare enough that it's still experimental and I think dealt with case-by-case even in malfunctions. I don't see that continuing if it's going to be generally marketed to the public, and I do think settling liability/responsibility questions isn't going to be so easy. "When used as intended" is a pretty slippery starting point.
Autonomous vehicles seem to be very, very different than elevators and escalators.
The risk/potential damage is significantly higher for AVs.
The frequency of use and potential different types of damage is significantly higher for AVs.
The fact that AVs will still likely involve some type of user input, even if very minor, is significantly different.
All of these represent major problems for this comparison.
Assigning well-settled procedures for one set of circumstances to a different set of circumstances is not so simple.
Only if you don't understand who the driver is.
If it's autonomous, the company providing the software that is driving is responsible.
My brother works for Google, so he is of course well aware of Waymo, but he also bought a 2021 Model Y because as noted, Waymo's are not a vehicle you can pick up at your local car dealership.
Granted, you have to keep your eyes on the road/camera that is looking at your eyes, but the self-driving feature is (a bit to my surprise) legit. I was stunned that it was easy for him to use self driving for over 60% of the trip we took from the Bay Area to Los Angeles to visit my mom (with a stop over in Bakersfield, so lots of driving that day).
No, you cannot sleep or just goof around on your phone/tablet, or watch a film, but being able to not constantly hit the brake/gas (so to speak) in stop/go traffic made the trip much less exhausting for him. We took a day trip with my Mom to the Chumash Casino, and again, surprised at how easy it was for my brother to let the car do its self-driving thing. Even windy roads were a breeze for the car to self drive.
I really wish Musk did not align himself with Trump who is uncomfortably close to being installed as President for life (like Putin), as Tesla did light a fire under car manufacturers rears to get on the electric bandwagon.
That being said, it sounds like the robo taxi event was a bit of a fail that even Musk fanboys would be willing to acknowledge. Will it hurt Musk in the long term, nope, unfortunately, he will still be able to say a lot of inaccurate stuff and end up remaining beyond filthy rich.
For many of us, at least it kept him off Twitter for 2 hours. Which is a total win any way you look at it.
No kidding. Guy needs more things to destress him, he sucks at social media.
I am in my second month of using FSD, and all I can say is: If you want to criticize Tesla’s self driving tech before actually using it, DON’T. I just drove from NJ to southwest VA and back, and FSD drove the car 99% of the time. I never had to intervene for safety reasons. The only times I took control were when the lane it chose was not the lane I would have chosen if I were driving - which I wasn’t! I have used it in all kinds of situations (except snow), and it is amazing. No, it isn’t so perfect that it is ready for totally autonomous driving, but it is very close. I also used the new smart summon feature, which is a hoot. I don’t even have the very latest version of FSD yet, so the few quirks I noticed (going too fast or slow on off ramps when no speed limit is posted) will probably get better soon. I don’t care how long Musk has been promising this. And I don’t care that he has become a Trumpist asshole. All I know is that what I have experienced for the past 2 months has been mind boggling.
A couple years ago I went down to Los Angeles to pick up some stuff from various points in the area to move to my new home in the Pacific Northwest. Having driven the L.A. freeways in the past -- I spent several years commuting there in the 1990's, first on the Santa Monica freeway and later up and down the Sepulveda Pass -- I was dreading all the stop-and-go traffic. After my encounter with bad traffic that exceeded even my worst memories, I decided to see how the Adaptive Cruise Control in my Toyota Sienna would respond. It was fantastic! It kept a safe distance between me and the car in front of me literally for hours (while I warily kept my foot ready to stomp on the brake petal, of course). Not only did it relieve me of the constant shift from throttle to brake, I wound up with far better gas mileage than I'd achieved on my own. (And, of course, no leg cramps, a boon it had already provided on the long trek down I-5.)
So one doesn't need a Tesla to enjoy some of the benefits of driving automation. Most other car manufacturers provide Adaptive Cruise Control (Level 1 autonomous operation as defined by SAE) and Lane-Keeping (Level 2). These alone make driving much more safe and comfortable.
Tesla's claims of Level 5 autonomous operation (informally, "steering wheel optional") still need to be verified. They've definitely taken things quite far in that direction, however.
Tesla has never claimed their current cars are doing or have level 5 enabled,
Waymos are driving around my neighborhood all the time (vehicle is a Jaguar). I've watched as they make left turns with traffic approaching and pedestrians out on the sidewalks. They are impressive. Only minor glitch was when I was driving behind one on a street that was narrowed due to construction, so there was less "lane-sense" at the time. Waymo didn't move as much to the right as it could, preventing the car in the other direction from slipping through. Couldn't (didn't) back up because I was right behind it*. So the other driver had to back up to allow the Waymo through. Waymos do well making right turns onto busy streets (especially when that traffic slows, leaving some still in the intersection). Overall vibe: careful, smooth, driven-by-a-chauffeur.
* also, no driver-to-driver communication that we occasionally use (e.g wave through, or the reverse). w/o driver (Waymo) that can't be done. At least untl Wayos have mannequin-like androids to move their arms and hands and faces with expressions. Which will probably never happen.
Waymo is testing all on their own liability, while Tesla isn't. This allows Tesla to collect more data to test virtually but also release flawed versions while Waymo racks up more confidence miles.
That's really all the difference. Doesn't make either of them vaporware.
That's got to be a disappointment for Tesla. I figured the reason why they kept the October presentation date despite not having a ready model for Cyber-Taxi is because they were really hoping it would provide a share price increase before Musk has to talk about the 3rd Quarter earnings again (which will probably disappoint, although Cyber-Truck sales are up).
At least on paper, Tesla should have an advantage here over Waymo. They don't have to build a ton of specialized hardware at a competitive price because they don't use LIDAR, and they have tons of data from Full Self-Driving folks for machine learning. But in practice, it seems like they've struggled to pull it off.
I also think the idea of "your car can make you money during the day while you work!" is not viable. Back when the "Sharing Economy" startups were a lot more common, there were attempts at "You can rent your [insert equipment] out for money!", and they all foundered except for the well-established ride-share and housing rental companies because most folks don't want to lend out their stuff to strangers - especially something as valuable as a car, to say nothing of folks leaving their personal stuff in the car all the time.
What they need is a five-person, four-door vehicle that can be sold in bulk to taxi services, the Ride-Share giants, and car rental companies in general.
The van is cool, though. And a self-driving EV bus would be tremendously useful - labor costs are the biggest expense of operating bus lines, so it would let you run far more bus routes far more frequently.
Waymo can go 10,000 miles between "interventions"
Tesla - not quite as far
But "interventions" for Waymo involve somebody driving out to where the car is stuck and manually driving it
"Interventions" with Tesla involve somebody tapping the brakes and then the accelerator
Waymo’s are in contact with human backups who, although they can’t take over the steering, help with identifying objects and making decisions when the car gets confused.
There’s that last 20% again
There's a bit of irony involved with driverless Waymo, though. Each Waymo actually has someone monitoring it and activities inside of the vehicle at all times. Until Waymo's AI model is worked out, this is what it is.
Took a Waymo in Phoenix. It let my wife and I out several blocks from our resort with no option but to hoof it. Don't know why.
We could have gotten lost, could have had to drag luggage, my wife had a broken foot at the time. It may have been a bad neighborhood...
Waymo offered us $10 off our next ride. Thanks for nothin.