I guess he has to keep saying it's imminent, or investors will start treating all the FSD sales as liabilities.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Elon Musk explains the delay for Tesla’s autonomous coast-to-coast trip
- Thread starter conman
- Start date
Unfortunately, that is true. I know a surprisingly large amount of people who ask me for a ride in my "self-driving Tesla". I hate to correct them with "Well actually, it doesn't do that just yet, but some time within the next months it will". I get a lot of "oh really" and "wait so then what can it do". Many many people truly believe that all Tesla's are fully self-driving, which is quite frankly a little scary
stopcrazypp
Well-Known Member
Meh, I think it's akin to people complaining about the "autopilot" name, but that haven't caused any real issues (although it is used as scapegoat to try to escape blame for accidents). When they operate the vehicle or see it being operated they will immediately will see the limitations and that it's not self driving.
but that's not what he's talking about.
he's talking about how musk is strategically and knowingly hyping up non-existing products because he knows the mindshare he can garner from the unawares " non-Tesla-owner technology enthusiasts".
Journalism today is complete garbage, elon knows that and he knows what ever he says/ tweets, people will just run with it and print it without actually analyzing it. So he can say stuff like All teslas will be self driving level 5 coast to coast without a human driver in 2 years (back in 2015) without getting any push back from the press. Plus the amount of ridiculous statements he makes monthly about what a tesla can do.
Now if trump were to make a statement it would be analyzed and fact checked to death. All the experts from the woodwork would have an opinion on the article.
You would be amazed at how many people at my job (a tech company) actually believe Teslas are self driving car.
Each time i correct them, they are shocked. shocked!
I find it concerning that everyone who is working on self driving is using Lidar, except Elon things differently and admits in the end he could be wrong. Once they have developed a lidar module that doesn't need moving parts (spinning mirror system) I thing the winner is obvious. I know several companies are working on those.
stopcrazypp
Well-Known Member
I'm only commenting on the part about it being "scary" (which I presume he meant that it might be a safety issue, akin to people complaining how the name "autopilot" is actually dangerous).
As for media treatment, Elon Musk had done plenty of things deemed impossible and dismissed by others, so people give him trust proportionate to that. Yes, some in the tech media and industry do worship him as they did with Steve Jobs. For skepticism of his claims in the media, you can see them in mainly financial related media, where Tesla had been on the receiving end of skepticism pretty much ever since it started.
As for fact checking his claims, many of them are forward looking claims that can't really be fact checked at the moment they are posted because they haven't occurred yet. Trump is different in that he makes frequent claims that are just outright false and can be verified as false at the moment he said it. And after the thing happens (for example Tesla is late with a update or feature) the media does report on it (as they did for this delay).
stopcrazypp
Well-Known Member
Well not everyone, AutoX, AImotive, and Comma.ai are working on camera-only systems.
Camera only sounds like a challenge. But then, we humans rely mostly on our eyes when driving.
stopcrazypp
Well-Known Member
Well no one said it would be easy, but the idea behind all of the companies are that using the camera-only (or "vision-first") approach is that the hardware is already cheap and commoditized, which makes it more viable for mass adoption. I should note that some these companies are open to also using cheap commodity sensors like radar and ultrasonics too to supplement information, so it's not completely pure camera (although the cameras are the primary sensors).
It's kind of similar to Tesla's strategy with batteries when choosing the commoditized small cylindrical cells rather than waiting for high density prismatic cells like everyone else in the industry did. That pretty much gave Tesla a nice lead in developing long range EVs. Elon is likely betting the same here.
The other difference I should mention is that the big players like Waymo and Cruise are developing autonomous taxis rather than consumer owned vehicles, so the economics are different. They are betting car ownership will change drastically in the future.
Last edited:
TaoJones
Beyond Driven
That would be... let’s find a censor-friendly word here... disappointing. What’s the next revision - that off ramp to SC to on ramp won’t “count” either?
If that transcontinental trip doesn’t include stop sign and traffic signal reaction (not just recognition), then they just shouldn’t bother.
AP1 circa 2016 already did the bulk of the highway driving part after they worked out the initial bugs (e.g., with Autosteer abruptly exiting the interstate when you didn’t want to.) Many of us have already logged tens of thousands of miles of highway driving with AP1 and sadly with AP2. Nothing new in essence to see there.
And we’ve already seen the snake charger prototype. Wouldn’t exactly be a bank-breaker to deploy temporary units to the ~20 stops needed to support said journey. Minus the Barry White music to save the few bucks for licensing. Would actually be a cool beta test and it would add legitimacy to the endeavor. Supposedly there are all kinds of additional mobile rangers and service techs added across the country, and scheduling pretty much down to the hour wouldn’t exactly be a problem if they wanted to babysit the snake chargers or disassemble them after the test car passes through.
Anyway, as disappointing as this has been, there’s a certain wisdom in waiting for this transcontinental test and you can’t blame Elon for waiting. We’ve already seen one video (12/2016) that apparently was staged to, in part, demonstrate stop sign reaction. I don’t need to see another fake.
Besides - a lot of people placed orders based upon that video and the clear promises made at that time. I’m surprised there haven’t been more class actions as a result. The moment a successful and hopefully non-bogus transcontinental trip happens, people are going to line up around the world to order. So if that drive doesn’t happen until the Tesla-designed SoC/board is ready for those of us who paid for full FSD, that’s okay. I realize that’s a couple of years away, but hey, the date’s already been missed.
Frankly at this point, if Tesla delivers just the stop sign and traffic light reaction feature, that would be a win. All this other marketing hokum that’s on the website about auto-lane changes and such is just noise by comparison.
When I recently looked at job openings at Tesla, they were looking for someone helping the AP team. The requirement was being familiar with the Unreal engine. It's a game development tool that is known to render photorealistic environments in realtime. It is basically a simulation and visualization tool. Not at all something to would take anything from the real world and interprets it. Made me wonder if they are in need to create more fake videos. LOL
whitex
Well-Known Member
I think focusing on something so minor as charging, is like worrying about what SPF sunscreen you should be wearing on Mars before you figured out how to get there in the first place. If an AP2.5 HW (no extra sensors) Tesla manages to make it from a busy city center to busy city center (say LA to NY) with the only human intervention being someone plugging it in and unplugging it at the SC, then they'd have a successful demo. If the car needs help off the highway and/or through construction zones, then it's a fail - we already know AP1 and AP2 can handle just highway driving.
This kind of reminded me of a joke I heard when I was a kid. A spy takes a mission to parachute behind the enemy lines where a bicycle is left hidden for him. He is to ride the bicycle 60 miles to take a secret location and take surveillance pictures. As he jumps out of the plane, the main parachute fails. He cuts it lose and realizes they didn't pack him a secondary one. The next thing that pops into the guys head is "I sure hope they didn't forget to pump the bicycle tires".
Here is hoping Tesla actually does the ride by mid this year (Elon's latest deadline) and the manual SC charging is the only thing that people can pick on.
whitex
Well-Known Member
It's a common training technique for vision systems (I remember reading somewhere that Google even applied for an exemption to the NHTSA to allow the simulated miles to count towards their driven miles). To train a neural network based vision systems you need labeled videos (videos where someone labeled where the cars are). Taking videos and paying people to draw rectangles around cars, signs, pedestrians and other objects is expensive, time consuming, and error prone. If you have a way to generate those by a computer (like a computer driving game, the computer knows where all the object are since it had to draw it in the first place, now take that video, combine with locations of all objects, and you have a great source of training data). Now the only thing limiting the number of training videos is how much computing power you can rent in the cloud. Of course, the more realistic the game engine, the better the training results (so for example the neural network doesn't learn to estimate the speed of the car based on MPEG compression effects).
I think focusing on something so minor as charging, is like worrying about what SPF sunscreen you should be wearing on Mars before you figured out how to get there in the first place. If an AP2.5 HW (no extra sensors) Tesla manages to make it from a busy city center to busy city center (say LA to NY) with the only human intervention being someone plugging it in and unplugging it at the SC, then they'd have a successful demo. If the car needs help off the highway and/or through construction zones, then it's a fail - we already know AP1 and AP2 can handle just highway driving.
This kind of reminded me of a joke I heard when I was a kid. A spy takes a mission to parachute behind the enemy lines where a bicycle is left hidden for him. He is to ride the bicycle 60 miles to take a secret location and take surveillance pictures. As he jumps out of the plane, the main parachute fails. He cuts it lose and realizes they didn't pack him a secondary one. The next thing that pops into the guys head is "I sure hope they didn't forget to pump the bicycle tires".
Here is hoping Tesla actually does the ride by mid this year (Elon's latest deadline) and the manual SC charging is the only thing that people can pick on.
Yea, charging isn't an issue at all. I guess it would be enough to show the automated charger again and tell people that's what they are going to install at each SC.
On the other hand I don't think it will be from one city center to another, but rather just highway driving, with some charging stops on the way. That would be way easier to do and it would show the next step for AP, having a system that can get you from on rap to off ramp w/o any human intervention. Cities will be way harder to do, since you have pedestrians crossing the streets anywhere and intersections with traffic from all sides.
I just hope this time they make it more public, with a lifestream, or maybe even some journalists to observe it. I would definitely be disappointed with just another video. Fool me once...
Yes, listen to this interview with Hotz CEO of comma.ai where he talks about self driving being much more a software and not hardware challenge among other things:
Episode 6: How George Hotz plans to win the self-driving race (go to about the 3 minute mark if you are in a rush to get part Hotz starts talking about why lidar is not the answer)
Last edited:
stopcrazypp
Well-Known Member
I never expected Tesla to have snake chargers at every stop. Rather I expected that they might have attendants along the way (as they already did in some stations). I don't think the charging is a big deal for the demo.
I think you are stretching a bit to extend that to off ramp to SC not counting.
stopcrazypp
Well-Known Member
I do recall seeing a post somewhere that said that charging wasn't going to count and that it would still be done manually. That said, the post was from a random person online who remained anonymous claiming to have inside information so I'm not sure how much weight should be given to it.
TaoJones
Beyond Driven
Think whatever you want. The two cases also exist independently.
The snake charger was publicized well after the 10/2014 event. Now, while possible that snake charging will go the way of the I-5 battery swap station prototype, I'd suggest that there's more utility for the former these days. See Arizona's recent position with regard to cars without drivers. Never mind that they're lurking in the back seat at the moment.
It's a stone cold fact that unless and until there's the stop sign (and by extension, traffic signal) reaction mentioned publicly at the 10/2014 event, ain't gonna be no transcontinental trip even if cheating with the SC charging occurs.
Is a transcontinental trip a win without snake charging now? Sure - it'll pass the popular press test at least.
Is it a win without off ramp to on ramp? Absolutely not. We could do that *in essence* 2+ years ago with AP1. The regression with AP2 notwithstanding.
Neither Tesla nor Elon get grief from me for waiting *for that trip*, especially given the disgrace from the 12/2016 video debacle.
I'm actually tickled to death that we'll have the hardware to support that trip, even though SvC interaction will evidently be required. Now do I believe that those of us who bought, say, an early 2017 AP2 car with prepaid EAP and FSD will be made whole? Doubtful. Which is why I'd be happy with stop sign reaction. I have no idea what the class action would settle for, however. I haven't even gotten a software update in going on 3 months since *.50.3. Yes, I've read the excuses, recall JonMc's forecast from the Amsterdam talk (winter 2017) and have seen the latest Elektrek happytalk.
Any day now. I'm not even annoyed that that's taking awhile - what will be annoying is that I probably will still have loading errors for audio and terrible, and I mean terrible touchscreen pane management. 6 taps when 2 or 3 used to suffice? C'mon now.
Similar threads
