Ulmo
Active Member
I remain optimistic too. I just want to be realistic at the same time. I generally agree with your 5 points, if not exactly.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Should FSDC (Full Self-Driving Capabilities) be an option right now?
- Thread starter CTgigawatt
- Start date
-
- Tags
- auto pilot 2.0 fsdc Model S
Lastly folks - has it ever occurred to any of you that the cries of "That tech won't work" often come from auto companies which do not have anything for sale remotely capable of competing with Tesla? It's really easy to keep saying "Lidar is a must" when you have no product to sell. In fact it's almost a nice excuse to keep you from having to actually put up and shut up in the marketplace, isn't it? Just keep saying "Cameras can't do it, nope, not safe. Pay no attention to the 40% accident reduction data over there. That's meaningless. Listen to me - need LIDAR or you will die." "What?" "Can you buy it now? Um, no - get back to me in a couple years."
Why is giving Elon the benefit of the doubt even part of this discussion? Hell why is Elon himself even part of it?
If we're both sitting down at Tesla months ago arguing about this neither one of us is going to say "Well Elon wants this so I think we should do it based on his track record". His track record might be important to consumers like us, but I specifically said to put ourselves in their shoes. To be the engineer or the marketing person. To try to argue from their perspective.
That's an important part because it removes the tendency for a consumer to get emotional over a product or idea. We also know as time goes on the consumers that buy a Tesla are going to be less willing to overlook problems/defects/short-comings. We can already see this from new HW2 owners that are way less willing to overlook problems than we were.
As to your points -
Sunny days? Sorry I live in Seattle, and I don't know what those are. Rumor has it we get those occasionally, but you gotta be quick to catch one.
Upgrading the Vision Computer isn't going to do you any good if the Camera sensors get dirty really fast. The reason the computer became part of the discussion is there is a disconnect between NVidia and Tesla as to what's required. Tesla addressed this by saying that they can always upgrade the computer. This is a good point, but I wasn't worried about the computer.
They won the market for the Model S/X. So why was FSD needed at the time? There wasn't anyone even close to releasing something EAP can do for a long time.
At the time of the FSD option release didn't Tesla already have hundreds of thousands of Model 3 reservations? Where they made the reservation with no expectation of full self driving?
You keep using the words modestly, but that isn't what FSD says to people. It especially doesn't say it when Tesla describes a scenario where the car drives without you in it. Where you can earn money with your car.
There is a big difference between what NVidia claims versus what Tesla claims. NVidia's claim only has to do with the vision computer hardware, and not the entire sensor suite.
Where did I put my money? On NVidia last early last year.
Still don't feel bad about that bet. As to HW2 I'm going to wait until Tesla actually starts to deliver on what they've promised before I part with my hard earned NVidia money. It was a lot of work sitting on it.
Okay I'll unpack it further. Giving someone the benefit of the doubt if they have a long track record of succeeding when others could not see the upside of their actions is rational. It becomes reasonable to assume, until proven otherwise, that the person with the track record has a history of thinking things through - and therefore this time has also thought things through - even if ye little outside hobbyist with no inside knowledge can't see it.
Elon runs the show. That's why he's part of it. He has the vision - the engineers and marketeers execute his vision as best they can. That's their job. It is not the engineer's job to decide what the market wants or the overall direction in which the company should head. The CEO may run up against the laws of what is possible - the engineer may find those limits for the CEO. But it's way above the engineer's pay grade to have large scale vision. And without a visionary CEO pushing the limits the engineer will never find those limits.
I don't understand your point, to be honest. I did argue it from a marketing perspective already - FSD is sexy as hell and Tesla is sexy as hell (sexy + green is Tesla's brand) - and sexy-as-hell sells cars. From an engineering perspective? What about the engineering perspective do you specifically want to discuss? That's a different discussion.
Actually we don't know that. That is another data-free piece of conventional wisdom - nothing about Tesla's success has met with conventional wisdom thus far. We will see what the consumers think. What I know, my little bit of data - is that I have 7 friends with Model 3 reservations. All will convert. Tesla has tremendous sex appeal with them, and I imagine these folks will forgive its shortcomings.
Again, I don't know if that is true. Where is the data which says a larger portion of new Tesla buyers are not satisfied with their purchases vs 1 year ago, or 2?
Sorry your rain is not relevant to my life experience in So Cal nor that of millions of other consumers in the sun belt. Sorry you may have to wait longer than the rest of us for self driving to become a reality in your climate. But I'm quite glad Elon is not holding back on self driving until he has a system capable of operating in all conditions. Very glad he doesn't let the perfect get in the way of the good.
Soooo - so what? Tesla should have waited to release a system until they had little windwshield squeegees swinging back and forth? Well crum - maybe I'll have to take 30 seconds routinely as a walk-around to wipe my lenses with a microfiber before setting off on a multi-hour journey. Maybe if I encounter a bug or dust storm I'll have to pull off the road to clean them or steer myself. Oh heavens. The horror.
Again, I'm very glad a reactive guy like you is not the guy in charge of Tesla's road map - or Tesla would be just another player in the race and we wouldn't have gotten where we are now. You'd be the guy who would've killed Autopilot 1.0 before it got out the door. "Hey guys, um, why are we doing this? Where's the upside? Nobody else has anything like this? Guys? Guys where are you going! You're fired!"
I AM very glad proactive Musk is running the show - the guy who wants to drive forward as fast as possible as hard as possible until his mind and body simply give out.
You don't know that nobody is on EAP's heels, actually. What we DO know is MBZ is working furiously to catch up and this fall in the 2018 S Class will release the third Drive Pilot system in four years - only 12 months after the prior generation in the 2017 E class. Also you are implicitly discounting the R&D value of fleet learning. Maybe FL is a lot of hype - maybe not. But if it is NOT hype then Tesla is wise to be doing it now.
Last edited:
Even before I bought my car I knew that blind/side spot monitoring, summons, and autopark would never work reliably on it.
They wouldn't work reliably because of my experience with Ultrasonic sensors.
As expected these "features" don't work, and interesting enough my main excitement in HW2 is the potential it has in fixing those things.
So I don't think you can quickly dismiss my fears of the Camera sensors not working well in the rain. If the FSD driving option doesn't work in the rain people are going to be livid.
I don't dismiss your fears about the rain - you may be right about that. I do disagree that enough people will be livid that Tesla would have been better off not releasing an FSD system until they had one which will work well in the rain. Of course there are also varying degrees of rain - perhaps it will only work well in light rain during the day, etc etc.
Also - even if it doesn't work well in the rain - what else will they buy that works better in the rain? A Bolt? A 3 series BMW?
BTW @S4WRXTTCS - I have no idea if - assuming this system will not work well in rain - what WILL work well in the rain. Nevertheless I'm glad Tesla is not letting that hold back progress.
Just saw this reply you made - didn't even notice it before in the furious back-and-forth.
Yes - fair point. Well - except my fellow nerd friends. I think the friends in line with me in Pasadena that morning expected it would be FSD based on the time lag, but as for everyone else? Fine. [/QUOTE]
Fair point again. The Tesla Network stuff does seem far fetched to me - but again, I am not a computer engineer. Maybe it's possible - I really don't know. My only direct experience with a person working daily with AI is a friend inside the Amazon Echo team - a phd computational physicist. He is a strong believer in the power of neural networks to iron out corner cases over time - and that something like sunny day self driving with cameras is simply a matter of network training and learning corner cases. I personally do not have the expertise to make that call.
Personally I think the tech is going to wow the consumer so much that all will be forgiven in terms of overall customer satisfaction rates - as long as it does do enough magic that other cars can't do.
Again, fair enough. But also remember NVIDIA is in the business of selling computers to automakers - many of whom want to use more sensors. It's NVIDIA's best move to remain sensor agnostic publicly. It's also the safe move politically to say "I dunno if what Tesla is doing will work."
Congrats - that's quite a return and a lot of work to sit on to hit 5X I can imagine. If i didn't need a second Tesla right now I probably would be waiting myself and be perfectly happy with AP 1.
Yes - fair point. Well - except my fellow nerd friends. I think the friends in line with me in Pasadena that morning expected it would be FSD based on the time lag, but as for everyone else? Fine. [/QUOTE]
Fair point again. The Tesla Network stuff does seem far fetched to me - but again, I am not a computer engineer. Maybe it's possible - I really don't know. My only direct experience with a person working daily with AI is a friend inside the Amazon Echo team - a phd computational physicist. He is a strong believer in the power of neural networks to iron out corner cases over time - and that something like sunny day self driving with cameras is simply a matter of network training and learning corner cases. I personally do not have the expertise to make that call.
Personally I think the tech is going to wow the consumer so much that all will be forgiven in terms of overall customer satisfaction rates - as long as it does do enough magic that other cars can't do.
Again, fair enough. But also remember NVIDIA is in the business of selling computers to automakers - many of whom want to use more sensors. It's NVIDIA's best move to remain sensor agnostic publicly. It's also the safe move politically to say "I dunno if what Tesla is doing will work."
Congrats - that's quite a return and a lot of work to sit on to hit 5X I can imagine. If i didn't need a second Tesla right now I probably would be waiting myself and be perfectly happy with AP 1.
It's a little hard to have a reasonable discussion with you because you keep exaggerating where I'm coming from.
I wouldn't have killed every aspect of HW1, but I definitely wouldn't have signed off on things that I felt wouldn't work correctly. I would have pushed for side radar to improve blind spot monitoring as an example. That's one thing we know that doesn't work right. As to AP itself I'm not sure what I would have said. I'm a strong proponent of the driver has to take responsibility, but I feel like it was a weird ask of a driver to remain engaged while taking away the primary (steering) component of driving. So I'm not sure what I would have argued for. After using AP1 for awhile I came to the conclusion that it was unacceptable to have a Level 2 car that couldn't see stopped cars or large stopped objects. But, I didn't know that when AP was first released.
As to fleet learning they already have that with the hardware that is there so I'm not sure what you're getting at with that one. The customer doesn't even need to have EAP active for that to take place.
I also fail to see how this has anything to do with progress. Like I stated before they can't use this money. So it's not even like a kickstarter where they could use the money for development.
What purpose does it actually serve beyond the temporary marketing sexiness? At some point the claim has to be backed up.
It can lock a person into the car or into a buying decision earlier, but like I previously said there wasn't anyone in the vicinity. The approach MB is taking is a lot more of German like approach.
Sure other manufactures are closing in, and within 2 years or so we should see competitors that are better aligned to compete with Tesla.
But, that's a long time in development time.
Elon has a tendency to stir the pot every few months with something new or exciting. But, with FSD he announced it along with EAP thereby robbing the chance of announcing it 6 months later to stir the pot again. Instead they just added a lot of things to their plate with no indication as to when they'll be done. There is no timeline given to the customer for FSD at all. There is no attempt to keep the customers up to date with its development. There is no real DMV record for how well it's doing like we can see for Uber, or for Googles
I honestly haven't seen that many people buying off on this whole FSD option hook line and sinker. There are a few of course, but most of use here seem to be in the wait and see position. Or like you or I where we know there might be some serious limitations. But, we still plan on getting it because were passionate about autonomous driving.
But, for lots of us I don't think they even needed to go all the way to FSD to excite us. They could have simply said Level 3 hands off driving when approved by local regulatory agencies. That maybe have even excited us more because we could see that happening at least with reasonably good weather.
You said not to let perfect be the enemy of good enough. But, level 3 was good enough for the moment.
What's funny about this as it relates to this conversation is that I didn't invest a lot of money.
So I spent the year adding more, and more to it.
I didn't invest a lot because I was being careful, and what is this entire conversation about?
Risk.
It's all a bit pointless talking about this without determining what self driving really means and for that matter what EAP means. For some, self driving is doing nothing from door to door like the video, but I've a suspicion based purely on being a sceptic that Full Self Driving will be most of the features which people thought might be in EAP and very few of the subtle extras in the video, like pulling out of side junctions or leaving your car to roam around a car park looking for a spot will ever be implemented on this hardware.
So should FSD be sold now? On the simple basis that we've no idea in any substantive (ie contractual) language what that means, let alone the fact it, whatever it is, doesn't currently work, then no.
So should FSD be sold now? On the simple basis that we've no idea in any substantive (ie contractual) language what that means, let alone the fact it, whatever it is, doesn't currently work, then no.
Tam
Well-Known Member
SpaceX has never sent any human into space or the earth's orbit.
It has never sent any ship beyond the earth's orbit.
It has never sent any ship to the moon.
But it has already sold out two tickets to the moon.
Thus, if it doesn't work now, it doesn't mean people don't buy them now, even with a very high price.
There is reason to speculate there are possibly three levels of EAP/FSD in development at Tesla:
1) Current EAP, imitating AP1 with one active front camera (plus ultrasonics), may eventually use backwards looking side cameras for better blind spot detection, but not much more than that...
2) Full self-driving, a separate team and effort at Tesla using all eight cameras
3) Future EAP, a subset of full self-driving code running a more advanced EAP - the eventual goal for the EAP
This may not necessarily show to the customers, but understanding the possible architectures/projects behind it may be useful. The source for this speculation: Current HW2 Autopilot using 2 of 8 cameras * Testing Inside *
This may mean any greater advances in FSD or additional autonomous EAP features may need to wait until that full self-driving is more ready to be running on the production cars... and that current AP1-like EAP is really a bit separate from all that.
That said, what may give Tesla an earlier opportunity to give FSD buyers some additional value are the stated numbers of active cameras for both systems. For EAP it is four, I guess we assume two looking forward and the backwards looking side cameras for changing lanes. FSD is promised to make use of all the eight cameras.
So basically all Tesla needs to do is use a fifth camera, for something, anything really, to give FSD owners something a bit more.
p.s. As for the ethics of this sale, I find the target descriptions of AP1/FSD having been the problem of asking pre-payment, not the fact that they offer the option to pre-pay itself. Tesla not meeting those AP1 promises, that was the unethical part. Same with FSD, if they can not deliver on what is written... If they can, offering the option is fine IMO, people can choose whether or not to pay less now or more later.
1) Current EAP, imitating AP1 with one active front camera (plus ultrasonics), may eventually use backwards looking side cameras for better blind spot detection, but not much more than that...
2) Full self-driving, a separate team and effort at Tesla using all eight cameras
3) Future EAP, a subset of full self-driving code running a more advanced EAP - the eventual goal for the EAP
This may not necessarily show to the customers, but understanding the possible architectures/projects behind it may be useful. The source for this speculation: Current HW2 Autopilot using 2 of 8 cameras * Testing Inside *
This may mean any greater advances in FSD or additional autonomous EAP features may need to wait until that full self-driving is more ready to be running on the production cars... and that current AP1-like EAP is really a bit separate from all that.
That said, what may give Tesla an earlier opportunity to give FSD buyers some additional value are the stated numbers of active cameras for both systems. For EAP it is four, I guess we assume two looking forward and the backwards looking side cameras for changing lanes. FSD is promised to make use of all the eight cameras.
So basically all Tesla needs to do is use a fifth camera, for something, anything really, to give FSD owners something a bit more.
p.s. As for the ethics of this sale, I find the target descriptions of AP1/FSD having been the problem of asking pre-payment, not the fact that they offer the option to pre-pay itself. Tesla not meeting those AP1 promises, that was the unethical part. Same with FSD, if they can not deliver on what is written... If they can, offering the option is fine IMO, people can choose whether or not to pay less now or more later.
In fairness, Elon Musk could still be wrong on what is needed for driverless driving, though. No matter how brilliant, great minds still get some things wrong. It is perfectly fair to discuss what those may be.
An entire industry is currently betting on lidar and 360 degree radar in addition to cameras (as well as ultrasonics), whereas Tesla and some other software startups have ideas of mere cameras with possible limited radar/ultrasonic assistance.
Unlike with EVs (where there are conflicts of interest when you make ICEs), the established automotive industry has no incentive to bet on the wrong horse on autonomous driving.
Are Elon Musk and a couple of software startups right and the rest of the automotive industry wrong? It is possible, but it still has to be noted, there is a lot more collective "access and influence on the power brokers" on the lidar/360 radar FSD suite than behind Tesla's.
Absolutely
Fair enough. My personal guess is that lidar and more radar will also come to Tesla 18 months from now (almost 6 months have passed since Hardware 2's release), and we'll be told that for yet another reduction in crash risk - or for bad weather capability - we should upgrade to Hardware 3. And I probably will, sigh...
I'd feel it would be a stupid move in terms of driving both sales and the motivation of his teams to reach beyond what they think themselves capable of achieving.
They absolutely can. They can't declare the money as revenue but that is purely an accounting issue.
My two cents: annoucing FSD as Elon did was an absolute masterful marketing move. Like all brilliant marketing it painted an enticing picture that made (some) customers dream of a reality that wasn't necessarily there. And like all marketing, there is blurred line between painting a nice picture for your customers versus deceiving them. Reasonable people can reasonably differ of where that blurred line exactly lies. As time goes on while Tesla doesn't deliver EAP or FSD more people will start to ask questions like OP.
Similar threads
