-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Will Tesla ever do LIDAR?
- Thread starter diplomat33
- Start date
That’s true, that is the topic of this thread, and not my question. I didn’t mean to suggest it was the topic of this thread. On that note, I don’t expect Tesla under Musk to use Lidar either, no.
What you miss — and what I feel Musk has gambled against mostly for cost reasons — is that machine learning can be used to improve sensor redundancy as well. That too is something you can continually improve through machine learning and the end-result is a more redundant system.
So goes the theory anyway. Nobody knows for sure what mix will win in the end.
No. I think it is rather important to understand my meaning.
It is VERY different to claim SpaceX lands on Mars before Tesla delivers FSD... than to say, which I meant, I feel SpaceX’s PR is more accurate on Mars than Tesla’s PR is on FSD. In other words, I believe SpaceX’s PR on Mars more, than I believe Tesla’s PR on FSD.
Very different things.
StealthP3D
Well-Known Member
You don't always get an award when you win (unless you're a Millenial
).If you've been following the field of autonomy, you will see more and more of the experts, as they gain experience, are tending towards the path of machine vision carrying more of the load. It has to do with the way machine vision can harness the power of AI and the fact that the world of driving is designed around vision.
I don’t see anyone here arguing against that. Obviously machine learning and vision are increasingly important part of autonomy, of any kind.
That is not where Musk ended his argument though. He went on to say Lidar was useless (appendices...) and HD maps too — on that there may not be quite as much agreement.
StealthP3D
Well-Known Member
I think Musk is Musk whether he's talking about interplanetary space travel or autonomous driving. He tends to be realistic on everything except for the exact timeframe.
And he has made a fool of more people that have doubted he could achieve a certain goal than anyone I can think of. So naysay at your own risk.
StealthP3D
Well-Known Member
That is not where Musk ended his argument though. He went on to say Lidar was useless (appendices...) and HD maps too — on that there may not be quite as much agreement.
The context in which he referred to LIDAR as useless appendices and the folly of HD maps was relying on them to get you to full autonomous driving. He believes the focus needs to be on improving the artificial intelligence and processing reams of data from extraneous sensors is counterproductive to that goal. In the end, the reason humans can drive at all is because they can make sense of the visual driving environment. Once you have enough insight into Tesla's AI and machine vision technology you have a pretty good idea of what it's capable of with continuing training. And I have to believe that Musk has more insight into their technology than anyone else who is publically proclaiming an opinion.
Pretty much the only people who think Musk doesn't know what he's talking about are those who think he's a lying fraudster. For crying out loud, he launches rockets into space and re-lands the boosters on unmanned ships bobbing in the sea!
StealthP3D
Well-Known Member
As I said, he tends to be realistic on pretty much everything except for the exact timeframe.
Timeframe does become a problem when it involves a product already sold — that can not usually be changed (hardware-wise), if it needs changing for the goal to be reached. That’s why it is also relevant here today: The cars have been sold already and are sold every day...
Many Tesla products like P85D and various P90DL revisions that never reached Musk’s announcements are testament to this. P100D finally made it but that was after many people were sold cars on Musk’s lofty words that did not meet their specs.
How many cars sold with EAP and FSD will end their leases or be sold before even reaching EAP and FSD features as marketed by Tesla/Musk in late 2016 early 2017? Many. NoA + Smart Summon (aka EAP) is not even available in most markets yet, let alone any FSD differentiating features promised by summer 2017. These are cars reaching 3 years this year.
What will the Model 3s sold today get in their ownership/lease period... timeframes matter for the customers.
When FSD finally is here, how different will the hardware sold for it actually be compared to where it was in 2016 when Musk promised the same things...?
Last edited:
StealthP3D
Well-Known Member
But I never said being later than projected was not a bad thing. He would do well to temper his timelines. The point is, he does have extraordinary vision and insight into how to achieve complex feats that many say are impossible or impractical. Then he makes complete fools of those same people who said he couldn't do it. FSD will be just one more feather in his cap.
That would be nice for our cars so lets hope. However that isn’t quite the point in this thread I think: For Musk to be right about Lidar would not just mean delivering FSD but also others failing with or indeed because of Lidar. So it is not just that he is making a bold claim for Tesla, he is also ridiculing the choices of others in the business.
Daniel in SD
(supervised)
If he delivers anything close to what he is promising next year then I would say that everyone else working on autonomous vehicles is DOOMED.
I suggest that everyone who thinks that should use half their life savings to buy call options on Tesla. If his predictions come true I would expect the the stock to be worth $1000 on Jan 15th 2021. You can buy a $600 call option for $5 so you could get a 7900% return!
CarlK
Active Member
StealthP3D
Well-Known Member
I have to call that out as really bad advice. Even if you believe Musk's timeline. Even if you believe his timeline. Terrible investing advice. For obvious reasons, you don't put 1/2 life savings into any instrument that speculative.
That said, I don't recall one person here who said they thought his timelines were likely. People believe his FSD estimations are probably too optimistic. So please don't mix up two separate questions: 1) Timeline, 2) first successful solution. Your suggestion of putting half of your life savings into 1/21 calls a very ill-advised move either way.
Unless half of your life savings is easily replaced and you're not afraid to lose it. Because it's definitely not out the question for those $600 call options to pay off handsomely. IMHO, this is actually a fun bet with any dollars that you don't mind taking a big chance with because it favorably leverages the overly pessimistic opinions regarding the future value of TSLA into potentially fabulous returns. No one should place bets on a roulette wheel when bets like this are available!
StealthP3D
Well-Known Member
There's nothing wrong with ridiculing poor choices when you know better. I would say it's actually his DUTY to call them out, try to wake them up. It's up to them whether they want to listen or not. In the end, the results will speak for themselves. He's certainly not hurting them by giving his honest appraisal.
Perhaps, perhaps not. My point was merely that for Musk to be right, he not only has to succeed with FSD, the approach of others’ with Lidar will have to be wrong. In other words: If others succeed with Lidar too, then Musk was wrong and there were many paths to success.
He is of course not doing it for any such altruistic reasons. He is doing it to highlight Tesla’s (alleged) excellence. I agree he is not really hurting anyone with it because his words are in general not taken seriously enough on this for it to matter at this time — the market reaction made that clear enough.
Still, personally I think it is perhaps an unnecessarily bold claim. I don’t think anyone really knows yet what is optimally needed for autonomous driving. It is one thing to not choose to use Lidar, it is quite another to be seen actively campaigning against its use, which is something Musk has become known for.
As for me, I’m less convinced that he is right on Lidar use in general, than I am of Tesla’s lack of it in particular.
Monday musings:
I wonder if maybe Musk is correct that LIDAR is not strictly necessary for FSD but that LIDAR is not the actual stumbling block either. After all, there are different ways for a self-driving car to "see" the world. LIDAR is perhaps the easiest and most obvious way but it is not the only way. LIDAR is just a shortcut to getting to the more challenging parts of FSD. We see that in the companies that used LIDAR. They solved the vision part, but are now stuck on some tough edge cases. The stumbling block of FSD is not seeing and tracking the environment around the car, it's what you do with that information, ie making those intelligent driving decisions in an often chaotic and even unpredictable driving world. Perhaps, Tesla Vision will actually work too but then Tesla will also get stuck on solving those last pesky edge cases?
I wonder if maybe Musk is correct that LIDAR is not strictly necessary for FSD but that LIDAR is not the actual stumbling block either. After all, there are different ways for a self-driving car to "see" the world. LIDAR is perhaps the easiest and most obvious way but it is not the only way. LIDAR is just a shortcut to getting to the more challenging parts of FSD. We see that in the companies that used LIDAR. They solved the vision part, but are now stuck on some tough edge cases. The stumbling block of FSD is not seeing and tracking the environment around the car, it's what you do with that information, ie making those intelligent driving decisions in an often chaotic and even unpredictable driving world. Perhaps, Tesla Vision will actually work too but then Tesla will also get stuck on solving those last pesky edge cases?
Edge cases will be a challenge. A lot of human drivers don't do terribly well with a lot of the edge cases, even when wide awake and completely focused.
I'm sure Tesla will struggle with some of them too. Their advantage from the fleet means that they will see more examples of those cases, more quickly, which gives them the best chance of learning a solution.
Yes, I remain optimistic that Tesla's big fleet will really help them to solve those difficult edge cases.
Similar threads
- Article
- Article
- Article
