FSD Beta Videos (and questions for FSD Beta drivers)

[WattPwr]

Some of the information being recently posted on this thread is not of a high quality. Yes, some of the information may be from Tesla's own website, but this website often contains information that is years old (see the issues concerning Tesla's advertising of Premium Connectivity). Years ago it was stated that Tesla was moving from Radar to vision as a primary sensor (I think it was after the first fatality), with Radar and ultrasonics simply confirming to the computer that its predictions were correct. It has been long known that Tesla plans to move to a fully vision based system, with Radar confirming/assisting in some circumstances. Therefore, the ultimate range of cameras will have more to do with pixels and software perception limitations than whether they are placed on a roof and called 360 degree vision.

Research is non-linear and difficult to extrapolate. Elon has optimistic timelines. However, when you are venturing into the unknown, you can't know what is going to happen. It is impossible. You can only guess. Am I disappointed in the rate of advancement of FSD (something I paid for years ago)? Yes, I am. However, some of the critiques on this page indicate a very poor understanding of the groundbreaking challenges.

 

[Battpower]

when you are venturing into the unknown, you can't know what is going to happen.

I agree that negativity for its own sake is pointless. We should all want Elon to be right, shouldn't we?

But this is not Kickstarter either! $10k to be part of something that may not deliver during your ownership of the car is a pretty big loss to swallow.

I would be far happier if when a subscription model is introduced, they offer existing FSD owners the option to convert to a lifetime subscription at no cost.

 

[Dan Detweiler]

But this is not Kickstarter either! $10k to be part of something that may not deliver during your ownership of the car is a pretty big loss to swallow.

Just to be clear, we all (me included) ponied up our money knowing full well that there was no hard guarantee of any timeline. I'm anxious too, but I realize I plunked my money down for FSD knowing I might not see the fruits of that decision within the lifetime of my ownership.

Dan

 

[run-the-joules]

Just to be clear, we all (me included) ponied up our money knowing full well that there was no hard guarantee of any timeline. I'm anxious too, but I realize I plunked my money down for FSD knowing I might not see the fruits of that decision within the lifetime of my ownership.

Dan

Sure, but I think most of us expected that the company would at least try and take care of their customers in exchange for our early faith and dollars. But instead, they just basically tell us to go F ourselves, when they have a number of options at hand to make things right. Instead, I found myself not buying a Model Y like I had planned, so they lost out on a sale and additional revenue, and I *know* I'm not the only person who's decided to stick with their old car.

 

[yerEVan]

Wasn’t FSD promised at an earlier date when they went crazy with sales in 2nd Q 2018, on the website during the ordering process? NOA and autopilot on side roads is decent, but it’s not worth 10k. I didn’t invest in the technology to never use it. I invested in it because it was promised at an earlier date.

I’m not complaining, but I agree that I truly expected to have something more than what is currently offered by now. I doubt in the next few months that we will have a better version of what the beta testers have right now (released to the public).

But I’m pretty patient, and am glad I can sell FSD with my car. And again, not complaining here, but I cannot disagree with @Battpower

 

[n.one.one]

 

[TIppy]

ears ago it was stated that Tesla was moving from Radar to vision as a primary sensor (I think it was after the first fatality), with Radar and ultrasonics simply confirming to the computer that its predictions were correct.

The quote from Elon about only needing radar in the forward direction because of high forward speed was from the Apr. 2019 autonomy day presentation.

 

[Battpower]

Just to be clear, we all (me included) ponied up our money knowing full well that there was no hard guarantee of any timeline. I'm anxious too, but I realize I plunked my money down for FSD knowing I might not see the fruits of that decision within the lifetime of my ownership.

Dan

I agree, but it surely makes for a tricky decision if / when replacing a car that you bought with FSD. Once bitten, twice shy.

If FSD actually had a solid core functionality then it would less of a concern. Having the choice to retain FSD or sell with the car would cover all options but not very agreeable to be forced to take a hit selling an FSD car if you intend buying FSD again on your next car.

 

[Dan D.]

Some of the information being recently posted on this thread is not of a high quality. Yes, some of the information may be from Tesla's own website, but this website often contains information that is years old ... It has been long known that Tesla plans to move to a fully vision based system, with Radar confirming/assisting in some circumstances. Therefore, the ultimate range of cameras will have more to do with pixels and software perception limitations than whether they are placed on a roof and called 360 degree vision.

Research is non-linear and difficult to extrapolate. Elon has optimistic timelines. ... However, some of the critiques on this page indicate a very poor understanding of the groundbreaking challenges.

When you have painted yourself into a corner by claiming your cars have the hardware to perform FSD you are then in the unenviable position of having to create novel software using this hardware. Yet, they are doing wonders.

A better strategy might be to build the best system first and then install it into your cars. Tesla chose their path and hopefully they can achieve it. IMO it's a stubborn method to design FSD software using existing hardware, without fully knowing beforehand where the challenges will be found.

Perhaps they could have used headlight sideview cameras. Perhaps narrow view rear cameras. Perhaps 50-80m is not really enough. Perhaps it is. Thing is, they have installed the cameras and are kind of stuck in having to use them and only them for the current software cycles. All the groundbreaking software you can write may still find the cameras are visually impaired for the task. Or maybe not. I'd love seeing them succeed too.

 

[WattPwr]

When you have painted yourself into a corner by claiming your cars have the hardware to perform FSD you are then in the unenviable position of having to create novel software using this hardware. Yet, they are doing wonders.

A better strategy might be to build the best system first and then install it into your cars. Tesla chose their path and hopefully they can achieve it. IMO it's a stubborn method to design FSD software using existing hardware, without fully knowing beforehand where the challenges will be found.

Perhaps they could have used headlight sideview cameras. Perhaps narrow view rear cameras. Perhaps 50-80m is not really enough. Perhaps it is. Thing is, they have installed the cameras and are kind of stuck in having to use them and only them for the current software cycles. All the groundbreaking software you can write may still find the cameras are visually impaired for the task. Or maybe not. I'd love seeing them succeed too.

If Elon had not made these exciting claims we wouldn't be here disagreeing. If Elon hadn't sparked our imagination by guessing (and we gambled along with him) we wouldn't be passionately trying to guess whether our hard earned dollars were wasted or not. It isn't a technical advantage Elon brings, but a passionate excitement that drives engineers to try to do more than they thought was possible. And then he demands more. So yes, I think Elon and Tesla enjoy painting themselves into a corner. Isn't it exciting. The strategy you suggest would not succeed because it has always failed to make radical advancements. History proves this.

As for the technology/hardware not being perfect, all engineering design incorporates iteration at all levels of design, even on complete production ready models. This is why we have new car models each year. If Tesla had chosen the safe route you wouldn't be here discussing it. You wouldn't care. And a self-driving car based on vision technology would be maybe 50 plus years away and battery cars would still be an unbelievable joke.

This isn't a technical problem we are discussing. Heck, it isn't even a socio-technical problem. It is a problem pertaining to inspiration and risk.

 

[run-the-joules]

This isn't a technical problem we are discussing. Heck, it isn't even a socio-technical problem. It is a problem pertaining to inspiration and risk.

it's also an ethics problem and that is where Elon is falling flat on his face. Very easily solved, at no cost to Tesla, but they'd rather just shrug it off and burn bridges.

 

[Electroman]

Thanks to everyone posting videos here..much appreciated

 

[Todd Burch]

Have a listen at 8:16 where Elon says radar is needed in the forward direction because of high speeds. He then says they don't need any other sensors. But speeds from the side because of other vehicles can be just as high as the forward speed of your vehicle.

Relative speeds are what matter here. With oncoming traffic, you have closing speeds of 120MPH and potentially greater. All other scenarios are typically half that or less.

 

[mark95476]

This is guaranteed: Whatever TSLA releases won't be "true" FSD. The self-proclaimed purists will be moving goal-posts, splitting hairs, lying/misinterpreting, ...etc...

The ~$38B the shorts have lost in 2020 is continuing to sponsor a lot of trolls and low quality FUD here and elsewhere.

- TSLA added to SP500
- TSLA at all-time highs
- TSLA hit 500k production
- Excellent margins
- 2 new factories coming online
- Multiple FSD videos being posted daily

I understand the intense media dislike since TSLA doesn't market and advertise normally. The media has been unsuccessful in shake-downing TSLA for their $$$.

 

[TIppy]

Relative speeds are what matter here. With oncoming traffic, you have closing speeds of 120MPH and potentially greater. All other scenarios are typically half that or less.

That's not what he said. You can listen at 8 minutes and 16 seconds where he says, "You only need radar in the forward direction cause that's the only direction you're going real fast." Nothing about closing speeds; you're going real fast. Apparently, the cars forward speed is enough to make radar necessary, but of course it would certainly be even more so for closing speeds.

I'm not sure this is a real issue. It just struck me when I heard him make this comment, because it is surely incorrect. I was being somewhat provocative with my post. Generally, I'm a fan boy and can't wait to get this version of fsd.

Earlier in the same talk, someone asked him about there being no radar to the side, and he talked about the procedure the car uses with the cameras to enter crossing traffic. He totally ignored the radar part of the question. A little later he circled back, I believe he was shooting from the hip, to give a reason why they don't need sideways looking radar. I think in the heat of the moment to answer the side radar question, he might have gotten it wrong. But if radar is indeed required for the forward direction, they are going to have a problem with crossing traffic.

 

[JulienW]

Hi all,

I’m one of the lucky FSD Beta testers. Since we are seeing a lot of videos from Beta testers now, I was curious to see what YOU all wanted to see now. I’ll be happy to review and test those ideas. Note that I might not be able to do everything based on the agreement we sign with Tesla.

Here are a few ideas that I already had:
- Record a “regular” drive but in French (for French Canadian friends)
- Take the same turns/routes and try to replicate the same condition to see how FSD evolves through the versions. I have to say I started on Beta 7 and we are on 9 now so not much to compare. Plus I don’t have a lot of footage for Beta 7 so it could take a while before this can happen
- I liked the drone FSD footage. Maybe do another one but with a stand still drone pointing camera down so we can see how well the car takes turns.

Thanks!

Just watched you latest and...

• Can you get a longer/lower camera mount? It is nice seeing your handsome face but the mirror blocks a lot of the upcoming road, traffic signals and don't need to see visors/headliner.
• Also a little steadier mount? I get the roads there are rough but a lot of shaking going on.
• More snow and ice covered roads
• Like the inner city driving
• Make a long routine route by making each new Start Favorite at the end of the preceding route.
• Can you give an English translation too?
• Maybe a one a week or so with drone video

...and of course I'm an expert and know it all sitting in my basement. THX

 

[aronth5]

Wasn’t FSD promised at an earlier date when they went crazy with sales in 2nd Q 2018, on the website during the ordering process? NOA and autopilot on side roads is decent, but it’s not worth 10k. I didn’t invest in the technology to never use it. I invested in it because it was promised at an earlier date.

I’m not complaining, but I agree that I truly expected to have something more than what is currently offered by now. I doubt in the next few months that we will have a better version of what the beta testers have right now (released to the public).

But I’m pretty patient, and am glad I can sell FSD with my car. And again, not complaining here, but I cannot disagree with @Battpower

What I find interesting is Tesla's opportunity when they released the Model Y to make hardware changes. One would think by then they had enough data to know whether the hardware was adequate.
They didn't so I'm hoping they are confident they can handle the difficult issues that FSD beta testers are posting, particularly the unprotected left hand turns.
On the other hand can you imagine the uproar from existing owners if Tesla had added cameras and/or radar locations to the Model Y and wondering what that meant for everyone else?. Now that would have resulted in some interesting posts!
Same question for the Cybertruck. Will Tesla make hardware changes? That will be telling.

 

[82bert]

Most of the valuation seems to be based on FSD revenues but if they don't come it's all going to collapse.

This is simply not true.

 

[buyleonard]

I didn't follow the full story on whether Elon and/or Tesla really believe that only forward approaching traffic is the only important traffic with a need for radar to detect the speed differential. However if true it could illustrate another stunningly poor assumption to add to the list.

In the forward direction, you have your car's speed + the traffic driving in the opposite direction. So if both are traveling 65, that's a 130 mph change that the car has to be able to 'see'. That is why forward needs radar and side doesn't,

 

[mikes_fsd]

When you have painted yourself into a corner by claiming your cars have the hardware to perform FSD you are then in the unenviable position of having to create novel software using this hardware. Yet, they are doing wonders.

Just because you are unable to see how to solve a problem with the hardware set that Tesla has delivered in ~1million cars and counting, does not mean that Tesla has painted themselves into a corner, it just means that you've limited ability to see the solution that they are working on.