FSD Beta Videos (and questions for FSD Beta drivers)

[gearchruncher]

Reality is the vast majority of those fatalities are happening in countries that are far less wealthy and less developed, the stats are publicly available

Technologies such as seat belts, airbags, stability control, ABS, AEB, etc are already proven highly effective at reducing deaths. Tesla's own numbers show this. Until these are standard in these countries you discuss, there is zero way they can afford FSD. FSD is not the $10K software cost- it's a lot of hardware in the car that you don't see the cost of buying.

The way to make safety things cheaper always follows the same path:
1) Some company makes it actually work (note, Tesla has not done this yet and isn't close at all)
2) They sell it for a premium (airbags, abs, seat belts, traction control, backup cameras were all premium features originally)
3) The volume from this lowers the price
4) Companies start competing over price or making it a standard feature
5) Enough cars get it to show it is statistically and financially worth it
6) Governments start mandating it
7) Price falls further

The real way Tesla can save all those lives is to make FSD actually work, and ship it to more than 71 people. The current price is irrelevant, once it works, competition will take over.

 

[AndreP]

Technologies such as seat belts, airbags, stability control, ABS, AEB, etc are already proven highly effective at reducing deaths. Tesla's own numbers show this. Until these are standard in these countries you discuss, there is zero way they can afford FSD. FSD is not the $10K software cost- it's a lot of hardware in the car that you don't see the cost of buying.

The way to make safety things cheaper always follows the same path:
1) Some company makes it actually work (note, Tesla has not done this yet and isn't close at all)
2) They sell it for a premium (airbags, abs, seat belts, traction control, backup cameras were all premium features originally)
3) The volume from this lowers the price
4) Companies start competing over price or making it a standard feature
5) Enough cars get it to show it is statistically and financially worth it
6) Governments start mandating it
7) Price falls further

The real way Tesla can save all those lives is to make FSD actually work, and ship it to more than 71 people. The current price is irrelevant, once it works, competition will take over.

I'd wonder how much of the FSD hardware is in all Tesla vehicles by default and if that's part of what's driving the price higher.

When you're deploying hardware specific to a function but the vast majority of people aren't paying for that functionality, the few who do pay are essentially subsidizing everyone else's hardware at the front end.

Conversely, if more people were paying for functionality that utilizes hardware that's in all cars by default, you'd expect the price to drop as many more people are sharing the burden of those costs.

 

[Knightshade]

I'd wonder how much of the FSD hardware is in all Tesla vehicles by default and if that's part of what's driving the price higher.

When you're deploying hardware specific to a function but the vast majority of people aren't paying for that functionality, the few who do pay are essentially subsidizing everyone else's hardware at the front end.

Conversely, if more people were paying for functionality that utilizes hardware that's in all cars by default, you'd expect the price to drop as many more people are sharing the burden of those costs.

All cars get the same HW (when built at the same time- obviously cars in 2016 got older stuff than cars built in 2020)

This is what has enabled Teslas massive data collection ability- since they don't need the owner to have FSD to collect data used to improve FSD.

 

[linux-works]

its usually a win to NOT have multiple board types and just write different firmware to them to turn features on and off.

given that, all hw would be designed to support the max config, at least that's what same vendors do.

 

[gearchruncher]

I'd wonder how much of the FSD hardware is in all Tesla vehicles by default and if that's part of what's driving the price higher.

When you're deploying hardware specific to a function but the vast majority of people aren't paying for that functionality, the few who do pay are essentially subsidizing everyone else's hardware at the front end.

Base Autopilot is standard in all Teslas. The computer, some cameras, steering rack, brake booster, and ultrasonics are all needed for those functions, even if someone does not buy FSD. Only a couple of the cameras are extra if you don't buy the FSD license.

 

[AndreP]

Base Autopilot is standard in all Teslas. The computer, some cameras, steering rack, brake booster, and ultrasonics are all needed for those functions, even if someone does not buy FSD. Only a couple of the cameras are extra if you don't buy the FSD license.

Didn't you just say the $10k FSD cost constitutes a bunch of hardware that we're not seeing? Maybe I misread that

If only a couple of cameras are being added with the FSD license, what is this hardware cost?

 

[qdeathstar]

I think (and Elon tweeted some time ago) that eventually it won’t be an option on any new Tesla but rather rolled into the purchase price. It will still be expensive but it won’t seem like it as much if it’s bundled.

He said it was going to cost more and be worth 100k

 

[Knightshade]

Didn't you just say the $10k FSD cost constitutes a bunch of hardware that we're not seeing? Maybe I misread that

If only a couple of cameras are being added with the FSD license, what is this hardware cost?

The physical HW is identical in all cars. Nothing gets "added" HW-wise with buying FSD other than if you have an old pre-3 driving computer, they swap in the new computer for the old one.

I think he was suggesting that only if you have FSD are a few of the cameras "used" but that's not really true- they're all on all the time for data collection purposes (and things like blind spot warning, FCW, etc)

 

[Chazman92]

#FSDBeta 9.1 - 2021.4.18.13 - 360 Degree Camera Viewing Angle Test using Drone View

 

[linux-works]

really love the idea of drone-view in the tests.

 

[ZeApelido]

#FSDBeta 9.1 - 2021.4.18.13 - 360 Degree Camera Viewing Angle Test using Drone View

Can you provide the google map location of this intersection?

 

[Daniel in SD]

Can you provide the google map location of this intersection?

Google Maps

Find local businesses, view maps and get driving directions in Google Maps.

www.google.com

 

[mhan00]

#FSDBeta 9.1 - 2021.4.18.13 - 360 Degree Camera Viewing Angle Test using Drone View

It looked like the car was seeing pretty far down the street with the pillar cam.

 

[MrBadger]

#FSDBeta 9.1 - 2021.4.18.13 - 360 Degree Camera Viewing Angle Test using Drone View

Could you put some cones and wheelie bins up the road with the problematic intersection and see if they get detected or are in a blind spot?

OK, so that is meant as a joke, but it may not be as stupid as it sounds.

Interesting video btw.

 

[linux-works]

again, I have to say, there is not enough overlap and redundancy in the cameras.

not enough to BET YOUR LIFE ON IT.

I will keep saying this until telsa finally gives in and adds more sensors to the outside of the car.

 

[linux-works]

I don't see where my car needs more sensors.

I understand you feel this way. many people feel this way. many people dont know the subject and are just tesla fans.

do you have any data or is this just your gut feeling?

 

[AlanSubie4Life]

It seems to me that the "stitching" is either not done for the rear facing cameras. This jumping around does not happen when he walks around the front half of the car.

There are multiple BEV NNs and they don’t all use all the cameras, it seems? No idea if this is related to your observations though.

https://twitter.com/x/status/1423361674963259397

 

[linux-works]

I am in warehouse automation, we design and build software that directly drives the hardware... before you start lecturing me on some hypothetical.

I hope I'm not near any of the things you design, then. what is sufficient sensor for you is CLEARLY not enough for my liking.

btw, I do work in this very field. not for warehouses, but for cars that people drive, so the stakes are as high, or higher.

but thanks for playing.

 

[linux-works]

Good thing that the world is not dictated by your whims and tastes.

ok, I'll let you have the last word. I know you need that.

*(oops)*

 

[mhan00]

360 degree video

1:50 - car under bridge gets confused by the sun patterns and pulls up short at a stop ligh, waaaaay too far back. Driver disengages And moves up to the line, re-engages and the car tries to run the red.

2:50 - car continues to be confused by the sun patterns on the road. Car wants to make a left but it’s displaying a sun pattern as a solid yellow so it won’t change lanes to the left turn lane. Driver disengages.

4:30 - car doesn’t change into the right turn lane and still executes the turn from the wrong lane. Driver disengages.

5.00 - car drives by a number of people jay walking. I’m really surprised it didn’t stop. The people were shown it just kinda keeps driving. Driver doesn’t disengage or comment on it, so maybe I was missing something.

7:30 - car is stopped at a stoplight and is displaying a car waiting behind it. If you rotate to see the view behind the car, there is no car there so I have no idea why the car thinks there is another one behind it.

11:00 - major screw up. Crossing an intersection the car starts drifting to the left. It was either drifting into the island or worse the car sitting in the opposing left turn lane. Driver disengages.

11:35 - car fails a right turn because it doesn’t appear to see the pillars and looked like it was going to turn right into it. Driver disengages. So it’s not just the pillars from the Hyperchange video that Tesla doesn’t see.

12:50 - car fails again on a slightly complex move to turn right and then slight turn left. It looked like it was aiming towards the car in the opposing lane of traffic. Driver disengages.

There are some more screw ups, but I’m out of time. This was a bad drive. Cool camera and well done by the driver to stay alert, but FSD Beta failed badly here.