FSD rewrite will go out on Oct 20 to limited beta

[Knightshade]

According to John, who's been posting the videos on Tesla Owners SV, the car goes back to normal NoA once it enters the freeway. The new FSD beta visualizations change back to the normal NoA visualizations once on highway / freeway.

I think it's going to take Tesla some time to polish up everything into an unified approach, probably 6-9 months.

IIRC Green mentioned the FSD Beta firmware (2020.40.8.10) was just the same 2020.40.8 everyone has, with the drive-on-streets stuff slapped on top of it, so that's not too surprising nothing else changed.

On that twitter link he says he will post in TeslaClubLA so I believe he is located in Los Angeles. Thus far everyone has been in Sacramento/Bay Area so maybe they are expanding geographically?

Edit: His Twitter bio says he lives in Los Angeles. So looks like we'll get some videos from SoCal.

If I had to guess, and it's only that, they've tested the most in CA so that's where the early public beta release is happening.

You obviously haven't watched the 2016 Tesla FSD video.

You mean the one they needed to drive 500 miles, repeating the route dozens and dozens of times, in order to get enough short clips they could edit together to fake it looking like a single take?

Interesting point to consider. Shouldn't high growth companies like Tesla prefer the up front cash to fund growth?

We'll find out next year when monthly prices are revealed. My wild guess is per month FSD will cost 2.5% to 3.0% of the flat amount.

Tesla has like 14.5 billion in cash- more cash than debt at this point (I think the only car company that's true of?)

They mentioned on the Q3 earnings call they're literally spending cash as fast as they can in any way that makes sense, and yet cash on hand keeps increasing.

 

[Bladerskb]

I’m actually taking it easy. I’m not celebrating or popping bottles. I tempered my statement with the fact that is worse than Waymo right now and that I expect some mistakes.

My statement was for those that doubted that this even existed (there were a fair number of people who expressed that). I’m all for criticizing things that are issues, but too many were blindly making statements when they had no clue what the future held. The verdict is still out on who ultimately reaches level 5... but for those who doubted and said it was vaporware, eat a sock. When I’m wrong and data is shown to me, I have and will eat my own sock.

No one doubted you can have a system with only camera that was capable of performing all the functions of driving on the road. We have seen dozens of Mobileye videos with absolutely NO lidar or radar. What we are saying is, its currently impossible to get the reliability to around ~1 million miles per disengagement using only cameras. The videos being posted actually proves our points.

 

[Bladerskb]

You are forgetting that Waymo is only driverless in well mapped modern cities like Phoenix. Do they have any others?

Tesla is general purpose and works anywhere. Quite different.

Edit: And Tesla doesn't have $100K of Lidar and other crap on it. Waymo effectively has prototype cars (how many, 4?), while Tesla has production cars (hundreds of thousands). HUGE difference right there.

Waymo's sensor suite cost anywhere from $20k to $40k. No where near $100k and they have 600 cars. With most driving around with absolutely no drivers.

 

[bdub85]

Teslascope asks you what your HW version is, so the likelihood of that being false is pretty big

Ahh yeah then that makes sense. It's probably wrong in Teslascope.

 

[zecar]

You mean the one they needed to drive 500 miles, repeating the route dozens and dozens of times, in order to get enough short clips they could edit together to fake it looking like a single take?

Wasn't the 2016 demo done on the nvidia board and software?

 

[qdeathstar]

Moving the goal post? What? everything we told you came true. You are looking at ~10 milers per disengagement versus waymo's ~1 million miles per disengagement.

1 million miles over the same 50 miles of road.

 

[tivoboy]

Smart Summon the new FSD Beta.

This is getting close to the quality that we’ll get once we begin to start teaching chimpanzees to drive.. after a few years of training. ;-)

 

[Webeevdrivers]

looks like he’s had the most problems so far of all the vids I’ve watched. He was forced to take over quite a few times. Some very impressive stuff nonetheless (how many objects were detected, how the car detected the truck with the crap parking job and maneuvered around it), but some very clear issues, the most important of which was the car apparently not detecting the dude speeding while it was at a stop sign and starting forward. The car might have stopped when it did detect the car, but the driver did the right thing not chancing it. Sure hope that they can fix that detection better when cross traffic is speeding.

Not sure if it has changed but I was under the impression that FSD should not be used in construction areas. At least one of the fails appears to be in a construction area. Possibly two. Hard to tell in the light.

 

[M109Rider]

Yes. Krafcik said that Waymo is planning to launch public driverless ride-hailing in San Francisco or Los Angeles next, then maybe Miami after that.

Wow.

 

[Knightshade]

Wasn't the 2016 demo done on the nvidia board and software?

Probably why they needed so many runs of the route to be able to fake a single-take video of it working properly.

 

[mikes_fsd]

Not sure if it has changed but I was under the impression that FSD should not be used in construction areas. At least one of the fails appears to be in a construction area.

Why not? There are a lot of disengagements in these videos. (for now this is expected)
And many have nothing to do with construction.

FSD must handle construction, even full on blocked roads and re-route.

 

[Webeevdrivers]

Why not? There are a lot of disengagements in these videos. (for now this is expected)
And many have nothing to do with construction.

FSD must handle construction, even full on blocked roads and re-route.

Of course. But what I’m asking is if the manufacture recommends it not be used in construction zones. In which case it shouldn’t as it is outside of the manufacturers design parameters.

It would be nice if it worked someday in slippery snowy winter conditions as well but I would never consider using it presently under those conditions.

I seem to recall reading something about these kind of conditions when we got the car but that was a while back. We both are in the habit of turning off Autopilot as soon as we are in construction it slipper conditions.

 

[tivoboy]

Something about all these videos makes me wonder, were people told to only record their experiences at night? Or is it just so much excitement from users and they all got it just prior or after nightfall? maybe PR wanted everyone to be able to clearly see the Matrix style visualizations clearly?

 

[john5520]

Something about all these videos makes me wonder, were people told to only record their experiences at night? Or is it just so much excitement from users and they all got it just prior or after nightfall? maybe PR wanted everyone to be able to clearly see the Matrix style visualizations clearly?

Could be. But I'd figure nighttime would be more challenging for the recognition system and is probably a better time to gauge how well the system is coping. Just a guess.

 

[82bert]

There are more daytime videos already, so certainly that doesn’t seem to be a “thing”. Pretty sure it came late and people were understandably excited to test it out.

 

[tivoboy]

Could be. But I'd figure nighttime would be more challenging for the recognition system and is probably a better time to gauge how well the system is coping. Just a guess.

Also, fewer cars on the road, and fewer pedestrians, bikes, etc. Might have been a directive.

 

[Rockster]

Those of us in M3 and MY prefer stuff in landscape mode.

Seeing video taken vertically gives me hives. It just needs to stop.

 

[bdub85]

Daytime roundabout

https://twitter.com/kimpaquette/status/1319663469860356096?s=20

 

[tivoboy]

Does anyone know if these early beta access users are clearly "early adopter" status, or just a more distributed target audience of users?

 

[bdub85]

Does anyone know if these early beta access users are clearly "early adopter" status, or just a more distributed target audience of users?

They seem to be targeting people who have large followings on social media.