FSD Beta Videos (and questions for FSD Beta drivers)

[mark95476]

Cars are very safe these days and Tesla is almong the safest. Look at all of the crash testing and Tesla gets top grades consistently.

People not driving Teslas die every single day, but that's expected and common so it's ignored.

Accidents at autosteer city speeds likely won't cause death or serious injury to occupants.

If you want to avoid death, don't drive on freeways due to the speeds involved. I use AP on freeways almost daily and it's super safe.

Tesla/Elon are hyper aware of the trolls that are ready to pounce on anything negative they can spin. FUD and bald-faced lies are old hat and they're used to it. I'm confident they won't fall for TSLAQ and trolls goading them to release autosteer city widely unless they're very sure it's safe.

Exactly. If the failures are fatal, it is unacceptable. I feel like I'm taking crazy pills: are people really saying that a device that is documented to cause fatal mistakes within 24 hours of a very limited release is ok to release to the public (??)

I know it's disappointing. But if you find yourself excusing these problems, please reconsider your intentions and potential subconscious biases. I'm rooting for Tesla too, but we have to stay grounded in reality.

Once FSD can prove that it won't drive straight into solid objects, and demonstrates a high level of safety when it fails (maybe dropping into a failsafe autosteer-only mode, or pulling over to the side, etc), then a wider release is more appropriate. I'm not saying it has to be perfect. It just has to not kill people.

 

[gearchruncher]

As is the design that the Tesla driver intervene!

I love the goalposts move by Tesla supporters.

Full Self Driving is clearly meant to be a full L4 system. It was sold this way until 2019. It's name says that. Elon says napping in a car is solved.

...But the system isn't "designed" for it, so it can do whatever it wants, and all these mistakes it's making are no big deal. FSD is just a marketing name. Read the disclaimers. Nothing you can do about stupid people, it's why electrical cords come with bare wires on the end and steering wheels have knives sticking out of them.

...But Tesla is really close to actual FSD, and are the clear leaders, everyone else is crap, they just need billions of miles of people driving on this software that tries to kill them, so they can prove it won't kill them.

All you're doing by saying "it's safe because look how many times people have taken over without a crash" is saying that you know the current "FSD" is still just a party trick. It demonstrates just how far away from true FSD Tesla is. There is zero way Tesla / Elon thought this is where they would be in 2021, so it's only "designed" this way because they have been unable to achieve L4, not because they set out for L2 at the outset.

You really think there was ever a design goal set at Tesla for "An L2 city driving system that fatally fails about once every 10 miles unless a driver intervenes"? It wasn't designed this way. The design failures have been mitigated by the presence of a human, not designed in from the beginning.

 

[Dan D.]

Accidents at autosteer city speeds likely won't cause death or serious injury to occupants.

I'm assuming you also count pedestrians and other road users. Surely you're not only counting Tesla drivers and their occupants.

In any case I don't know how to respond to your statement that death or serious injury likely won't be caused by crashes in the city. I'll just leave your statement as is.

 

[diplomat33]

I found this thread on twitter that dives deep into the whole "not seeing pillars" incident in one of the FSD Beta videos.

Short version: Tesla's depth maps do correctly detect the pillars as non-driveable space. The depth maps are actually really good. But it appears that FSD Beta V9 is not using the depth maps or only looks at curbs for path planning so since there were no curbs around the pillars, it did not "count" the pillars.

Here is full thread:

https://twitter.com/x/status/1415034007222317057

https://twitter.com/x/status/1415034015006949377

https://twitter.com/x/status/1415034024943259650

 

[gearchruncher]

Accidents at autosteer city speeds likely won't cause death or serious injury to occupants.

If you want to avoid death, don't drive on freeways due to the speeds involved. I use AP on freeways almost daily and it's super safe.

You really, really need to look up fatality statistics. Highways are about 3X as safe per mile as any other driving, and almost 50% of fatalities happen at below 30 MPH. You're literally actively endangering yourself by avoiding highways.

https://crashstats.nhtsa.dot.gov/api/public/viewpublication/810625

People not driving Teslas die every single day, but that's expected and common so it's ignored.

Many people have died in Teslas as well. Tesla's own crash stats only show a Tesla a bit safer than another modern car with the same features, and actual fatality stats don't really show them safer at all (1:96M vs 1:90M miles).

 

[Dan D.]

I found this thread on twitter that dives deep into the whole "not seeing pillars" incident in one of the FSD Beta videos.

Short version: Tesla's depth maps do correctly detect the pillars as non-driveable space. The depth maps are actually really good. But it appears that FSD Beta V9 is not using the depth maps or only looks at curbs for path planning so since there were no curbs around the pillars, it did not "count" the pillars.

Sorry to be dense, I'm almost with you

What is meant by "it did not 'count' the pillars"?
Is the bottom line that there is a risk that FSD Beta will hit the pillars or will not?

 

[jabloomf1230]

You really, really need to look up fatality statistics. Highways are about 3X as safe per mile as any other driving, and almost 50% of fatalities happen at below 30 MPH. You're literally actively endangering yourself by avoiding highways.

https://crashstats.nhtsa.dot.gov/api/public/viewpublication/810625

I'm having trouble finding the information you quoted at that link. Are you assuming that crash speed is the same as speed limit? Highways vs. city streets are not compared. Most of the data at that link indicates that rural driving is the most dangerous.

 

[helvio]

Sorry to be dense, I'm almost with you

What is meant by "it did not 'count' the pillars"?
Is the bottom line that there is a risk that FSD Beta will hit the pillars or will not?

I presume it's not counting the pillars as undriveable space, because they are not surrounded by curbs, only lines, and depth perception is not yet evaluating them as "do not drive" space.

 

[diplomat33]

Sorry to be dense, I'm almost with you

What is meant by "it did not 'count' the pillars"?

I am referring to this tweet:

"If you go a few seconds further until the pillars are surrounded by curbs you can see there's a purple/red line + what appears to be the gray non-drivable haze. It seems like the FSD beta can't actually detect the pillars just curb/medians"

We know the depth maps did see the pillars as non-driveable but FSD beta did not mark the pillars as non-driveable because they were not curbs/medians. That is what I mean by "did not count" the pillars. FSD Beta did not treat the pillars as non-driveable even though the depth maps did see the pillars as non-driveable.

Is the bottom line that there is a risk that FSD Beta will hit the pillars or will not?

Yes, there is a risk that FSD Beta might hit the pillars. In this FSD Beta video, we see the car steer into one of the pillars and would have hit it if the driver had not grabbed the wheel in time.

 

[gearchruncher]

I'm having trouble finding the information you quoted at that link. Are you assuming that crash speed is the same as speed limit? Highways vs. city streets are not compared. Most of the data at that link indicates that rural driving is the most dangerous.

You can interpret it via the speed vs crashes in that link.

But want it more explicit? The Irony of Road Fear - Freakonomics

For instance, in 2007 0.54 people were killed for every 100 million vehicle miles driven on urban interstates, compared with 0.92 for every 100 million vehicle miles driven on other urban highways and arterials, and 1.32 killed on local urban streets.

That's normalized for miles. Local urban streets are the most dangerous by far.

IIHS: https://www.iihs.org/topics/fatality-statistics/detail/urban-rural-comparison

Motor vehicle crash deaths by road functional class and land use, 2019
	Rural	Urban	Total*
Deaths	%	Deaths	%	Deaths	%
Interstates and freeways	2,360	14	3,773	19	6,133	17
Arterial	3,913	24	11,578	59	15,491	43
Collector	6,809	42	1,773	9	8,582	24
Local	3,218	20	2,443	12	5,661	16
Total*	16,340	100	19,595	100	36,096	100
*Total includes other and/or unknowns

 

[helvio]

You can interpret it via the speed vs crashes in that link.

But want it more explicit? The Irony of Road Fear - Freakonomics


That's normalized for miles. Local urban streets are the most dangerous by far.

IIHS: https://www.iihs.org/topics/fatality-statistics/detail/urban-rural-comparison

Deaths	%	Deaths	%	Deaths	%
	Rural	Urban	Total*
Motor vehicle crash deaths by road functional class and land use, 2019
Interstates and freeways	2,360	14	3,773	19	6,133	17
Arterial	3,913	24	11,578	59	15,491	43
Collector	6,809	42	1,773	9	8,582	24
Local	3,218	20	2,443	12	5,661	16
Total*	16,340	100	19,595	100	36,096	100
*Total includes other and/or unknowns

Your table got HTML-screwed. Do you mind posting a screenshot as well? I'm interested but my brain is going for a loop with the misaligned headers.

 

[daktari]

Comments like these tell me you don't understand the enormity of what v9 is accomplishing. They're rewritten the way the car understands the world from the ground up. That it's better out of the the gate than the previous instance tells you that it's a rock-solid foundation to build upon in the near future via machine learning.

Or be slightly worried since this might be total rewrite number 4 or so since autumn 2016. How many more total rewrites are we gonna see?

 

[mark95476]

All vehicles have minimum safety standards.

Simply put: speed kills. That shouldn't be difficult to understand.

Hit at pedestrian or other road user at 30mph vs 60 mph. It's more likely an impact at 60mph will cause death and/or severe injury.

I'm assuming you also count pedestrians and other road users. Surely you're not only counting Tesla drivers and their occupants.

In any case I don't know how to respond to your statement that death or serious injury likely won't be caused by crashes in the city. I'll just leave your statement as is.

 

[stopcrazypp]

I am referring to this tweet:

"If you go a few seconds further until the pillars are surrounded by curbs you can see there's a purple/red line + what appears to be the gray non-drivable haze. It seems like the FSD beta can't actually detect the pillars just curb/medians"

We know the depth maps did see the pillars as non-driveable but FSD beta did not mark the pillars as non-driveable because they were not curbs/medians. That is what I mean by "did not count" the pillars. FSD Beta did not treat the pillars as non-driveable even though the depth maps did see the pillars as non-driveable.



Yes, there is a risk that FSD Beta might hit the pillars. In this FSD Beta video, we see the car steer into one of the pillars and would have hit it if the driver had not grabbed the wheel in time.

Note another thing that may not be apparent to most people, but the spaces between the pillars are legal drivable space (drivers can legally cross them to get to a lane on the other side, which goes in the same direction). That's why it would have been improper to mark then as purple (unlike other continuous medians) and also why the city didn't make them as curbs in the first place.

Seattle news, weather, sports, events, entertainment | seattlepi.com

Seattle local news, traffic, weather, business news, sports, real estate, photos and events.

blog.seattlepi.com

However, the system doesn't seem to have marked in the visualization the areas under the pillars (although we have no way to tell what's happening behind the scenes, it could be the same as verygreen's sample that does have it marked as undrivable space).

 

[helvio]

Note another thing that may not be apparent to most people, but the spaces between the pillars are legal drivable space (drivers can legally cross them to get to a lane on the other side, which goes in the same direction). That's why it would have been improper to mark then as purple (unlike other continuous medians) and also why the city didn't make them as curbs in the first place.

Seattle news, weather, sports, events, entertainment | seattlepi.com

Seattle local news, traffic, weather, business news, sports, real estate, photos and events.

blog.seattlepi.com

However, the system doesn't seem to have marked in the visualization the pillars (although we have no way to tell what's happening behind the scenes).

You're right, but let me clarify something. The new stack considers drivable space even the wrong way, which is right. The cars can make use of that space to swerve away from construction stuff or stalled vehicles. The point in question here are the pillars THEMSELVES. The lack of depth perception could, in theory, make the car drive itself into one, aka, crashing lol

 

[stopcrazypp]

You're right, but let me clarify something. The new stack considers drivable space even the wrong way, which is right. The cars can make use of that space to swerve away from construction stuff or stalled vehicles. The point in question here are the pillars THEMSELVES. The lack of depth perception could, in theory, make the car drive itself into one, aka, crashing lol

But the posts above show there is depth perception that correctly identifies the pillars, just that it's unclear how Tesla is using it.

 

[helvio]

But the posts above show there is depth perception that correctly identifies the pillars, just that it's unclear how Tesla is using it.

You're right again. I meant the lack of marking the pillars themselves as obstacles. Didn't mean not detecting them. Thanks for pointing that out!
The main point was just to make sure we were talking about the obstacles, not the space between them.

 

[daktari]

All vehicles have minimum safety standards.

Simply put: speed kills. That shouldn't be difficult to understand.

Hit at pedestrian or other road user at 30mph vs 60 mph. It's more likely an impact at 60mph will cause death and/or severe injury.

Both will probably die or have severe injury at those speeds, that is a high energy trauma.
Model 3 pedestrian head and leg crash test at 1:58

 

[stopcrazypp]

Both will probably die or have severe injury at those speeds, that is a high energy trauma.
Model 3 pedestrian head and leg crash test at 1:58

I linked something in another post. Risk of death or serious injury is low under 15 mph, but goes up drastically as speed gets above that (age also plays a big factor).

Impact Speed and a Pedestrian’s Risk of Severe Injury or Death - AAA Foundation for Traffic Safety

This study examines how vehicle speed influences the probability that a pedestrian struck by a vehicle will sustain severe injuries or die.

aaafoundation.org

 

[rijc99]

As is the design that the Tesla driver intervene!

People lobbing these handfuls of of crap hoping something sticks are certainly predictable.

I guess we’re pretty far from Elon’s fleet of autonomous ride share Teslas.