FSD Beta 10.69

[EVNow]

I’m glad someone does! It’s a real pain in my can and I’ve almost given up on it a few times.

Check this out. There is a crowdsourced tracker that tracks 10,000s of miles of FSDb driving by some 100 testers.

Here is the latest crowd sourced FSDTracker.

It is interesting how the miles/disengagement hasn't changed much over the months.

Home

FSD Community Tracker

www.teslafsdtracker.com

 

[mtndrew1]

Check this out. There is a crowdsourced tracker that tracks 10,000s of miles of FSDb driving by some 100 testers.

Here is the latest crowd sourced FSDTracker.

It is interesting how the miles/disengagement hasn't changed much over the months.

View attachment 902566

Home

FSD Community Tracker

www.teslafsdtracker.com

I wonder if some of this is self selecting bias. I think anyone rational would have tried my test route once and just turned off FSDb entirely. It’s hilariously awful.

FSDb has never come close to approaching even the illusion of usefulness in my car. It is often terrifying. My husband doesn’t like me engaging it because it’s so erratic.

I can’t imagine I’m alone. I have three friends with FSDb (we bought it back in the day before it was obvious this wasn’t going anywhere) and all three turned it on once, were terrified, and never tried it again.

I think we’d all love to see Tesla’s internal data on this. Out of 400k cars I bet 350k of them have tried FSDb three or fewer times and now just ignore it. At least in my case it’s hand-clappingly remarkable if the car can go from one destination to another without an intervention to avoid a collision or a honk from another motorist.

 

[Dewg]

I wonder if some of this is self selecting bias. I think anyone rational would have tried my test route once and just turned off FSDb entirely. It’s hilariously awful.

FSDb has never come close to approaching even the illusion of usefulness in my car. It is often terrifying. My husband doesn’t like me engaging it because it’s so erratic.

I can’t imagine I’m alone. I have three friends with FSDb (we bought it back in the day before it was obvious this wasn’t going anywhere) and all three turned it on once, were terrified, and never tried it again.

I think we’d all love to see Tesla’s internal data on this. Out of 400k cars I bet 350k of them have tried FSDb three or fewer times and now just ignore it. At least in my case it’s hand-clappingly remarkable if the car can go from one destination to another without an intervention to avoid a collision or a honk from another motorist.

Weird... I'm in the Irvine area and have a good experience. The freeway widening project on I-405 causes some problems around on ramps, but I just disable around them.

Can you give me an example route (start and destination)? I'd love to test next time I'm in the area for Brodard spring rolls.

 

[mtndrew1]

Weird... I'm in the Irvine area and have a good experience. The freeway widening project on I-405 causes some problems around on ramps, but I just disable around them.

Can you give me an example route (start and destination)? I'd love to test next time I'm in the area for Brodard spring rolls.

My test route is here https://goo.gl/maps/RLTCvYDJUjS5AGnP7

 

[KArnold]

Out of 400k cars I bet 350k of them have tried FSDb three or fewer times and now just ignore it.

Plus I know 2 folks who bought FSD but don't want to try it until it's "out of beta". Tried to tell them they may wait a while, as taking something out of beta is unprecedented for Tesla. But that number is not insignificant.

Then maybe one of them gets brave and tries it. I'm sure most then go into @mtndrew1 s group. I'd love to see those stats.

 

[JohnB007]

About mapping issues - how does Tesla manage it's maps? I sent mapping error long ago on a near-by street to TomTom (they accepted) and that has not been received by my car, while my home location (I'm in a new subdivision) was submitted to GOOGLE MAPS only 2 weeks ago (they accepted yesterday), and all of a sudden, it is correct on my car, without any updates...

Confused ???

 

[MP3Mike]

About mapping issues - how does Tesla manage it's maps? I sent mapping error long ago on a near-by street to TomTom (they accepted) and that has not been received by my car, while my home location (I'm in a new subdivision) was submitted to GOOGLE MAPS only 2 weeks ago (they accepted yesterday), and all of a sudden, it is correct on my car, without any updates...

Confused ???

My understanding is that there are three sets of maps, only one of which is visible to you:

• Google maps: What is displayed on the screen.
  • These are live from Google.
• Navigation maps: Maps that the Tesla navigation systems uses to plan your route. Your route is displayed on top of the Google maps while you are navigating.
  • These traditionally have been updated once or twice a year. (Big download, with the version you have listed on the Software page in the car.)
• ADAS maps: Maps/meta-data used by the AP/FSD system on the FSD computer. Not visible to us end-users.
  • I haven't heard anything about the update schedule for these.

Part of the ADAS maps seem to come from OpenStreets, like what is used for Smart Summon in parking lots.

 

[sleepydoc]

Check this out. There is a crowdsourced tracker that tracks 10,000s of miles of FSDb driving by some 100 testers.

Here is the latest crowd sourced FSDTracker.

It is interesting how the miles/disengagement hasn't changed much over the months.

View attachment 902566

Home

FSD Community Tracker

www.teslafsdtracker.com

So the FSD team has made no progress in the last year. Not encouraging.

 

[mtndrew1]

So the FSD team has made no progress in the last year. Not encouraging.

From my driver’s seat there have been huge improvements in many parts of the system. Effectively none of those improvements have resulted in FSDb becoming more useful, however.

I think this trips up a lot of people tracking FSD progress. I’ve been in the program since September of 2021 and the perception and rendering and so forth are quite different now. Way more sophisticated. It’s easy to be amazed by what’s on the screen only to realize that the actual system performance (point A to point B) remains a novelty, at best. In that regard there’s been little progress over the last year and a half.

This thing needs to be able to go tens of thousands of miles between disengagements before it results in “the greatest asset value increase in history” as one chief twit promised.

It’s like smart summon. It’s…neat.

 

[FOOLSLFDRVNG]

They took away speed bump slow down and rendering. It just pummels through now, feels like 4X4 big foot and the muscle machine monster car.

 

[EVNow]

So the FSD team has made no progress in the last year. Not encouraging.

Personally my disengagement rate has gone down. But apparently not for the people submitting to that crowdsourced dashboard.

 

[Tronguy]

Weird... I'm in the Irvine area and have a good experience. The freeway widening project on I-405 causes some problems around on ramps, but I just disable around them.

Can you give me an example route (start and destination)? I'd love to test next time I'm in the area for Brodard spring rolls.

And, THIS.

On this thread I read end-of-the-world-posts about how bad FSD-b is, how it can't go a mile without endless interventions, and how it's Not Ready for Prime Time.

I'll be the first to admit that running red lights has happened to me, a couple months back in particular, and now and then.

Phantom braking, of the, "it slows down by sure doesn't screech the wheels" type does happen. Panic stops? Nope.

Getting confused sometimes about which lane it should be in? Yep.

But, having said all that: Compared to, say, last March, the car drives tons better. Back then, it was harum-scarum interventions per mile; a 15 mile trip on local road involved 20+ interventions, all for good reason.

Now.. A 15 mile trip is at most two, maybe three interventions. With the occasional it-made-it-all-the-way trip. Those "did it all the way" trips did occur last March, but they were very, very rare. Now it's closer to 50:50.

Yeah, it's still not ready for prime time. But it's not the disaster it used to be, at least with my winging around in central NJ, one of the most congested areas in the U.S..

Heck, we even get the occasional New Yorker from Manhattan who posts, in amazement, that he managed to get significant distances downtown.

So, in the glass half-full/half-empty sweeps, I'm leaning towards 75%-85% full. Why have we got 10%'s around here? Something weird about the maps, the hardware, or what?

 

[mtndrew1]

They took away speed bump slow down and rendering. It just pummels through now, feels like 4X4 big foot and the muscle machine monster car.

Mine no longer renders speed bumps but it still slows for them.

 

[Supcom]

Personally my disengagement rate has gone down. But apparently not for the people submitting to that crowdsourced dashboard.

As are mine. I regularly have drives with zero, or one, disengagement. Most of these are a mix between FSDb and NOA and most disengagements are at a few common locations where the car makes the same mistakes. Few of these are safety related, mostly poor lane selection or inability to make multiple lane changes in a short distance.

I'll often have a few interventions when I'm impatient with the car being slow to make a right turn on red or similar hesitation.

Perhaps those who contribute to a crowdsourced dashboard are more motivated by poor behavior that good? One should be skeptical of any conclusions if the collection methods do not provide controls for bias.

 

[AlanSubie4Life]

So the FSD team has made no progress in the last year. Not encouraging.

If the design intent were to have it be a useful system, I would be discouraged.

But all the signs are that that is not the design intent.

When it can stop and go reliably and smoothly and promptly, I am going to really start to take interest, since that will suggest a change in design intent, and things could get interesting.

 

[Daniel in SD]

If the design intent were to have it be a useful system, I would be discouraged.

But all the signs are that that is not the design intent.

When it can stop and go reliably and smoothly and promptly, I am going to really start to take interest, since that will suggest a change in design intent, and things could get interesting.

Things aren’t that encouraging if you think the design intent is robotaxis either. The only encouraging news on that front is that they’re not trying to make HW4 compatible with older cars.

 

[AlanSubie4Life]

Things aren’t that encouraging if you think the design intent is robotaxis either. The only encouraging news on that front is that they’re not trying to make HW4 compatible with older cars.

I don’t think that robotaxis are the design intent either.

That’s a classic misdirection ploy, while Tesla focuses on solving a much harder problem.

 

[valem]

I wonder if some of this is self selecting bias. I think anyone rational would have tried my test route once and just turned off FSDb entirely. It’s hilariously awful.

FSDb has never come close to approaching even the illusion of usefulness in my car. It is often terrifying. My husband doesn’t like me engaging it because it’s so erratic.

I can’t imagine I’m alone. I have three friends with FSDb (we bought it back in the day before it was obvious this wasn’t going anywhere) and all three turned it on once, were terrified, and never tried it again.

I think we’d all love to see Tesla’s internal data on this. Out of 400k cars I bet 350k of them have tried FSDb three or fewer times and now just ignore it. At least in my case it’s hand-clappingly remarkable if the car can go from one destination to another without an intervention to avoid a collision or a honk from another motorist.

I had family came to visit from overseas, and I was excited to show them FSDb, and the progress in autonomous driving... well, even when there was no need for interventions or disengagements, they just kindly asked me to stop using it and just drive the car.
Their words: "Can you please drive and not let the car drive anymore when we are in the car?"

I don't blame them; it has become nearly unusable (very uncomfortable) since the wide release.
Every time it sees a scary shadow it phantom brakes.
Weird lane changes, slams on the brake to get behind a car (nearly causing rear ends) when it could get in front instead.
Continues speeding towards stopped cars at red lights instead of coasting etc.

I don't understand how the cyberlyft passengers can't tell he is not driving but the car is... most people I know would be like, hey dude, get me out of this.

 

[AlanSubie4Life]

they just kindly asked me to just stop using it and just drive the car

Same reaction for my tech-savvy out-of-town guests, one of whom has greatly enjoyed taking passenger rides on an autocross course in the past.

 

[Daniel in SD]

I don’t think that robotaxis are the design intent either.

That’s a classic misdirection ploy, while Tesla focuses on solving a much harder problem.

Solving a harder problem (I assume you mean driver assist that people will pay $15k+ for) instead of solving a problem that will make FSD worth $100k. Makes sense.