Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

FSD Beta 10.69

I get the feeling that you're overestimating the abilities of simulation and/or don't understand how much goes into a simulation. (Don't take this wrong, I don't mean it disparagingly) I work in healthcare and we do simulations but even with decades of experience the simulations are still poor models of a complex biological system. This is similar - it's a highly complex and dynamic system with numerous inputs that Tesla is still developing and has relatively limited experience with. If you don't understand the complexities at this point there's not much point in continuing the conversation.
It's actually the opposite problem, as a mixed signal circuit designer I do simulations all day so I'm biased by that design flow. In circuit design you must rely heavily on simulations because the cost and time of fabricating chips is so high (months and millions of dollars). Of course there are often unexpected issues that show up in real life so you must add them to your simulation flow.
With regards to AV simulations I'm only going by industry presentations. They talk about simulation of scenarios like Chuck's ULT with millions of variations. It's hard for me to imagine that they would bother with all that if it didn't work. Nobody seems to be able explain what about Chuck's ULT is uniquely hard to simulate. It's really not a highly complex and dynamic system since if the maneuver is done correctly there is no interaction with other agents (at least in light traffic situations where we still see failures).
 

AlanSubie4Life

Efficiency Obsessed Member
Oct 22, 2018
14,945
19,293
San Diego
It does seem like I may need to get two beers from @AlanSubie4Life in the future since what we're seeing now isn't the final 10.69...
Well, it wouldn’t be double or nothing at this point. But under what terms do you want to bet for two beers? I think we need to be more rigorous in defining the terms this time. We can hash out the details offline and then present them
here.

The original bet, of course, was for the first set (or sets) of turns done by Chuck. Maybe he’ll do some more turns before 10.69.1/2 come out and then I will owe you two beers. I think I would definitely still win though; FSD Beta is oddly inconsistent. It’s very interesting; the behavior is not apparently deterministic, or at least not in an obvious way. Maybe they left some dither turned on.

Chuck really should do same more turns in heavy traffic though. It is hard to tell how FSD Beta performs on his turn.
 
Last edited:
Nowhere did he say anything was great. He essentially said you are misallocating resources and working on the wrong problem. In other words, he said that Tesla/Elon/Ashok don't know what they are doing, and they should let him prioritize their work queue.

That is a valid interpretation, but it’s kind of silly for anyone internal to the company, let alone the chief executive, to spar with what is essentially a comments section. No one wins that type of thing.
 

Mardak

Active Member
Oct 13, 2018
1,907
4,258
USA
From Elon's latest tweet it sounds like they're working on a series of known issues before releasing 10.69 wider
My interpretation of the tweet is that he's quite literal in what he said/meant:

Early beta has many known issues. The reason we release it to a limited number of cars is to discover unknown issues.​

Nothing there about working on known issues. ".0" initial releases to ~1000 vehicles is to find new issues / regressions. And even if any are found, that doesn't necessarily require any changes before releasing ".1" as now it's a known issue added to the long list of "many known issues." What could delay the next step of release is if there's a serious previously unknown issue such as the false AEB from 10.3's initial wide release to both new and existing testers.

If James' tweet was a couple weeks later and specifically said "10.69.2" instead of "10.69," would there have been a different reaction? 🤷‍♂️

So unless there's a blocking issue with 10.69.0 that is much worse than 10.12.2, I don't see why 10.69.1 would be delayed.
 

JulienW

Active Member
Jul 7, 2018
4,606
7,099
Atlanta
soooo why doesn’t it go internal then external? Why is it internal, shills, external?
Probably doesn't want another 10.3 debacle. That was a less than fun, anxiety producing update and I still remember day one (almost everyone got updates on day one) and my several scary and perplexing FCWs.:oops:

 
Last edited:
  • Like
Reactions: sleepydoc
personally don't care about Chucks VPL or whatever, really just want fewer (none?) unwanted slowdowns at 70mph from the AP stack.
Conversely want slowdowns to happen at all with the FSDb stack when the speed limit drops
:rolleyes:
any word on whether this has been fixed? I have a spot on my way to work where the speed limit drops from 50 to 35 and it takes a full quarter mile for the car to slow down. It's so gradual you can't even tell it's decelerating.
 
any word on whether this has been fixed? I have a spot on my way to work where the speed limit drops from 50 to 35 and it takes a full quarter mile for the car to slow down. It's so gradual you can't even tell it's decelerating.
In one of the 10.69 videos published so far (don't remember which one, as usual 😅), a video has confirmed that this has not been fixed. Of course, that doesn't mean 10.69.1 won't fix it, but at least on their latest published code, it's not.

It is annoying - when FSD is on, and when I see a warning sign saying it is about to go from 80 to 60, I usually end up manually reducing the max right there, so that by the time it reaches the new lower limit, I'm not at a speed-ticket range...

May be reporting this more frequently may bump up the priority of this fix? Must be a trivial fix...
 

Yelobird

Active Member
Aug 24, 2020
1,583
2,112
Illinois
I’m not sure how they may or might not even be related but since receiving the latest map data my fsd beta 10.12.2 has been driving much better. It’s choosing lanes it’s never done better; ones I would choose.
Can you share which map update version you have? Last week mine said I had an update but didn’t update and now says it’s current?
 
I'd venture that it wasn't caught because the safety performance of any given maneuver can be really subjective.

[….] You can model this scenario in a simulation and figure out how often the maneuver results in a collision; but you cannot use a simulation to model how safe a human in the car feels during the maneuver.
Not just subjective, but driven by factors beyond what are available to the car. Specifically, “how safe a human feels” can vary wildly by situation, length of delay, and confidence in judging distances.

The classic example is “driving way beyond the speed limit to get pregnant wife to the hospital”. Do we really expect Tesla to program the internal camera to watch for evidence of the passenger’s broken water and contractions and adjust how aggressively it drives—up to and including running stop signs and passing on freeway shoulders?

A less extreme example is how I’ll pull up to an unprotected left turn and wait for a few tens of seconds to see whether a nice big gap opens up in crossing traffic. But after a minute or two I’ll start looking for smaller and smaller gaps in traffic, and might even consider pulling out into the intersection. Objectively that pulling-out-into-the-intersection collision risk profile hasn’t changed in the minute or two since I stopped. But my tolerance for risk has changed.

There’s a sweeping right curve on my drive home from work that FSD tends to ride much closer to the center line than I feel comfortable about. The car is inhumanly good at knowing exactly where the side of the car is at any point in the trajectory. I ended up downloading the side camera video and playing it back a couple of times to confirm that the car wasn’t actually wandering across the center line, and was almost surprised to see how accurately it tracked and stayed in the marked lane. Taking the curve as widely as it does means less lateral force on the tire contact patches, and is therefore objectively safer. But it sure doesn’t feel safer to a human like me who routinely drives multiple cars and doesn’t have as finely calibrated a sense (e.g., to the millimeter) where exactly the car is in the lane!
 
Last edited:
Well, it wouldn’t be double or nothing at this point. But under what terms do you want to bet for two beers? I think we need to be more rigorous in defining the terms this time. We can hash out the details offline and then present them
here.

The original bet, of course, was for the first set (or sets) of turns done by Chuck. Maybe he’ll do some more turns before 10.69.1/2 come out and then I will owe you two beers. I think I would definitely still win though; FSD Beta is oddly inconsistent. It’s very interesting; the behavior is not apparently deterministic, or at least not in an obvious way. Maybe they left some dither turned on.

Chuck really should do same more turns in heavy traffic though. It is hard to tell how FSD Beta performs on his turn.
No, it would be you replacing the beer I bought you and buying me a beer! :p
I went back and looked at the terms of the bet and it's unclear if this current beta of 10.69 beta should count. It has known issues.
I was thinking and I wonder how early testers know which issues are known and which issues are unknown? Do they get special release notes that list known issues?
 
  • Funny
Reactions: AlanSubie4Life
But it sure doesn’t feel safer to a human like me who routinely drives multiple cars and doesn’t have as finely calibrated a sense (e.g., to the millimeter) where exactly the car is in the lane!

I think that as FSD Beta develops, this understanding of what makes the human driver nervous will become more and more important. This is because as the system gets better, the most dangerous thing won't be bad behavior by FSD that requires a driver intervention, it will be technically safe behavior from FSD that elicits a disengagement from a nervous driver in a dangerous way.

And I think we're already starting to see this. Take for example this article from Electrek: We tested Tesla Full Self-Driving Beta in the Blue Ridge Mountains, and it was scary where he describes a "scary disengagement." But if you frame-advance through the video, you can see that this is the point where FSD disengaged (far before it even entered the turn):

1661436274047.png


And on the previous frame, you can clearly make out his foot on the brake:

1661436335693.png


So while I cannot say with certainty that it happened this way, I think it's a possibility that he was nervous going into the turn at 23 MPH, subconsciously tapped the brakes, and disengaged FSD himself.
 
  • Like
Reactions: FSDtester#1
It's actually the opposite problem, as a mixed signal circuit designer I do simulations all day so I'm biased by that design flow. In circuit design you must rely heavily on simulations because the cost and time of fabricating chips is so high (months and millions of dollars). Of course there are often unexpected issues that show up in real life so you must add them to your simulation flow.
With regards to AV simulations I'm only going by industry presentations. They talk about simulation of scenarios like Chuck's ULT with millions of variations. It's hard for me to imagine that they would bother with all that if it didn't work. Nobody seems to be able explain what about Chuck's ULT is uniquely hard to simulate. It's really not a highly complex and dynamic system since if the maneuver is done correctly there is no interaction with other agents (at least in light traffic situations where we still see failures).
I got my bachelors (and spend 2 years working on a masters) in electrical engineering. That was several years ago (do they still use SPICE?) but I've actually run circuit simulations so I have an idea what you're talking about. The simulations you're using are quite different from those for biologic systems for a variety of reasons. I consider the FSD simulations to be much more akin to biologic systems just based on the number and degree of variables as well as what you can control vs not control.
 
I got my bachelors (and spend 2 years working on a masters) in electrical engineering. That was several years ago (do they still use SPICE?) but I've actually run circuit simulations so I have an idea what you're talking about. The simulations you're using are quite different from those for biologic systems for a variety of reasons. I consider the FSD simulations to be much more akin to biologic systems just based on the number and degree of variables as well as what you can control vs not control.
The tools still use spice format for netlists but there are have been much faster simulators for decades. One nice thing about FSD simulations is that they can run in realtime (or way faster if you're not including the perception stack) whereas circuit simulations are millions to billions of times slower than real life.

I'm agreeing in general but not for this case. I think what makes FSD simulations difficult is the modeling of actions and reactions of other drivers but that is not the case in Chuck's ULT.
 
Last edited:
  • Like
Reactions: sleepydoc

About Us

Formed in 2006, Tesla Motors Club (TMC) was the first independent online Tesla community. Today it remains the largest and most dynamic community of Tesla enthusiasts. Learn more.

Do you value your experience at TMC? Consider becoming a Supporting Member of Tesla Motors Club. As a thank you for your contribution, you'll get nearly no ads in the Community and Groups sections. Additional perks are available depending on the level of contribution. Please visit the Account Upgrades page for more details.


SUPPORT TMC
Top