Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
Unnecessary wear on brakes? You bet.
Think about the energy dissipated. It’s basically zero and the brakes contact and sweep the disk for a completely minimal portion of the rotor circumference. The additional wear on the brakes is totally minimal. What you have to “worry” about is the car touching the brakes at 20-40mph. That will create enormous wear.
making that full zero MPH stop right on the stop line for a fraction of a second
It is surprising how much you really have to press the brake pedal to get a solid stop. If done correctly it can be fairly smooth but it definitely highlights one issue Tesla will eventually need to deal with.

Fortunately for them there is a ton of low-hanging fruit they can deal with around stops before having to fret over that. Like slamming brakes early, eliminating regen and coasting unnecessarily, etc. We all know the horrors of an FSD stop at this point and have experienced the several seconds of lost time at each stop.

The one thing that is clear is that NHTSA requirements have nothing to do with it! It does provide an excuse though for Tesla which they can foist on those who haven’t thought about it though, I suppose.
 
Last edited:
You might get a ticket for not coming to a complete stop, but I've never heard of one for not stopping before the line.
An example with the case details is posted up thread. It happens. You need to stop at the stop line. If the cop doesn’t see you stop there, you’ve potentially got problems.

Not really an issue for FSD most likely since it usually stops about five times (yes this is potentially an exaggeration) but stopping at the actual line might be missed I suppose.
 
  • Like
Reactions: cyborgLIS
I think it was a lot more cautious than v12.3.x! In 12.3.x it would stop in the middle of traffic lanes with oncoming traffic, assuming that the traffic would stop. In 12.4.1 it appears to want to clear the traffic lanes.

I'd have to review the details on the gap tolerance of v12.3.x (perhaps it was very conservative - here's an example showing an aggressive commit from 12.3.x), but really allowing 4-5 seconds of space as 12.4.1 does is ample. Just have to cross quickly (this was a problem with 12.3.x). It's not really throwing caution to the wind if executed and committed properly. You have to cross the road somehow.

I agree it needs to clear the aft part of the vehicle from traffic lanes with greater alacrity.

Basically just a quick review of 12.3.x showed major issues:
1) Insufficient crossing speed.
2) Stopping in near traffic lanes due to traffic from the right.
3) Bad pose in the median (sticking out too far in the front sometimes, slow to clear the rear from traffic lanes).
4) Aggressive pre-rolling?
With all the dissection's of Chuck's UPL the only thing that mattered to me was Chuck's statement that v12.4.1 is much better than v12.3.6.
The rest is just noise.
 
An example with the case details is posted up thread. It happens. You need to stop at the stop line. If the cop doesn’t see you stop there, you’ve potentially got problems.

Not really an issue for FSD most likely since it usually stops about five times but stopping at the actual line might be missed I suppose.
You got me there. Still, there's stopping at the line and there's stopping 5 feet before it.
 
With all the dissection's of Chuck's UPL the only thing that mattered to me was Chuck's statement that v12.4.1 is much better than v12.3.6.
The rest is just noise.
To be fair, it would be hard to be worse than the last outing. Still, not at the first 9. That is another thing that matters. It does not appear they were holding back until they had the first two 9s, as some have surmised.

Still, there's stopping at the line and there's stopping 5 feet before it.
Yes, it it’s important to stop at the line. If you stop before the line you may get a ticket if you do not also stop at the line. The law says to stop at the line. Five feet before it doesn’t count, of course. That was the specific case referenced above - the defendant stopped before the line. This was deemed to not be legal (by the appeals court as I recall). Plenty of discussion up thread.
 
Yes, it it’s important to stop at the line. If you stop before the line you may get a ticket if you do not also stop at the line. The law says to stop at the line. Five feet before it doesn’t count, of course. That was the specific case referenced above - the defendant stopped before the line. This was deemed to not be legal (by the appeals court as I recall). Plenty of discussion up thread.
How far before the line did that person stop?
 
How far before the line did that person stop?
10 feet or so, by that person's admission. At the position of the stop sign (in line with it), and the "STOP" lettering.

This is within the range of outcomes possible with FSD - in fact we have pictures showing it stopping in exactly such a spot, just at/over the "STOP" letters, as you are no doubt familiar with.

Definitely needs to be fixed. I don't think there is any debate about that! There's just no good reason for the current behavior, and most certainly has nothing to do with the NHTSA. Make it a reasonable couple feet behind the line, sometimes closer (it's context dependent), and call it good. Be respectful to pedestrians & other road users and stop where you're supposed to (just listen to the court's opinion; they have very excellent legal arguments explaining where one should stop). It's not that hard to do in general. There will always be tricky special cases so may as well get the easy ones right.

I think stopping very far back (which still happens with 12.3.6!) is neither legal nor appropriate.
 
Last edited:
The thing is, there is actually zero truly useful information in the driving visualization. It’s all eye candy and in recent versions it actually includes less information than it used to (e.g. cones and construction barrels are no longer shown.)
The map shows your route, traffic and other data you can actually use. Even if they are designing it for a robo taxi as you surmise, shouldn’t they get the Software that actually drives the car done before they worry about graphics?
At least until the update, I found the visualization enhanced my awareness of my surroundings. Often I would notice cars alongside or behind me in the visualization before noticing them in the real world. It also helped with gauging distance for lane changes (especially in slow moving traffic) and checking my blind spot without signaling for the repeater cam. Mirror, repeater cam, and 3rd person visualization each have their own strengths and weaknesses. It helps to make use of all three. The one thing that should be a thing of the past in a tech car is turning your head away from the road to check your blind spot. We know the visualization was designed for blind spot monitoring because it zooms out for lane changes.

Disappointingly, after the update it does not sit zoomed out as far as it used to. I'm really hoping they revert this. Even better would be if they allowed you to fix the camera in place to your preference.
 
High resolution mode meaning full screen FSD? Or is there something else?
Meaning visualization showing both lanes of a two lane country road with the double/dotted yellow line down the middle and the white fog lines on either side. Garbage cans and traffic cones and all the other details you don't see while using AP or manual. At least twice the level of resolution. And this is with an Intel chip.

But just in the regular map window.
 
  • Like
Reactions: zoomer0056
Yes, it it’s important to stop at the line. If you stop before the line you may get a ticket if you do not also stop at the line. The law says to stop at the line. Five feet before it doesn’t count, of course. That was the specific case referenced above - the defendant stopped before the line. This was deemed to not be legal (by the appeals court as I recall). Plenty of discussion up thread.
I actually thought FSD was following the law by stopping a few feet before the line, I guess not. The way I learned to drive is to creep up to the point of visibility, then come to complete stop there, not necessarily at the stop line itself.
 
  • Like
Reactions: aronth5 and JHCCAZ
Meaning visualization showing both lanes of a two lane country road with the double/dotted yellow line down the middle and the white fog lines on either side. Garbage cans and traffic cones and all the other details you don't see while using AP or manual. At least twice the level of resolution. And this is with an Intel chip.

But just in the regular map window.
Still a little confused. This was in the map window only? What was in the left pane? Also not sure what you mean by manual mode. AFAIK the visualization doesn't change based on whether or not AP/FSD is engaged.

EDIT: Or are you saying you had AP engaged but somehow got the FSD visualization instead?
 
At least until the update, I found the visualization enhanced my awareness of my surroundings. Often I would notice cars alongside or behind me in the visualization before noticing them in the real world. It also helped with gauging distance for lane changes (especially in slow moving traffic) and checking my blind spot without signaling for the repeater cam. Mirror, repeater cam, and 3rd person visualization each have their own strengths and weaknesses. It helps to make use of all three. The one thing that should be a thing of the past in a tech car is turning your head away from the road to check your blind spot. We know the visualization was designed for blind spot monitoring because it zooms out for lane changes.

Disappointingly, after the update it does not sit zoomed out as far as it used to. I'm really hoping they revert this. Even better would be if they allowed you to fix the camera in place to your preference.
Not to be insulting, but part of paying attention while driving is scanning your mirrors and knowing where traffic is around you.

As far as the visualizations go I’ve found that the placement of cars on the screen is fairly inaccurate and you can’t depend on it to know much more than the fact that there’s a car in that general direction. They are certainly not accurate enough to gauge a safe merging distance.

Interestingly, I’ve had FSD almost cut off cars when merging on the highway a couple of times. Maybe FSD needs to quit looking at the screen, too.
 
They are certainly not accurate enough to gauge a safe merging distance.

I've never seen it be so off as to be a risk of collision during a merge. But I also try to avoid merging into super tight gaps too, and avoid using only the visualization because it doesn't show rearward far enough for my taste.

The way I learned to drive is to creep up to the point of visibility, then come to complete stop there, not necessarily at the stop line itself.

That's how most prudent drivers behave (watching for pedestrians who naturally would be closer).
 
Not to be insulting, but part of paying attention while driving is scanning your mirrors and knowing where traffic is around you.

As far as the visualizations go I’ve found that the placement of cars on the screen is fairly inaccurate and you can’t depend on it to know much more than the fact that there’s a car in that general direction. They are certainly not accurate enough to gauge a safe merging distance.

Interestingly, I’ve had FSD almost cut off cars when merging on the highway a couple of times. Maybe FSD needs to quit looking at the screen, too.
I regularly scan my mirrors as I'm sure you do, and yet vehicles end up in our blind spots. There's a reason why we check our blind spots before performing a lane change.

If the visualization helps you identify that a car is in a given direction, how does it offer zero useful information? The mirror is more reliable for gauging distance (except possibly the right mirror which is distorted), but the visualization works if you are not aiming for a crack of clearance.

The mirror has a blind spot and can make it difficult to identify which lane a visible car is in. The camera is poor for gauging distance and has the same issue of discerning a vehicle's lane. The visualization is better for gauging distance than the camera and placing vehicles in the correct lane. Each tool has its strengths and weaknesses so you can't rely on just one.
 
  • Like
Reactions: JHCCAZ and rlsd
I regularly scan my mirrors as I'm sure you do, and yet vehicles end up in our blind spots. There's a reason why we check our blind spots before performing a lane change.

If the visualization helps you identify that a car is in a given direction, how does it offer zero useful information? The mirror is more reliable for gauging distance (except possibly the right mirror which is distorted), but the visualization works if you are not aiming for a crack of clearance.

The mirror has a blind spot and can make it difficult to identify which lane a visible car is in. The camera is poor for gauging distance and has the same issue of discerning a vehicle's lane. The visualization is better for gauging distance than the camera and placing vehicles in the correct lane. Each tool has its strengths and weaknesses so you can't rely on just one.
Yes, yes, yes. I use the visualization screen to monitor (supervise) lane changing. I wish I can see at least 3 cars behind my ego. Right now I can see 2 or 2.5.
 
OT: Here is a great video about the way AI works, its limitations and future iterations of it. Even shows some Tesla driving to stay OT.

Please check me on this. This video appears to confirm what I said about the immense calculation needed to train current AI models, such as FSD. The video then speculates on future, as yet unproven ideas about "liquid" and "spiking" models, which, it suggests might be able to learn after they are trained. Promising, yes, but FSD does not incorporate either of these hypothetical technologies. And the video does not give any clue as to how the liquid model incorporates any experience after training into the parameters of the "core" of the "liquid" model called the "reservoir". I'm not calling BS, but this video sounds a lot like pseudo-science AI. That is to say there may be some reality behind this, but this video does not illuminate.

Did I miss something?

I do hope that some sort of learning in the Tesla cars is eventually implemented, beyond "Home" and "Work" and other favorite destinations and manual settings. Things like how much acceleration after a stop is comfortable for this driver, where the pot holes in my neighborhood are so I don't have to disengage every time I get near home, and which turns should be taken slower so as not to veer over the lane divider line.

So far, Telsa's approach is one style, one set of learning fits all cars. This has never been the reality, where each driver has a different style as well as a different family, pets and friends as critics of their driving style. Safe is necessary, but comfortable for each family is the metric critical for market acceptance, which metric is different for each family. Learning in each car may be necessary, in which case, we certainly still have quite a way to go.
 
Please check me on this. This video appears to confirm what I said about the immense calculation needed to train current AI models, such as FSD. The video then speculates on future, as yet unproven ideas about "liquid" and "spiking" models, which, it suggests might be able to learn after they are trained. Promising, yes, but FSD does not incorporate either of these hypothetical technologies. And the video does not give any clue as to how the liquid model incorporates any experience after training into the parameters of the "core" of the "liquid" model called the "reservoir". I'm not calling BS, but this video sounds a lot like pseudo-science AI. That is to say there may be some reality behind this, but this video does not illuminate.

Did I miss something?

I do hope that some sort of learning in the Tesla cars is eventually implemented, beyond "Home" and "Work" and other favorite destinations and manual settings. Things like how much acceleration after a stop is comfortable for this driver, where the pot holes in my neighborhood are so I don't have to disengage every time I get near home, and which turns should be taken slower so as not to veer over the lane divider line.

So far, Telsa's approach is one style, one set of learning fits all cars. This has never been the reality, where each driver has a different style as well as a different family, pets and friends as critics of their driving style. Safe is necessary, but comfortable for each family is the metric critical for market acceptance, which metric is different for each family. Learning in each car may be necessary, in which case, we certainly still have quite a way to go.
Do you backseat drive your Uber drivers. Or just let them get you to your destination?

Edit: not saying we don’t need improvement, just that FSD doesn’t need to learn every persons individual driving style. Chill mode, and the three options for following distance (forget the other two because I always use assertive) should suffice. Saying that market acceptance is tied to to learning everybodies driving style would mean Uber doesn’t exist.
 
Last edited:
part of paying attention while driving is scanning your mirrors and knowing where traffic is around you.
I was taught to look over my shoulder too. It is getting less feasible as I age and as freeway madness becomes more common. Yet another place to look is not an improvement. I'd prefer a loud beeping when I turn on the blinker or start a lane change when someone is lurking where my mirrors and swivel head don't cover, or someone is rabbit lane changing behind me.