After 5-weeks, I turned FSD Beta off!

After years of waiting I finally got my FSD Beta! After 6-weeks I turned it off.

I found the promise of the FSD Beta to be overblown. Yes it is very cool that the car will navigate on city streets to a destination. But the car is far too timid to be usable. It is worse than driving with a student driver. Also, the car’s hesitation causes a lot of problems with other drivers who get confused when they see the car stop or slow, when it should not. But that is not what made me turn it off.

I turned it off because it made the overall driving experience annoying, and more dangerous.

With the beta turned-on, the car makes speed changes at nearly every speed limit sign, even when only running TACC. Not only is this annoying but these changers often alarm my passengers and sometimes surprise the cars behind. The car does not do this when the FSD is disabled.

And then there is a notice at nearly every traffic light that one is coming, even when running only on TACC. I find I am spending time clearing nuisance alerts instead focusing 100% of my time on driving the car. And the sudden speed changes, if you miss one of the alerts, is just another version of the phantom breaking problem with the same set of issues for the cars behind. Note that I have this feature disabled under AP. This does not occur if FSD is disabled.

Next we have speed based lane changes. I disabled this under AP and the FSD stack just ignores this as well, and there is no way to tell it not to do so.

So overall, the benefit of having FSD is very minimal, while the negative impact to quality and enjoyment of driving is very real. I definably will not be buying FSD on a future car, unless these problems are fixed.

"Tesla Autopilot Engaged in Model X" by Ian Maddox is licensed under CC BY-SA 4.0.
Admin note: Image added for Blog Feed thumbnail

 

[ewoodrick]

You are just being argumentative. This is not about how we as humans drive, it is about automated driving systems. The point is that while vision can do a lot vision, has limitations. It is foolish to rely entirely on vision when you can improve on it. Consider a foggy road - your eyes can see say 50 feet, radar can see hundreds of feet.

No, I'm only stating reality.
A foggy road? that's funny, why does a car need to see more than a human does?

My car easily drove 40 miles without LIDAR or RADAR. How did it do it?

Again, I'm the one stating fact, you seem to not believe reality.

 

[johnnycnote]

Elon is a visionary. Unfortunately, he lacks common sense. Removal of radar and USS are primary examples.



Elon‘s vision is a car that uses 100% vision to navigate. Adding radar back will have nothing to do with cost effectiveness, it will appear when he is forced to realize a 100% vision approach will not let him reach the higher levels of FSD. This is why we are about to see radar reappear in Tesla cars.

Tesla will use different technologies as those technologies become more cost effective. LIDAR, as I understand it, is very expensive to implement, and will require vision to go along with it. Can they achieve full FSD with vision only? I have no idea, but it's arguably doing better than anything out there for a fraction of the cost

 

[BrendonHolt]

No, I'm only stating reality.
A foggy road? that's funny, why does a car need to see more than a human does?

My car easily drove 40 miles without LIDAR or RADAR. How did it do it?

Again, I'm the one stating fact, you seem to not believe reality

I find your positions rather aggressive and frankly repetitive, I will bail now. You seem to state reality, reality is FSD AIN'T THERE and thanks for playing, but really dude, if we don't agree with you does not mean you are right.

 

[BrendonHolt]

No, I'm only stating reality.
A foggy road? that's funny, why does a car need to see more than a human does?

My car easily drove 40 miles without LIDAR or RADAR. How did it do it?

Again, I'm the one stating fact, you seem to not believe reality.

Reality, my wife had Tesla with FSD, frankly, reality is that you have many more posts then me, hopefully, you are not so much smarter then everyone you post against, since obviously this is not a discourse of wit.

 

[BrendonHolt]

It blows my mind that anybody thinks it's acceptable for Tesla to arbitrarily remove functionality from our cars, temporarily or not.

Absolutely agree

 

[BrendonHolt]

No, I'm only stating reality.
A foggy road? that's funny, why does a car need to see more than a human does?

My car easily drove 40 miles without LIDAR or RADAR. How did it do it?

Again, I'm the one stating fact, you seem to not believe reality.

One last point, the goal is SELF DRIVING, not HUMAN ASSISTED DRIVING. With that goal in mind, yes the car has to see more then the human does, it has to be BETTER then the Human. So yes, the Car needs to have Technology to make it see things that the human cannot, and calculate actions of other vehicles BETTER then the human.

Your position that Vision Only is acceptable lacks any concept on what the entire Autonomous Driving Industry is striving for.
Here is where the industry is going in Commercial Trucking, and note the picture with LIDAR, RADAR, CAMERA, GPS and Speed and Gyroscopes to manage vehicles.

SO YES THE BAR IS FAR HIGHER THEN WHAT HUMANS CAN SEE AND DO, and MUSK will figure that out as his technology is passed up.

Research Guides: The Trucking Industry: A Research Guide: Autonomous Trucking

This guide provides useful starting points, including books, journals, databases, legislation, government agencies, and associations, for in-depth research on the U.S. trucking industry, with a strong focus on autonomous trucking.

guides.loc.gov

 

[BrendonHolt]

 

[BrendonHolt]

I have no idea, but it's arguably doing better than anything out there for a fraction of the cost

This is a true statement. They have a lead, but not significant lead, commercial vehicles are getting ahead of the curve.

Really cool how it sees the traffic signs, and light colors. Obviously Vision is critical technology, and Musk is correct it can be used for a significant advancement in Self Driving. My only argument is Full Self Driving will require more then one technology.

It is a widely held Industry Belief, but if anyone can do it Tesla/Elon can, but not so sure he will get this one right. Vision only, not buying it till it is proven.

 

[SalisburySam]

…never mind your customers with HW3 trying to park." and have had few issues doing so with previously-installed sensors.

FTFY.

 

[BrendonHolt]

…never mind your customers with HW3 trying to park." and have had few issues doing so with previously-installed sensors.

FTFY.

Here Here, but we talk with our wallets. Went into the Fort Lauderdale Tesla Dealership to look at 2023. I total my wife Tesla 3 LR RWD and decided on 2021 with 18K miles, with USS and Radar. Many would argue that for the price difference we should get new. Used had Enhanced Autopilot and Sensors for $48K. New it was like 60K when you add Enhanced Autopilot.

Money was not the issue, new had no USS, and my wife LOVES knowing where Columns in our parking garage and backing into park LOVES the distance. For me not having Radar at all, mean the 2023 cannot implement Radar it does not have. Bad Radar to me is better then no Radar at all.
She is already over that new car smell, I detailed it so it smells just fine and looks like new.

The argument was made that "No Problem" they are bringing it back. For 12K less got a car with all the features she wants and none of them removed.

Hopefully someone in R&D at Tesla reads these posts, because USS is pretty standard in high end and even middle end vehicles.

 

[ewoodrick]

One last point, the goal is SELF DRIVING, not HUMAN ASSISTED DRIVING. With that goal in mind, yes the car has to see more then the human does, it has to be BETTER then the Human. So yes, the Car needs to have Technology to make it see things that the human cannot, and calculate actions of other vehicles BETTER then the human.

Your position that Vision Only is acceptable lacks any concept on what the entire Autonomous Driving Industry is striving for.
Here is where the industry is going in Commercial Trucking, and note the picture with LIDAR, RADAR, CAMERA, GPS and Speed and Gyroscopes to manage vehicles.

SO YES THE BAR IS FAR HIGHER THEN WHAT HUMANS CAN SEE AND DO, and MUSK will figure that out as his technology is passed up.

Research Guides: The Trucking Industry: A Research Guide: Autonomous Trucking

This guide provides useful starting points, including books, journals, databases, legislation, government agencies, and associations, for in-depth research on the U.S. trucking industry, with a strong focus on autonomous trucking.

guides.loc.gov

You seem to be stuck on self driving having to be better than humans. That's not the general goal as far as I know.

The general goal is that self-driving does as good as humans, at least until you can get humans off the road. At that point, full autonomy is easy.

And look at the current FSD specs for Tesla, it's currently significantly safer than the alternatives. FSD is currently reacting to over 90% better and faster than humans do.

NO THE BAR IS NOT HIGHER THAN HUMANS. I have no idea where that came from.


By the way, you may want to read the article that you linked. I've highlighted a few sections.


Trucking is the dominant mode of US inland freight transport; it amounts to billions of tons and is projected to continue to climb in the coming years.1

The term autonomous trucks is applied to trucks that will be controlled from other sources such as satellites and advanced GPS (Global Positioning Systems), models currently on the road already provide a semi-autonomous mode of operation, in which the unmanned system and/or a human operator conduct a mission, have various levels of human-robot interaction. In the fully autonomous mode of operation, the unmanned system is expected to accomplish its mission, within a defined scope, without human intervention. In the teleoperation mode of operation, the human operator, using video feedback and/or other sensory feedback, either directly controls the actuators or assigns incremental goals, waypoints in mobility situations, on a continuous basis, from off the vehicle and via a tethered or radio linked control device. And, finally, in the remote control mode of operation, the human operator, without benefit of video or other sensory feedback, directly controls the actuators of the unmanned system on a continuous basis, from a location off the vehicle and via a tethered radio linked control device using visual line-of-sight cues.

By 2027 fully autonomous trucks, including truck platoons of two or more trucks in which all trucks have a driver, but only the driver of the lead truck has full control of the vehicle, are anticipated to appear on highways.3 In spite of this transition, it is anticipated that the delivery of goods and services will be much the same as it has been for decades in the sense that drivers remain critical to the process.

 

[ATPMSD]

NO THE BAR IS NOT HIGHER THAN HUMANS. I have no idea where that came from.

This came straight from Elon:

”So he wants the system to be ten times safer than a human driver”

Tesla News and Reviews | InsideEVs

Get the latest on new Tesla models including spy shots, first drives, in-depth reviews, concept car photos, new features and more.

insideevs.com

 

[BrendonHolt]

You seem to be stuck on self driving having to be better than humans. That's not the general goal as far as I know.

The general goal is that self-driving does as good as humans, at least until you can get humans off the road. At that point, full autonomy is easy.

And look at the current FSD specs for Tesla, it's currently significantly safer than the alternatives. FSD is currently reacting to over 90% better and faster than humans do.

NO THE BAR IS NOT HIGHER THAN HUMANS. I have no idea where that came from.


By the way, you may want to read the article that you linked. I've highlighted a few sections.


Trucking is the dominant mode of US inland freight transport; it amounts to billions of tons and is projected to continue to climb in the coming years.1

The term autonomous trucks is applied to trucks that will be controlled from other sources such as satellites and advanced GPS (Global Positioning Systems), models currently on the road already provide a semi-autonomous mode of operation, in which the unmanned system and/or a human operator conduct a mission, have various levels of human-robot interaction. In the fully autonomous mode of operation, the unmanned system is expected to accomplish its mission, within a defined scope, without human intervention. In the teleoperation mode of operation, the human operator, using video feedback and/or other sensory feedback, either directly controls the actuators or assigns incremental goals, waypoints in mobility situations, on a continuous basis, from off the vehicle and via a tethered or radio linked control device. And, finally, in the remote control mode of operation, the human operator, without benefit of video or other sensory feedback, directly controls the actuators of the unmanned system on a continuous basis, from a location off the vehicle and via a tethered radio linked control device using visual line-of-sight cues.

By 2027 fully autonomous trucks, including truck platoons of two or more trucks in which all trucks have a driver, but only the driver of the lead truck has full control of the vehicle, are anticipated to appear on highways.3 In spite of this transition, it is anticipated that the delivery of goods and services will be much the same as it has been for decades in the sense that drivers remain critical to the process.

Yep sticking with Full Self Driving needs to be Full Self Driving. The goal is autonomous vehicles. It is a Tesla Goal Also.

They should have rolled it into Enhanced Auto Pilot. If it is truly Beta they could have just put the features in there instead of offering something they don't actually have. This would have benefited them if they actually delivered it.

Read the last paragraph the end goal is to eventually eliminate the other drivers not in the lead, but it will take time.
My point was more the technology, the use of the LIDAR and RADAR, not to dissect the Article in and of itself.

 

[BrendonHolt]

This came straight from Elon:

”So he wants the system to be ten times safer than a human driver”

Tesla News and Reviews | InsideEVs

Get the latest on new Tesla models including spy shots, first drives, in-depth reviews, concept car photos, new features and more.

insideevs.com

And it should be. Frankly, overtime it can be, but lots of variables.

 

[ewoodrick]

This came straight from Elon:

”So he wants the system to be ten times safer than a human driver”

Tesla News and Reviews | InsideEVs

Get the latest on new Tesla models including spy shots, first drives, in-depth reviews, concept car photos, new features and more.

insideevs.com

The "bar" and what Elon wants is two different things.

At this point, it IS safer than humans.

And my sentence was not about safety, it was the basic ability to drive. The car doesn't NEED to be able to drive in denser fog than humans can. Sure it may be nice, but then again, if it does, it would tend to be unsafe. Why? that's because there are humans out there that won't be able to handle a car going 55 mph in a fog that is making then creep along at 10 mph.

 

[ewoodrick]

Yep sticking with Full Self Driving needs to be Full Self Driving. The goal is autonomous vehicles. It is a Tesla Goal Also.

They should have rolled it into Enhanced Auto Pilot. If it is truly Beta they could have just put the features in there instead of offering something they don't actually have. This would have benefited them if they actually delivered it.

Read the last paragraph the end goal is to eventually eliminate the other drivers not in the lead, but it will take time.
My point was more the technology, the use of the LIDAR and RADAR, not to dissect the Article in and of itself.

Then let's solve your problem. FSD is not here today, only the FSD beta. No matter how loud you complain about it, you won't make it here any faster. If you bought a car expecting it to have FSD, then either I'm sorry and you need to try to return or sale the car.

Again, FSD is not here today. We all know that. We don't need to hear people complaining about it.

Do you have anything positive to add to the forums?

 

[EVNow]

You are just being argumentative. This is not about how we as humans drive, it is about automated driving systems. The point is that while vision can do a lot vision, has limitations. It is foolish to rely entirely on vision when you can improve on it. Consider a foggy road - your eyes can see say 50 feet, radar can see hundreds of feet.

Tesla says the negatives of low-res Radar outweigh the positives. Show us you are right and they are wrong.

 

[ATPMSD]

Tesla says the negatives of low-res Radar outweigh the positives. Show us you are right and they are wrong.

They could have fixed the problem but Tesla made a decision to not address the issue since Elon’s vision is a 100% vision system. This was stated by their engineering team in one of their tech days video. This was, of course, a business decision, but I am not sure why you even pose the question. Radar + vision is better than vision alone, as evidenced by the fact that they are now adding radar back.

 

[EVNow]

They could have fixed the problem but Tesla made a decision to not address the issue since Elon’s vision is a 100% vision system. This was stated by their engineering team in one of their tech days video. This was, of course, a business decision, but I am not sure why you even pose the question. Radar + vision is better than vision alone, as evidenced by the fact that they are now adding radar back.

You are misrepresenting what they stated.

They said, radar was giving so many false negatives, it was more trouble than it was worth.

https://thenextweb.com/news/tesla-ai-chief-explains-self-driving-cars-dont-need-lidar-syndication

“We deleted the radar and are driving on vision alone in these cars,” Karpathy said, adding that the reason is that Tesla’s deep learning system has reached the point where it is a hundred times better than the radar, and now the radar is starting to hold things back and is “starting to contribute noise.”

 

[ATPMSD]

They said, radar was giving so many false negatives, it was more trouble than it was worth.

Agree, it was more trouble than it was worth to fix, for sure. But I have no idea what you took exception to in my post? I said:

“..while vision can do a lot vision, has limitations. It is foolish to rely entirely on vision when you can improve on it.”