FSD discussion

[Ogre]

My best guess is a tsla shareholder was the driver, was watching tsla stock price in the car in real time and passed out. /s

Nah, driver was TSLAQ. He’d been driving laps on the bridge for hours praying for a phantom braking event that causes an accident. There were 4 more of his buddies in the cars behind him. /tinfoilhat

 

[EVNow]

So does Autopilot/NOA actually turn off in tunnels? And if so how far into a tunnel, immediately?

No.

Like various other "FSD" accidents, we have to wait for investigation report / Tesla statement. This doesn't look anything like what you'd expect out of AP/NOA, let alone FSD.

 

[StealthP3D]

There was a time when Elon balked at the idea of even using cabin cameras to track eye movement, sensing steering wheel torque was the sole method of measuring driver engagement. The cabin cameras were initially built in to monitor passengers using your Robotaxi lol, or at least thats what Elon tweeted.

Elon did say the cabin camera is put there to monitor robo-taxi customers. But this is one of the things I really appreciate about having a leader with so much foresight. Because at a time when cars were becoming increasingly loaded with technology, and people were becoming increasingly concerned about privacy and other issues, and rightfully so, Elon knew there was a real chance of backlash about in-car cameras and it was a sensitive issue for people that were not yet necessarily fully comfortable with a camera in the cabin.

I'm sure Elon knew a camera in the cabin could be used in any manner compatible with its field of view/resolution, and that it was just good planning to position it in the least threatening manner possible. It's smart to ease skeptical people into new ideas in stages. People can always opt out of it if they are willing to give up the benefits associated with it but to introduce it to the public as primarily for monitoring the driver would have created an outcry from TSLAQ types that was not worth the trouble. Elon is smart the way he positions things like that. He was also smart to make the field of view wide enough that it works to monitor the entire cabin. Two cameras would have been overkill.

 

[Electroman]

Why post a video that is at best misleading and at worst completely fabricated? I remember watching that video and playing it for my children as proof of what was about to happen. Then 7 years later this story and no true FSD.

I don't understand what rock people are living under. There are hundreds of videos of tesla FSDb users driving in much more complex environment than that. In fact today a Tesla can drive the same section that was shown in that 2016 video, without much fuss. But in order for masses to use it in every other situation there is still some work required, but to say that Tesla has done nothing in this space is just outright false.

 

[traxila]

I don't understand what rock people are living under. There are hundreds of videos of tesla FSDb users driving in much more complex environment than that. In fact today a Tesla can drive the same section that was shown in that 2016 video, without much fuss. But in order for masses to use it in every other situation there is still some work required, but to say that Tesla has done nothing in this space is just outright false.

Who said that?

My issue would be with the production of the video, not their statements on FSD.

I said there is no true FSD. This is a fact. Not that "Tesla has done nothing in this space".

My opinion is that Tesla is the leader in FSD and will be the first to develop a true FSD.

 

[Singuy]

Who said that?

My issue would be with the production of the video, not their statements on FSD.

I said there is no true FSD. This is a fact. Not that "Tesla has done nothing in this space".

My opinion is that Tesla is the leader in FSD and will be the first to develop a true FSD.

No true level 5 FSD. Tesla can fully self drive in many areas across US, just not everywhere.

 

[EVNow]

Tesla can fully self drive in many areas across US,

Where ? It is a pretty decent ADAS for enthusiasts like us at this point. Nothing more.

 

[Singuy]

Where ? It is a pretty decent ADAS for enthusiasts like us at this point. Nothing more.

Disney Hollywood studios through Disney magic Kingdom then back door Disney university to my house. 18 miles, drive had one accelerator push. I can pick lots of routes and it self drives. It did really well in Tampa too. Point to point without you doing much. I sold a Tesla to a friend who witness my car went from house to restaurant without me doing a single thing. This is self driving to me. I'm sure you can pick many routes that can achieve perfect or near perfect driving.

You know plenty of these routes exist now, where the car achieves 100% self driving. So many that people are starting to defend FSD in online forums vs trashing it constantly half a year ago.

 

[EVNow]

Disney Hollywood studios through Disney magic Kingdom then back door Disney university to my house. 18 miles, drive had one accelerator push. I can pick lots of routes and it self drives. It did really well in Tampa too. Point to point without you doing much. I sold a Tesla to a friend who witness my car went from house to restaurant without me doing a single thing. This is self driving to me. I'm sure you can pick many routes that can achieve perfect or near perfect driving.

You know plenty of these routes exist now, where the car achieves 100% self driving. So many that people are starting to defend FSD in online forums vs trashing it constantly half a year ago.

We should take this to the other thread.

Unless you drive that "route" thousands of times in all kinds of traffic and weather conditions - it is simply incorrect to say "car achieves 100% self driving.". I also have a LOT of zero disengagement drives - but the same routes also have disengagements once in a while.

ps : I'm very surprised to see this kind of argument made in this forum. I thought we'd know better.

pps : BTW, you first said "areas", not particular routes.

 

[FSDtester#1]

We should take this to the other thread.

Unless you drive that "route" thousands of times in all kinds of traffic and weather conditions - it is simply incorrect to say "car achieves 100% self driving.". I also have a LOT of zero disengagement drives - but the same routes also have disengagements once in a while.

ps : I'm very surprised to see this kind of argument made in this forum. I thought we'd know better.

pps : BTW, you first said "areas", not particular routes.

You know I love you brother, but I have multiple routes I know I can take 99% of the time with 0 disengagements or interventions.
On the other hand, I have some I know when and where it will fail. Almost never safety related, just pressing go pedal not to tick off someone behind me while cars wait or turning off phantom blinkers.
So while it is subjective, they eventually will get to level 3 or higher, imo with hardware 4.

 

[EVNow]

You know I love you brother, but I have multiple routes I know I can take 99% of the time with 0 disengagements or interventions.
On the other hand, I have some I know when and where it will fail. Almost never safety related, just pressing go pedal not to tick off someone behind me while cars wait or turning off phantom blinkers.
So while it is subjective, they eventually will get to level 3 or higher, imo with hardware 4.

Moving the discussion here.

There are a couple of things to keep in mind.
- @Singuy was first talking about "areas" - not particular routes. A world of difference.
- Statistically you would have to drive a million miles on that route in all kinds of conditions to say "it can do 100% self-driving" (even though not really 100%). This is the same thing I've said multiple times about Merc "L3" on highways < 37 mph too.
- Even if FSD can drive on particular routes well, don't get complacent. All it takes is one unanticipated event to get into an accident.
- FSD simply doesn't handle heavy Rain / Snow. So, unless you are in a desert, its not handling any route 99%.

 

[Bet TSLA]

Any opinions on whether the continued appearance of new versions of V10 FSD (we're now up to 10.69.25.2) is a good or a bad sign for whether V11 will be released to customers soon?

Me, my expectations are quickly approaching zero. I haven't seen any significant improvements in V10 for months now and I don't see V11. My expectation is that V11 will be about the same on local streets and will be a mixed bag on highways. My hope is that the constantly poor choices in lane selection will get better, but I can't think of any reason to expect that.

 

[Electroman]

My prediction. V11 will start as a disappointment for many, where there would be some minor progress on a few use cases but also regression in my other situations... and then it will get better and better for every minor upgrade that a few iterations later if you compare it to where you are today, there will be significant progress.

And so it goes..

 

[Singuy]

Moving the discussion here.

There are a couple of things to keep in mind.
- @Singuy was first talking about "areas" - not particular routes. A world of difference.
- Statistically you would have to drive a million miles on that route in all kinds of conditions to say "it can do 100% self-driving" (even though not really 100%). This is the same thing I've said multiple times about Merc "L3" on highways < 37 mph too.
- Even if FSD can drive on particular routes well, don't get complacent. All it takes is one unanticipated event to get into an accident.
- FSD simply doesn't handle heavy Rain / Snow. So, unless you are in a desert, its not handling any route 99%.

LoL statistically humans can't even drive millions of miles on any route of any kind without making a mistake. So since I, as a human never passed that great filter, I must not be 100% driving the car.

For me, if the car can get me from point A to B without interventions, then it has full self driving CAPABILITIES
If it got me from point A to B without interventions, then it 100% fully drove me there, therefore it self drove.

I really don't put reliability in there as a qualifier because humans are not reliable, and yet zero people claim they don't fully self drive themselves. Self driving is a matter of ability and control. If it had full control of the car from point A-B, then it qualified as self driving to me anyways. Reliability comes in if you want to measure of how long the car can self drive before making a mistake.

My wife is in a car accidently once every 40k miles. She makes a mistake(like turned into the wrong lane or didn't lane keep correctly) maybe once a drive? If I put my wife in a brand new city, she will most likely make 5-10 mistakes in a day, not knowing where she is going. I'll make half as many mistakes. So both her and I may even perform WORST than FSD beta in those scenarios. I made plenty of U-turns or made the wrong turn while in a brand new city. However no one said I didn't fully self drive the car.

 

[jerry33]

LoL statistically humans can't even drive millions of miles on any route of any kind without making a mistake. So since I, as a human never passed that great filter, I must not be 100% driving the car.

For me, if the car can get me from point A to B without interventions, then it has full self driving CAPABILITIES
If it got me from point A to B without interventions, then it 100% fully drove me there, therefore it self drove.

I really don't put reliability in there as a qualifier because humans are not reliable, and yet zero people claim they don't fully self drive themselves. Self driving is a matter of ability and control. If it had full control of the car from point A-B, then it qualified as self driving to me anyways. Reliability comes in if you want to measure of how long the car can self drive before making a mistake.

My wife is in a car accidently once every 40k miles. She makes a mistake(like turned into the wrong lane or didn't lane keep correctly) maybe once a drive? If I put my wife in a brand new city, she will most likely make 5-10 mistakes in a day, not knowing where she is going. I'll make half as many mistakes. So both her and I may even perform WORST than FSD beta in those scenarios. I made plenty of U-turns or made the wrong turn while in a brand new city. However no one said I didn't fully self drive the car.

Our wives must be sisters, or at least related.

 

[EVNow]

LoL statistically humans can't even drive millions of miles on any route of any kind without making a mistake. So since I, as a human never passed that great filter, I must not be 100% driving the car.

You need to readup on AV testing, why they need to test for millions of miles to be statistically significant etc.

I've posted many of these in the AP sub-forum, you can search (or if I find some time I'll, and post).

Being ignorant about stats is not a bliss

 

[Singuy]

You need to readup on AV testing, why they need to test for millions of miles to be statistically significant etc.

I've posted many of these in the AP sub-forum, you can search (or if I find some time I'll, and post).

Being ignorant about stats is not a bliss

We are talking about two different things. You are talking about reliability high enough to be a robotaxis. I am talking about full self driving. In order for a car to qualify to fully self drive, it just needs to do the drive me to the destination I selected without my input and that drive was fully self driven. If a stranger asked me, "hey did the car drive you here or did you help it drive you here?". My answer would be "yeah the car fully drove me here".

 

[EVNow]

We are talking about two different things. You are talking about reliability high enough to be a robotaxis. I am talking about full self driving. In order for a car to qualify to fully self drive, it just needs to do the drive me to the destination I selected without my input and that drive was fully self driven. If a stranger asked me, "hey did the car drive you here or did you help it drive you here?". My answer would be "yeah the car fully drove me here".

But driving once doesn't make it "FSD". Because next time you drive, it might crash or you may need to intervene.

Common, this isn't rocket science.

Anyway, here's a quick link ....

Test-Driving Alone Cannot Demonstrate the Safety of Autonomous Vehicles

In order to advance autonomous vehicles into daily use, alternative testing methods must be developed to supplement on-the-road testing. Alternative methods might include accelerated testing, virtual testing and simulators, mathematical modeling, scenario testing, and pilot studies.

www.rand.org

ps : I've lots of zero disengagement trips too. I don't claim those routes to be "self driving". When you have a zero disengagement / intervention drive, it just means exactly that - a zero disengagement / intervention drive. Nothing more, nothing less. Just as one day of SP going up doesn't mean it goes up every day.

 

[Singuy]

But driving once doesn't make it "FSD". Because next time you drive, it might crash or you may need to intervene.

Common, this isn't rocket science.

Anyway, here's a quick link ....

Test-Driving Alone Cannot Demonstrate the Safety of Autonomous Vehicles

In order to advance autonomous vehicles into daily use, alternative testing methods must be developed to supplement on-the-road testing. Alternative methods might include accelerated testing, virtual testing and simulators, mathematical modeling, scenario testing, and pilot studies.

www.rand.org

ps : I've lots of zero disengagement trips too. I don't claim those routes to be "self driving". When you have a zero disengagement / intervention drive, it just means exactly that - a zero disengagement / intervention drive. Nothing more, nothing less. Just as one day of SP going up doesn't mean it goes up every day.

View attachment 897734

How would describe your car in relation to you if you arrived at a destination without your input?

What about sitting in a car with a friend, did the friend drive you? The friend can also crash, lower chance than the Tesla but not zero. Does the chance of your friend crashing being lower than a Tesla defines how you were driven there differently?

 

[JRP3]

In order for a car to qualify to fully self drive, it just needs to do the drive me to the destination I selected without my input and that drive was fully self driven.

Other than yourself I don't think most people would qualify a vehicle as "fully self driving" because it made it to one destination. To be "fully self driving" it needs to handle all routes and conditions as least as well as an average driver and it is far from that right now.