Poll: Is FSD a complete crock?

[S4WRXTTCS]

What I got from speaking to one driverless perception engineer is that the best traffic light recognition is at 99% now. That doesn't seem reliable to me, but if a human is suppose to augment it, then I guess it is passable. Still seems very dangerous to me, because people will tend to tune out and not watch the traffic lights closely.

Is that 99% internal at Tesla? Or were they referring to the industry as a whole?

In any case I'm firmly of the belief that self-driving cars need to be equipped with V2E (vehicle to everything) technology so they can be told not only the status of the light, but when it will change.

That way it accomplishes multiple things.

It augments the >99% accuracy of the cars ability to visually identify the light
It can help the car determine the correct speed to go to hit the green light
It can eventually be used to optimize traffic through a light

It's going to take awhile to get there so I see true self driving being a slowly expanding white list of areas that are modernized for it.

For L2 driving I share your concerns that it's going to be dangerous especially at that 99% level.

 

[diplomat33]

Interesting, I have the opposite opinion. I think it is a very high bar. I guess that speaks to our expectations. I have very low expectations and you have high ones.

Remember, you said driver intervention, not driver engagement. A driver intervention would mean that the driver has to disengage AP. So 25% of commutes with no driver interventions means that 75% of commutes would require the driver to disengage AP at least once, presumably because AP can't handle the type of intersection, or type of road, or construction zone, etc... So I could consider that a very low bar. That would not be feature complete at all.

Are you sure, it is not the other way around, 25% with driver intervention? That would make more sense because it would mean that FSD could handle 75% of commutes.

 

[DanCar]

... self-driving cars need to be equipped with V2E (vehicle to everything) technology so they can be told not only the status of the light, but when it will change. ...

I wonder if there is anything we can do to accelerate the adoption of traffic lights communicating directly to devices?

 

[Zythryn]

I don't think FSD is a crock. But I think it will take longer than 1-2 years before we can sleep in our Teslas while it self-drives.

My thoughts exactly.

 

[Dubes]

I think this poll is a biased crock. Yes Tesla releases incremental changes to FSD with the goal of eventually reaching level 3. For those of you who will only be happy when we reach level 3 with no risk of accidents, that will never happen. The goal is to continually improve safety beyond the unaided human driver, and that is currently happening.
If you don't like and appreciate the amazing steps Tesla has made so far then please don't use and purchase Tesla's amazing FSD. There are many of us who are grateful for the opportunity to use FSD and maybe it's just not for you. I personally think it is unbelievable what Tesla has accomplished so far and I am immensely enjoying the ride.

 

[DanCar]

Unless everyone in the industry is lying about their disengagement data there is no way the reliability is 99%. You can't go 10k+ miles between disengagements with 99% traffic light recognition ...

Good point. Perhaps the 99% is for a specific frame and the software averages out across frames to increase the accuracy. If I remember correctly Tesla's accuracy was at 97% about 6 months ago. Don't remember the source for that. Again they could be averaging across frames to improve accuracy.

 

[cucubits]

On the current hardware, yes it's 100% complete crock.

Maybe tesla will improve/add some features but full 100% self driving without driver intervention, no way. They rely way too much on image recognition from a few cameras.

 

[Dubes]

On the current hardware, yes it's 100% complete crock.

Maybe tesla will improve/add some features but full 100% self driving without driver intervention, no way. They rely way too much on image recognition from a few cameras.

Clearly not for you. Don't use it. I personally love Tesla FSD. It is awesome.

 

[Daniel in SD]

In any case I'm firmly of the belief that self-driving cars need to be equipped with V2E (vehicle to everything) technology so they can be told not only the status of the light, but when it will change.

It doesn't seem like it's worth the cost to me. It would be useful if it's what holding you back from achieving safety greater than a human but I doubt that will ever be the case. I think there are far more difficult vision problems to solve than recognizing red lights.

 

[DanCar]

I personally love Tesla FSD. It is awesome.

It is not FSD is it? That is the crock part. It is great you think it is awesome, but it is being marketed as something it is not. I think it is awesome also, but I think Tesla should lower expectations.

 

[splashrun]

Compared to the Volkswagen Golf GTE, that had lane and automatic speed assist, I found it assisted better than the Tesla M3, but was less autonimous. The first year i had the golf, before they updated it in the garage, it was far better, after that, it was worse every update. Garage told me it was because "regulations'. From this "assisted" car to the M3 "autonomous", I feel no difference, apart from the m3 being better most of the time of following the road, but so much unreliably, I rather drive myself.
Short of it: I am happy to have paid for improvement, but so far, not much seen vs ICE/hybrid Volkswagen stuff.
I think the ID.3 might yet be the Beetle of 2020. And Tesla the euhh, ford T?

 

[Daniel in SD]

Clearly not for you. Don't use it. I personally love Tesla FSD. It is awesome.

I think this thread is referring to pre-March 2019 FSD which is different from your FSD. It’s confusing! Maybe we should call it Real Full Self Driving.

 

[splashrun]

I think this thread is referring to pre-March 2019 FSD which is different from your FSD. It’s confusing! Maybe we should call it Real Full Self Driving.

FSD depending on local law and driving circumstances.
TBH, here in NL sometimes I dint know what traffic wants from me. I am often not surpised FSD doesnt underrstand either. Maybe in Nevada it does?

 

[Dubes]

It is not FSD is it? That is the crock part. It is great you think it is awesome, but it is being marketed as something it is not. I think it is awesome also, but I think Tesla should lower expectations.

I tried FSD and asked about it and researched it before I gratefully purchased it with all future upgrades. I can understand how some people are annoyed at it being called FSD, but would anyone purchase it based just on the name?

 

[Dubes]

I think this thread is referring to pre-March 2019 FSD which is different from your FSD. It’s confusing! Maybe we should call it Real Full Self Driving.

I didn't know. What is different?

 

[MichaelP90DL]

I have the feeling that federal regulators, states, cities and whatever else will allow full self driving only on roads equipped to keep those cars in certain lanes, through electronic means. The prospect of FSD prompted me to look at road conditions that would probably mess it up. When it's released, I expect there will be a lot of wrecks, injuries and deaths. I hope I'm wrong but I'm concerned that I'm not.

 

[Jack6591]

Technology is advancing at an ever-increasing pace. The premise of this poll is inane. Just as my bothering to respond to obvious FUD is a waste of time. Idiotic.

 

[Dubes]

I have the feeling that federal regulators, states, cities and whatever else will allow full self driving only on roads equipped to keep those cars in certain lanes, through electronic means. The prospect of FSD prompted me to look at road conditions that would probably mess it up. When it's released, I expect there will be a lot of wrecks, injuries and deaths. I hope I'm wrong but I'm concerned that I'm not.

FSD will continue to have new features and improvements, but one important aspect that it is already extremely good at is staying in a lane. I look around while using FSD and see human drivers swerving all over their lanes and out of them. Not just occasionally but all the time.
You are right to expect lots of wrecks and injuries because humans cause them. I am thankful that we are finally on a trajectory to eliminate most of them in the next 100 years. The bar has been set quite low by us.

 

[DanCar]

Good point. Perhaps the 99% is for a specific frame and the software averages out across frames to increase the accuracy. If I remember correctly Tesla's accuracy was at 97% about 6 months ago. Don't remember the source for that. Again they could be averaging across frames to improve accuracy.

Most traffic intersections have more than one light, so averaging across lights will improve accuracy also. Flow of traffic and what the walk sign says can also be used as clues.

 

[Daniel in SD]

I didn't know. What is different?

All the currently released FSD features are part of the Enhanced Autopilot option package that was sold before March 2019. The old FSD package was supposed to be fully autonomous. Tesla sold it beginning in Nov 2016 and there still aren't any released features for that package. The new FSD package is not promised to be autonomous, has actual released features, and is much less of a "crock" IMHO.