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

Full Self-Driving Capability

This site may earn commission on affiliate links.
My take on the "history of computer processing power" comment. I am getting the iPhone X and even though computers are getting cheaper it is the most expensive iPhone to date. Also, memory upgrades on iPhones are always much more expensive then they are to make. I would be shocked to see FSD being cheaper just because computers are cheaper per compute power of the chip. I did buy FSD for $3,000 on 9/28/17. When I ordered it said it would be $4,000 if I bought after delivery. But my understanding was the $4,000 was not a fixed number for the life of my ownership of the car. From the beginning I felt I definitely wanted FSD (or any part of it like stop lights etc) and did not want to to pay $4,000+ later. I also felt that they may need to replace my hardware (and I have AP 2.5) and wanted to make sure I would/could get that upgrade. Of course, they could come back to us and say "sorry" but we can not deliver FSD on your car and here is your money back (or something else). But, I think if you do not buy it up front and they for some reason feel it is not worth it to upgrade the hardware and add the software for only $4,000 then I would expect it to be more to get the feature.


I realize the cost to manufacture hardware for FSD vehicles will decline overtime but I don’t see that being the reason for setting FSD pricing. The expense in my opinion is in all the man hours it’s taken Teala to get HW 2.0 & 2.5 to its current state and the hundreds of thousands more hours of teala engineers to get Teala vehicles to FSD. Not to mention, it’s a little scary to think that something your paying for that has your life in its hands (FSD) becoming cheaper. Seems like just the liability alone by Tesla is enough to keep the FSD feature to continue to increase as the more miles in FSD the greater the likelihood of lawsuits, IMHO.
 
Technically, no, it's neither a promise that the option will always be able to be added later (prior case: Ludicrous retrofit/upgrade, was offered for a fee for a while, now just nope, gone, they don't feel like doing it anymore)…. nor is it that the price is stable (it could go up or down later. I would bet money it goes up especially if it requires hardware retrofit).


But I mean, it's all about whether or not that's enough justification to sink $3000 up front for a car that you may or may not even own when the feature gets delivered. A lot of people are saying skip it, and I can understand the reasoning.

If you come to the end of your finance deal and they still haven't delivered FSD, you should be looking for a refund plus interest.
 
One possibility to consider is that Full Self Driving (FSD) is not a precisely defined term with clearly defined functionality. Sure, Tesla could provide FSD with the current hardware, but they may eventually come out with FSD+ that can do a few extra things that FSD can't do.

Current AP2.5 will get you safely from point A to B, but FSD+ may be able to provide (a) some additional features or (b) higher safety levels or (c) operate in more complex situations (eg. snowstorms and shanty towns in India).

I would expect that the initial FSD roll-out will be limited to certain well-laid out areas and Tesla will expand the roll-out over time as they feel confident with the technology. They could limit the working areas of FSD if the current hardware is not sufficient and work-around the limitation by having the navigation software avoiding driving through it.

This is all speculation and my only point is that there is a lot of wiggle room available to Tesla.
 
One possibility to consider is that Full Self Driving (FSD) is not a precisely defined term with clearly defined functionality. Sure, Tesla could provide FSD with the current hardware, but they may eventually come out with FSD+ that can do a few extra things that FSD can't do.

Current AP2.5 will get you safely from point A to B, but FSD+ may be able to provide (a) some additional features or (b) higher safety levels or (c) operate in more complex situations (eg. snowstorms and shanty towns in India).

I would expect that the initial FSD roll-out will be limited to certain well-laid out areas and Tesla will expand the roll-out over time as they feel confident with the technology. They could limit the working areas of FSD if the current hardware is not sufficient and work-around the limitation by having the navigation software avoiding driving through it.

This is all speculation and my only point is that there is a lot of wiggle room available to Tesla.

Autopilot

Excerpt from the above Tesla URL said:
The system is designed to be able to conduct short and long distance trips with no action required by the person in the driver’s seat.

It's a little vague, but the fundamental difference is that EAP's target is advanced driver assistance, while FSD's target is autonomy.
 
Autopilot

It's a little vague, but the fundamental difference is that EAP's target is advanced driver assistance, while FSD's target is autonomy.

My earlier post wasn't comparing EAP to FSD. I was comparing FSD to some potential new FSD+. For example, FSD claims to be twice as safe as a human driver. FSD+ could claim to be over 10 times safer than a human driver. Another example, is that FSD+ may provide more route options (because it can handle more complex environments) and hence potentially get you to your destination faster. A last example, is that FSD+ could operate within tighter parameters (eg perform lane changes into tighter gaps).

Another way to look at it is, as soon as FSD has been reached, I doubt Tesla will stop improving it. I'm sure improvements will continue long after they achieve FSD, and some of those improvements may require updated hardware.
 
On the other hand, buying FSD is an insurance that you will get this feature for $3000 no matter what hardware upgrades are required. E.g. you bought the feature, not the hardware. With the obvious risk of the feature being delayed many years, you'll be the first to get it.

If you buy without FSD, Tesla does not have to give you this feature later for $3000. It might be $6000 or they don't want to sell it at all if it turns out adapting the hardware is more work than it's worth. Rendering the vehicle out of date wanting to buy a new Tesla if FSD is really something you want.

Good point... it is a gamble either way if owning an FSD Tesla is your end goal.

I have an inventory car (an MS) that came with the FSD option -- it was the quickest path to what I wanted in terms of other options -- I then bought an MX without the option otherwise it was 3K more, i have the future upgradability set at 4K.... will they raise the price? Dunno -- but it is a gamble.

Personally, I think the current HW is not FSD capable, I will be surprised (pleasantly) if it becomes reality (actually wondering how Tesla will compensate me on my MS if FSD fails to materialize).
 
Last edited:
One possibility to consider is that Full Self Driving (FSD) is not a precisely defined term with clearly defined functionality. Sure, Tesla could provide FSD with the current hardware, but they may eventually come out with FSD+ that can do a few extra things that FSD can't do.

Current AP2.5 will get you safely from point A to B, but FSD+ may be able to provide (a) some additional features or (b) higher safety levels or (c) operate in more complex situations (eg. snowstorms and shanty towns in India).

I would expect that the initial FSD roll-out will be limited to certain well-laid out areas and Tesla will expand the roll-out over time as they feel confident with the technology. They could limit the working areas of FSD if the current hardware is not sufficient and work-around the limitation by having the navigation software avoiding driving through it.

This is all speculation and my only point is that there is a lot of wiggle room available to Tesla.

To be fully self driving, I would think that a car would have to be able to communicate with all of the other vehicles on the roadway and vice versa.
It has to be able to see outside of the immediate vicinity. We are not there yet. The software and hardware are not there yet and certainly won't be there
for this entire generation of automobiles. I'd suggest not fussing about it and check back in another decade.
 
To be fully self driving, I would think that a car would have to be able to communicate with all of the other vehicles on the roadway and vice versa.
It has to be able to see outside of the immediate vicinity. We are not there yet. The software and hardware are not there yet and certainly won't be there
for this entire generation of automobiles. I'd suggest not fussing about it and check back in another decade.

Why do you think FSD requires the cars to communicate or to see outside of the immediate vicinity? Human drivers don't really communicate between cars, nor do we have nearly as many sensors as even AP 2.0 vehicles. There are two extremes for solving FSD... a fully connected road system where cars communicate with each other and the road (lights, signs, etc..) or a solution that mimics what humans do today (use our eyes and ears).

Elon has repeatedly said Tesla is shooting for the 2nd route and so it becomes a massive computer vision problem. Computational hardware isn't really the problem since these workloads a highly parallelizable and sure they may draw more power than we'd like but the software is the real challenge.
 
  • Like
Reactions: vandacca
I just wonder if their cameras and neural nets are going to be enough to build up a detailed map of the environment, or if they plan on fudging it somehow by accepting a higher failure rate.

I can just see all sorts of corner cases where AI trained on highways won't have a clue. They promised it will take your kids to school by itself so it will have to deal with all that.
 
To be fully self driving, I would think that a car would have to be able to communicate with all of the other vehicles on the roadway and vice versa.
It has to be able to see outside of the immediate vicinity. We are not there yet. The software and hardware are not there yet and certainly won't be there
for this entire generation of automobiles. I'd suggest not fussing about it and check back in another decade.
Totally agree. That's called V2X communication and a bunch of companies are working on the specs and standards. Our company is doing testing of it.
Human drivers constantly communicate, explicitly or implicitly. It might be a glance "I've seen you", or a flashing of the headlights "I let you in", etc. We're also good at guessing another driver's behavior. If you see an aggressive driver, you know he might cut you off or not let you in, getting on the highway. Artificial Intelligence is not up to that yet. But the problem with V2X is that it only works if nearly all cars have it and when the human has been removed from the equation (i.e. all cars are self-driving).
I'm also working on radar and lidar test systems where we emulate test scenarios with multiple cars around your car. We basically spoof the radar sensor with signals that look like there are all these cars around it. So then car companies can test how their car behaves (without driving the car) and they can make regression tests so they can see how the car changes behavior whenever they update the software. Most of this is in the R&D phase. It will be many years before there is a decent FSD system.
I've seen the Google, oops Waymo, cars for years driving around in Mountain View. They drive like grandmas (apologies to the grandmas), use many kW of computer power, have very expensive sensor systems and have now probably catalogued every bush, tree, house, traffic sign in the city. That's just not very scaleable.
Am I disappointed in Tesla: absolutely not. I like the Autopilot, use it all the time. But I did not get FSD as I don't expect it to get anywhere during the life of the car. There are several companies around here working on pilot-less drones for short-haul passenger flights and I think they will be successful before we have FSD. It's so much simpler flying around in a controlled airspace than driving a car under real-life conditions!
 
Do any of you think we will get some of the FSD features like we get some of the EAP features to get some value of FSD? I mean in advance of full FSD.

Example:
Street light detection
Stop sign detection
Auto lane change on service streets
Auto turns based on GPS

All above still requires the driver to be in control like now. “Beta”
I get a lot of benefit from the current features of EAP even though some are beta.

Maybe wishful thinking.
 
Auto turns based on GPS

I think that is a long way off. Think about how far they need to come from the current capabilities which can't even keep the car in a lane reliably. They need to recognize the type of junction, stop if required, look for other vehicles, traffic lights, figure out who has priority and when it is safe to move, predict what other drivers will do, and then calculate a turn that ends up in the correct lane.

Google did some video talks about all this stuff, it's pretty complex. Tesla are focusing on basic road following and lane keeping and still struggling to make that work reliably with their hardware.
 
I put in for the FSD. R&D costs money. The more folks contribute, the better/faster the technology will be. And plus I'll have the $1k discount when it releases ;)
I put in the for FSD as well, and I'd like to say I have no doubt it will be money well spent.... that is, until the free FSD uncorking of 2019, six months after FSD goes live! Do'h!
 
I was told yesterday that FSD will not ready ready for at least a year and that it still has to be approved by the FEDS.
By who? Not sure anyone knows for sure. Anyway, I am not "sure" EAP will be FULLY Complete in 2018 but I think it will get better and better during that time. Regarding FSD I believe (like with EAP) we will start to receive initial features in 2018 but it will not be complete for some time. Might even take until 2020. I.E. About the time The Tesla Network will be available.