How soon will any Tesla drive itself, level 5?

[someguy in KC]

The main reason I am considering buying a Tesla is so that it can drive me to work while I do something else (work while it drives). I understand we are guessing, but how soon do we think this will be possible? Are we talking 1-2 years? 3-5 years? or should I get real about this and understand that it could be more than 10 years before it happens? (Yes, I've seen the videos of these cars driving themselves, but as I understand it, this isnt available to me.)

 

[Tam]

There is some hint of the rapid development in 1/23/2017:

Tesla to transition from ‘Enhanced Autopilot’ to ‘Fully Self-Driving’ as soon as ‘3 to 6 months’, says Elon Musk

But the 3 to 6 month period has gone by and it wasn't even mentioned at Semi/Roadster announcement.

In April 2017 at Ted talk, Elon Musk says 2 years. That's 2019.

If you are familiar with Tesla timeline, you might want to add a little bit more time to it.

The press report of Tesla engineers quitting when Tesla decided to sell full-self driving capability to the public because experts don't think that could technically happen.

nVidia CEO says about 3 years or 2020:

2018: simulated environments and development systems
2019: robot taxis
2020:self-driving cars

Just like going to mars, I think it will happen but the exact timing is just a guess.

 

[thecloud]

Note that the wording of the poll question is slightly different than the thread title:
"How soon will a tesla be available for purchase that will drive itself, level 5?"

If previous experience is a guide, I suspect you'll first be able to purchase a tesla where it is claimed that it will drive itself, level 5, real soon, just as soon as the software is ready. Then it will actually drive itself, a year or two later.

Much will depend on how self-driving cars are regulated. I think we'll see this happen in small, cautious steps. Your self-driving car won't be able to take you to work right away, unless your route to work happens to include a self-driving zone.

 

[Max*]

This poll pertains specifically to Tesla, Tesla is trying to go about the problem their own way with Tesla vision or whatever it's called after the separated from MobilEye.

Will they get there? Sure. Will it before the competition? Who knows.

I voted 5+ years specifically to Tesla. I'd vote 2-5 years wrt to L5 in "general" based on things MobilEye and Nvidia forecast, as well as companies like Audi, Ford, etc.

 

[Tam]

...Much will depend on how self-driving cars are regulated...

There are numerous states that already allow autonomous vehicle testing but there are 3 states that allow operating/selling those cars to consumers as long as they comply to traffic laws: Florida, Michigan, and Georgia .

If the technology is not at its infancy, you would see cars without drivers roaming around in those 3 states.

So, currently, the delay is because of the immature technology, not because of the laws.

 

[someguy in KC]

If previous experience is a guide, I suspect you'll first be able to purchase a tesla where it is claimed that it will drive itself, level 5, real soon, just as soon as the software is ready. Then it will actually drive itself, a year or two later.

Much will depend on how self-driving cars are regulated. I think we'll see this happen in small, cautious steps. Your self-driving car won't be able to take you to work right away, unless your route to work happens to include a self-driving zone.

thecloud:

so, I take your post to mean that you don't believe that any current Model S (or model X) vehicles will have full self driving capability? is this your opinion or something that Tesla has stated?

 

[thecloud]

I take your post to mean that you don't believe that any current Model S (or model X) vehicles will have full self driving capability? is this your opinion or something that Tesla has stated?

I would love for current Model S and Model X vehicles to eventually have FSD in a future software update, but until they actually do, my opinion is that it's best to enjoy the car's existing capabilities and not buy the car for a feature it doesn't yet have.

There's a very lengthy thread already discussing whether AP2 will/won't be capable of FSD:
FSD video completely fake?

Tesla has stated that all AP2 cars "will have the hardware needed for full self-driving capability at a safety level substantially greater than that of a human driver."
All Tesla Cars Being Produced Now Have Full Self-Driving Hardware

But this is like saying all modern pianos have the full 88 keys needed to play Rachmaninoff's piano concertos. While true, you might need to spend a number of years gathering the necessary expertise to fulfill that promise.

 

[Tam]

...you don't believe that any current Model S (or model X) vehicles will have full self driving capability?

That's a good question!

Usually, a company has to prove that its product can function as specified with no speculation whether it will be able to do the work or not.

However, Tesla is different. It sells an autonomous vehicle without letting reporters ride on one to see whether it can pass a simple traffic cone test or not.

As mentioned above, even some Tesla engineers had doubts and quit:

Tesla's Autopilot team reportedly became divided over Elon Musk's self-driving car plans

Without a scientific proof that your Tesla can avoid hitting traffic cones on its own, it's just a guess of what it will be able to do in future.

 

[Ugliest1]

I voted 5+ years for level 5. But if a 10+ years was available, I'd pick that.

OTOH, it seems to me the current level 2 is a huge help to drivers (going from TMC comments since I have a pre-APn car), and level 3 or level 4 would be fabulous. I totally agree with the above comment about buying the car for what it has now, not what it might deliver some time in the future.

 

[Tiger]

Within 1 year, but nobody will actually trust it

 

[cwerdna]

I voted 5+ years for level 5. But if a 10+ years was available, I'd pick that.

I also voted 5+ years for level 5 but yeah, 10+ years is probably a safer answer assuming Tesla doesn't somehow do a partnership with Waymo and can use all of Waymo's IP (or that of some other company that isn't much behind Waymo or somehow further ahead than Waymo).

I also answered w/the assumption that the OP is referring to availability to actual Tesla customers (not employees) of consumer vehicles (not commercial vehicles) in the US, since the OP is from the US.

OP, if you haven't already, you may want to look over the doc linked to at Autonomous Car Progress. And, look at the links in post 41 in that thread. Also, skim Tesla vs. Google Auto/Waymo at Autonomous Vehicle Disengagement Reports 2015 and Autonomous Vehicle Disengagement Reports 2016. Pay attention to how many miles they did on public roads in CA and disengagements per mile.

 

[Tam]

...actually trust it...

In this modern age, science should be straight forward with consistently repeatability results.

If I set my cruise control at 70 MPH, the car should accelerate to that speed in a same scenario and should be verifiable, repeatable by anyone.

The same should happen in automatically avoiding hitting rubber traffic cones or a tractor trailer that's doing a Lateral Turn Across Path.

We need a scientific proof, not blind trust!

 

[Max*]

In this modern age, science should be straight forward with consistently repeatability results.

If I set my cruise control at 70 MPH, the car should accelerate to that speed in a same scenario and should be verifiable, repeatable by anyone.

The same should happen in automatically avoiding hitting rubber traffic cones or a tractor trailer that's doing a Lateral Turn Across Path.

We need a scientific proof, not blind trust!

Given the same inputs, you would expect the same outputs.

It's a completely different comparison to cruise control, which the goal is to just get you to the set speed and stay there. Regardless if there's something in the way.

Problem is that if you have 2 cars driving down the road, back to back, the inputs will not be 100% the same, so you can't expect the outputs to be 100% the same.

 

[Pezpunk]

In this modern age, science should be straight forward with consistently repeatability results.

If I set my cruise control at 70 MPH, the car should accelerate to that speed in a same scenario and should be verifiable, repeatable by anyone.

The same should happen in automatically avoiding hitting rubber traffic cones or a tractor trailer that's doing a Lateral Turn Across Path

the difference between traditional cruise control and self driving is comparable to the difference between being able to toss a football 20 yards, and being able to play quarterback in the NFL.

 

[Tam]

the difference between traditional cruise control and self driving is comparable to the difference between being able to toss a football 20 yards, and being able to play quarterback in the NFL.

I think some people missed what I meant.

I mean consistent scientific results do not require trust.

Whether I trust or not, using a hand gun at close range would consistently shoot through a paper book even when I trust it should not.

On the other hand, if I use bullet proof material, that same hand gun cannot pierce through it whether I trust it or not.

One results in death, while the other may result in injuries but the results are consistent: One material, paper, cannot stop the bullet while the other, bullet proof material consistently does.

So, for Self-Driving, I expect it to avoid hitting rubber traffic cones that redirect traffic in construction zones and the results should be repeatable.

It's a matter whether the system works or not.

When we have to guess that it will work in future or to use trust, that is not the same as repeatability in science.

It's a bet, hope...!

 

[TaoJones]

Joe Customer-accessible Level 5?

Heeheehohohahaheeheehohoho!

Bwhahahahahahahahaha*hic*!

Without restrictions as to road type and jurisdiction and without failure? Greater than 5 years for sure.

Level 5 for limited use cases with all sorts of disclaimers and caveats?

Soon.

*polite cough*

 

[cwerdna]

Level 5 for limited use cases with all sorts of disclaimers and caveats?

Soon.

*polite cough*

That wouldn't seem to meet the definition of level 5 at Wayback Machine.

 

[Phrixotrichus]

Level 5 for limited use cases with all sorts of disclaimers and caveats?

Soon.

*polite cough*

That would by definition be somewhere between lvl 3 and 4

@Topic
5+ years, and If there`d been a 10+ years I´d probably have voted that.
I can see lvl 4 in the next 5 years maybe, but full autonomy in all cases....well I kinda expect that to take a while longer.

I´d be seriously happy to be completely wrong here, though.

 

[Milkfloat]

Level 5... aka all roads all conditions >5 years.
I think they probably get FSD on freeways/motorways etc (aka Level 4) in <5 years. That will mean you can ask the car to drive on freeway but you take over on local roads.... thats a big chunk of time in many long distance and a large number of short distance drives. I'd be happy with that - I really dislike that type of monotonous driving.

 

[The Duke]

You want to work while driving, that happened long ago with Autopilot. FSD is a ways off, but guaranteed at purchase.

Wifey and I travel about 50 miles to her doctor every Wed. Once in the carpool lane the drive is bumper to bumper up to 80 mph and down to full stops, no lane changes. I get out the laptop for an hour after setting the "laptop" as driver to re-position seat, steering and mirrors. I still don't have 100% trust although Nicki has never failed me so not as productive as at the office.
You don't need 100% level 5 to work while driving, just 100% autonomous driving some of the time.