When do you think the Model 3 will no longer require a driver?

[daniel]

A related but different question to the "Do you regret buying FSD" poll. Tesla has promoted the "FSD" option package with the promise that at some undetermined time in the future, after the software is developed and regulatory approval is received, Model 3 cars whose owners have paid for the package will no longer require a driver. You will be able to send your car off by itself to pick up your spouse/child/friend/customer and take them where they want to go, drop them off, and then drive itself home again.

This poll is not about regulatory approval. This poll is about when you think the technology will exist to make the car capable of driving itself without a driver in the city and on the highway, and driving/navigating itself to the destination(s) you specify. This means dealing with such things as detours or other issues too recent to be in the map database. And let's say if that the car cannot get to the programmed destination it must be able to return to the starting location, and that cannot happen more than 1% of the time. And it must establish a better safety record than human drivers. I.e., you're statistically safer in this car than you would be if you were driving.

 

[daniel]

P.S. I voted 10 to 20 years.

 

[NerdUno]

HINT...

 

[suwaneedad]

I replied never because I think full autonomy is going to be beyond the technical development lifespan of Model 3.
And for seemingly simple reasons, like dodging random things in the road, or potholes that I'd prefer not to send my car into.
I expect that much in the same way Elon learned this year that humans are better than machines often on the factory floor, he's also now realizing that the same is true of humans vs computers in the role of driver.
That said, heavily assisted driving is here and will continue to get better, of course.
It may take Tesla a very long time to admit FSD isn't going to happen, since doing so will involve lots of refunds to lots of people. Tesla wins by halting sales of FSD now, and then letting the installed fleet on the roads taper away due to age, wrecks, etc. The FSD liability for Tesla has now finally started to gently fade away.

Even if I'm wrong about the above, without fail the regulators will then get involved and very likely spoil the party. One FSD car running over Timmy on his skateboard and things get ugly really fast, and I'm not sure that'd be wrong by the way.

I hope I am wrong about ALL of the above!

 

[JeffK]

3 years maybe, 6 years definitely.

 

[NeverFollow]

I put 'other'. IMO cities need to build first area where self driving car will be able to park when waiting for passengers.
Otherwise there will be everywhere cars waiting in the street with flashing blinker on, like the Uber driver do,
and blocking all the street for any traffic like it is already the case in San Francisco.

 

[Vitold]

I put in 5 years but it could be earlier if Model 3 got more cameras and more processing power.

 

[JeffK]

I put in 5 years but it could be earlier if Model 3 got more cameras and more processing power.

I'm curious why it needs more cameras.

 

[Vitold]

I'm curious why it needs more cameras.

For redundancy, stereo vision, full 360 coverage. I also think that near infrared could be helpful as well.

Many would argue that stereovision is not needed but I think it offers less false positives when car is not in motion and/or objects are not moving.

 

[iaflyer]

Too many edge cases - for example, last month here in Michigan they did a "chipseal" basic road refresh on a medium traffic road. One lane in each direction, probably 8 cars a minute use the road. To those who aren't familiar, they spray tar on the existing road, then immediately drop small gravel chips on the tar, to create a new surface. A lot cheaper than repaving but only lasts 2 years or so. Anyway, the point I'm getting to is that the road was without lane markings or edge markings of any kind for about a month. No delay, it takes a while for the cars to pound the chips into the tar, then they have to sweep it, then wait for it not rain, then they paint the lines. Chipseal - Wikipedia

But what would an autonomous car do? Eventually they might be able to program it to see the edge of the road surface without lane markings, but it gets pretty hard for even humans. Or what if it's turning into a driveway and a small branch has fallen? A human can decide to either drive over it, get out and move it, or stop, park and call someone to help remove it. Just two examples that I think an autonomous would have significant problems with, and are very difficult to solve.

But I really think the biggest problem is liability. Right now, if an Tesla crashes, the driver is at fault, even if the automatic systems were engaged. With an autonomous car, with no "driver" (because it could be a child, or a person who does not have a driver's license, or even empty) who is liable in an accident causes by the autonomous vehicle? Tesla... or your insurance company. Either way, there would be a huge, financially rich target for lawyers to sue heavily. I think that will be the ultimate problem that can't be solved by more technology.

 

[Vitold]

Too many edge cases - for example, last month here in Michigan they did a "chipseal" basic road refresh on a medium traffic road. One lane in each direction, probably 8 cars a minute use the road. To those who aren't familiar, they spray tar on the existing road, then immediately drop small gravel chips on the tar, to create a new surface. A lot cheaper than repaving but only lasts 2 years or so. Anyway, the point I'm getting to is that the road was without lane markings or edge markings of any kind for about a month. No delay, it takes a while for the cars to pound the chips into the tar, then they have to sweep it, then wait for it not rain, then they paint the lines. Chipseal - Wikipedia

But what would an autonomous car do? Eventually they might be able to program it to see the edge of the road surface without lane markings, but it gets pretty hard for even humans. Or what if it's turning into a driveway and a small branch has fallen? A human can decide to either drive over it, get out and move it, or stop, park and call someone to help remove it. Just two examples that I think an autonomous would have significant problems with, and are very difficult to solve.

But I really think the biggest problem is liability. Right now, if an Tesla crashes, the driver is at fault, even if the automatic systems were engaged. With an autonomous car, with no "driver" (because it could be a child, or a person who does not have a driver's license, or even empty) who is liable in an accident causes by the autonomous vehicle? Tesla... or your insurance company. Either way, there would be a huge, financially rich target for lawyers to sue heavily. I think that will be the ultimate problem that can't be solved by more technology.

I do not think that driving w/o line markings is particularly difficult. Se. this 2015 presentation:

 

[Vitold]

 

[daniel]

Too many edge cases - for example, last month here in Michigan they did a "chipseal" basic road refresh on a medium traffic road. One lane in each direction, probably 8 cars a minute use the road. To those who aren't familiar, they spray tar on the existing road, then immediately drop small gravel chips on the tar, to create a new surface. A lot cheaper than repaving but only lasts 2 years or so. Anyway, the point I'm getting to is that the road was without lane markings or edge markings of any kind for about a month. No delay, it takes a while for the cars to pound the chips into the tar, then they have to sweep it, then wait for it not rain, then they paint the lines. Chipseal - Wikipedia

But what would an autonomous car do? Eventually they might be able to program it to see the edge of the road surface without lane markings, but it gets pretty hard for even humans. Or what if it's turning into a driveway and a small branch has fallen? A human can decide to either drive over it, get out and move it, or stop, park and call someone to help remove it. Just two examples that I think an autonomous would have significant problems with, and are very difficult to solve.

But I really think the biggest problem is liability. Right now, if an Tesla crashes, the driver is at fault, even if the automatic systems were engaged. With an autonomous car, with no "driver" (because it could be a child, or a person who does not have a driver's license, or even empty) who is liable in an accident causes by the autonomous vehicle? Tesla... or your insurance company. Either way, there would be a huge, financially rich target for lawyers to sue heavily. I think that will be the ultimate problem that can't be solved by more technology.

The ability of a human driver to get out and move a branch is something I had not thought of before. Of course, some very old drivers could not do that either. Maybe those people should not be allowed to drive. You raise a significant issue.

Insurance is a non-issue. The owner of the car buys insurance that covers accidents. Just like now. If there's an accident, the insurance company pays. Tesla buys insurance that covers them in case a driver's insurance company sues them. Car companies probably buy insurance now in case they get sued over faulty cars.

An autonomous car running over little Timmy on his skateboard will give ammunition to the Luddites who oppose all technology, and in this age of talk radio and disinformation it could become a real problem. But as long as autonomous cars are safer than human drivers, leveler heads will be able to point out the statistics, and that autonomous cars kill far fewer kids on skateboards than human drivers kill. And this will not be Tesla against the world. This will be the entire automotive industry pushing and lobbying for autonomous cars, because they're all working towards them, and the entire insurance industry will be on the same side due to the demonstrable safety advantage. Legislators are seldom rational or knowledgeable, but they know who butters their bread, and no industry in the country is against autonomous cars.

The only question is when the technology will be ready. The branch in the driveway is just one of many problems that need to be solved.

Maybe the car stops, honks its horn repeatedly, and shouts "Will somebody please move this branch for me?!"

 

[JeffK]

For redundancy, stereo vision, full 360 coverage. I also think that near infrared could be helpful as well.

Many would argue that stereovision is not needed but I think it offers less false positives when car is not in motion and/or objects are not moving.

There is full 360 coverage, currently.

I'd argue a second camera at each spot is not needed for redundancy because if the car is disabled it should safely pull over, I'm sure it can manage with a camera down. Even with a second camera there's no scenario where you should proceed to your final destination with a disabled/blind sensor.

As for stereovision, in motion you have this with a single camera, when stopped you can still use a single camera to determine velocity and direction of cars around you. There are quite a few papers on vision based vehicle speed measurements.

I agree IR would be helpful, most sensors can detect IR already to some degree. I don't know if there's an IR filter on the sensors Tesla uses. That said, at night a Tesla normally has its lights on.

 

[Tam]

Just a reminder, there have been autonomous vehicles that have been running on their own without human operators for years.

They operate in strict geofenced locations such as airport rail link, Disneyland, automated metro and light rail systems... with strict exclusions of unexpected factors such as obstacles...

If technology is that good, we should see it in trains first but that won't happen for a long time.

Commercial airplanes should plot their routes free of obstacles so we should supposedly also see Commercial autonomous airplanes first.

Self-flying plane: Boeing to test pilotless aircraft in 2018

Thus, if cars are placed in the same environments as those automated airport people transporters, they can be autonomous too.

The speed of progress on trains, commercial airplanes might be a good gauge of how fast Autonomous cars will arrive.

 

[JeffK]

so we should supposedly also see Commercial autonomous airplanes first

This depends, if you look at it from the other perspective when considering the need for autonomous vehicles.

The highest need is in cars as that's where the fatality rate, due to human factors, is greatest. I think it's going to come as fast as possible, even if it's not perfect.

 

[Vitold]

It is 2018 and trains still do not even have positive control (PTC).

At this point self driving technology is either too expensive or unreliable.

 

[rachelhikes]

I think driverless operation in the full range of driving conditions is a long way off. (I voted 10 to 20 years.) But I see no well-defined upper bound as to how good driver assistance can get in the interim. I don’t see Tesla’s existing sensor system as a problem. It is primarily a matter of software and secondarily a matter of processing power.

 

[Maximilien]

In optimistic view, I say 2-3 years. Realistically 5-10 years.

 

[Tam]

...fatality rate...

That's what Uber said when it entered into the business of Autonomous Vehicles: Safety!

But it did the opposite: Money$$$.

Which resulted in the death of a pedestrian.