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

FSD rewrite will go out on Oct 20 to limited beta

This site may earn commission on affiliate links.
Waymo and Cruise only remove the safety driver AFTER it has been thoroughly tested that it is safe to do so and only in a geofenced area. They do a lot of testing and validation with professional safety drivers before going driverless. They are not just releasing "beta" software without a driver with no testing.

I would still argue that FSD beta with a vigilant driver is safer than a driverless Waymo. But of course that argument falls down as soon as the Tesla driver loses attention or trusts the system too much.

That being said, FSD beta is probably about as safe as Waymo was when they started testing systems with safety drivers. Just because one is a customer and the other is an employee doesn't make them any better at paying attention.
 
I would argue that I do not want my car to be hurt/scratched/damaged in any way, so I will be more vigilant.
When it's a company car, well you know how rentals are treated, right? :)

In the case of Waymo or Cruise, it's not like a rental car. The safety drivers are employees. It's their job. If they let the car get damaged or worse, they could cost the company a lot of problems and even possibly get fired. So I think they have very strong incentives to avoid any incidents.
 
As a professional pilot for 45 years, I love this stuff, and can't wait for it to come to my car, but I will never think of it as anything more than workload reduction software, no matter how sophisticated it becomes. I've been too long in cockpits staying prepared for anything and everything that might go wrong. I don't believe I would ever trust a car anymore than I would an aircraft to operate perfectly 100% of the time, I will always have to monitor them.

That said, I am very impressed with how far things have come in my life. Autoland in aircraft has been around fro a while in airliners, for landing at places like Heathrow in zero visibility and such, but it is monitored closely by the pilots when in use, and the aircraft has to be certified and pilots trained especially for these types off approaches.

Now Garmin has developed a system for small aircraft, that monitors the pilot for incapacitation, as well as can be initiated by a passenger in case of pilot incapacitation. It can find the nearest airport, contact ATC, change power settings, descend, lower flaps and gear, land, stop, all while avoiding terrain. Really impressive stuff.

You guys might enjoy this if you haven't seen it before, it's the promo video from Garmin, but there are separate demo videos on YouTube as well. Pretty cool stuff.


 
I’ve seen this assertion made many times without any obvious proof. Although it definitely is a possibility, Elon has also once said that it’s true “no matter how you slice and dice it” meaning it’s still safer when normalizing for all the obvious variables. Also, if you read the fine print about how Tesla determines what is an accident, it is quite conservative. The reporting rate for Tesla is something approaching 100% while NHTSA can only gather data via accident reports. If a Tesla is rear-ended it is considered an accident. If autopilot disengages 5 seconds before an accident, they also count it as an accident.

I also believe the Autopilot team would refuse to push out anything knowingly harmful. Stuart Bowers, the previous autopilot head also gave indications that safety actually was higher on autopilot at autonomy day.

I'm not sure which assertion is "this assertion," but I don't disagree that Teslas on AutoPilot are--by and large--safer than cars without AutoPilot, under the same conditions. I just don't think that Tesla is doing as good a job as it could to make the comparison bases as clear as possible (to Telsa's advantage). I think the real test will be when FSD is rolled out to (some form of) the masses, and we can get closer to an "apples to apples" comparison.
 
As a professional pilot for 45 years, I love this stuff, and can't wait for it to come to my car, but I will never think of it as anything more than workload reduction software, no matter how sophisticated it becomes. I've been too long in cockpits staying prepared for anything and everything that might go wrong. I don't believe I would ever trust a car anymore than I would an aircraft to operate perfectly 100% of the time, I will always have to monitor them.

Well now. You and I think alike. Probably because I was a former Navy pilot LONG AGO. Of course I had no type of automatic systems back then. I'm very cautious about what I let the car do and where. I'll leave experimentation to people with more desire to do such things.
 
  • Like
Reactions: Lindamon
Very low bar in 3 years, just freeway perhaps low speed only. Expect Waymo this year.

I don't like the term driverless unless there's no remote "driver" as well. It's only level 4 if the car can safely pull over or something like that before it phones home. For example, if Cruise has no safety driver but has to phone home for guidance and/or new paths every 5-15 miles in SF, is that really driverless?
 
Most of y'all seem to be confused about level 3 like me before diplomat33 and some others educated me.

With L3, you can define your own operational domain. For example, the following is level 3 (similar to the magical Mercedes level 3):

One lane driving on highway only
Under 25 mph
Requires follow car
Clear lane markings only
Not rainy or snowy weather
5 seconds to take over
If car stops for 10 seconds, driver has 5 seconds to take over

DRIVE PILOT | Daimler
Everything for level 3 is the same for level 4 except the take cover. So only difference is that Level 4 can pull over and Level 3 can't. Seems easy for Tesla to skip L3.
 
Last edited:
I don't like the term driverless unless there's no remote "driver" as well. It's only level 4 if the car can safely pull over or something like that before it phones home. For example, if Cruise has no safety driver but has to phone home for guidance and/or new paths every 5-15 miles in SF, is that really driverless?

According to the SAE, I think the answer is "yes".

Here is what the SAE says on the topic:

ADS-Dedicated Vehicle (ADS-DV)
A vehicle designed to be operated exclusively by a level 4 or level 5 ADS for all trips within its given ODD limitations (if any).

NOTE 2: An ADS-DV is a truly “driverless” vehicle. However, the term “driverless vehicle” is not used herein because it has been, and continues to be, widely misused to refer to any vehicle equipped with a driving automation system, even if that system is not capable of always performing the entire DDT (within given ODD limitations, if any) and thus requires a (human) driver for all or part of a given trip (see Section 7). Instead, this document defines the term “driverless operation,” which encourages specificity in usage to avoid imprecision and misunderstanding.

NOTE 4: ADS-DVs might be operated temporarily by a conventional or remote driver: (1) to manage transient deviations from the ODD, (2) to address a system failure, or (3) while in a marshalling yard before or after being dispatched.

REMOTE DRIVER

A driver who is not seated in a position to manually exercise in-vehicle braking, accelerating, steering, and transmission gear selection input devices (if any) but is able to operate the vehicle.

NOTE 1: A remote driver can include a user who is within the vehicle, within line of sight of the vehicle, or beyond line of sight of the vehicle.

NOTE 2: A remote driver is not the same as a driverless operation dispatcher (see 3.29.4), although a driverless operation dispatcher may become a remote driver if s/he has the means to operate the vehicle remotely.

NOTE 3: A remote driver does not include a person who merely creates driving-relevant conditions that are sensed by, or communicated to, the ADS (e.g., a police officer who announces over a loudspeaker that a particular stop sign should be ignored; another driver who flashes her head lamps to encourage overtaking, or a pedestrian using a dedicated short range communication (DSRC) system to announce her presence).

EXAMPLE 1: A level 2 automated parking feature allows the remote driver to exit the vehicle near an intended parking space and to cause the vehicle to move into the parking space automatically by pressing and holding a special button on the key fob, while s/he is monitoring the driving environment to ensure that no one and nothing enters the vehicle pathway during the parking maneuver. If, during the maneuver, a dog enters the pathway of the vehicle, the remote driver releases the button on the key fob in order to cause the vehicleto stop automatically. (Note that the remote driver in this level 2 example completes the OEDR subtask of the DDT during the parking maneuver.)

EXAMPLE 2: Identical situation to Example 1, except that the remote driver is sitting in the back seat, rather than standing outside the vehicle.

EXAMPLE 3: A level 4 closed campus delivery vehicle that has experienced a DDT performance-relevant system failure, which forced it to resort to a minimal risk condition by parking on the side of a campus roadway, is returned to its designated marshalling yard by a remote driver who is able to operate the vehicle using wireless means.
 
It's most definitely going to be mandatory for L3, and there is the question as to whether L3 will even be allowed to travel at freeway speeds. I also doubt L3 will ever be allowed on cities streets.

Non-geofenced L4 in the strictest sense is a just a dream for now. And even if it was possible with the current hardware, as mentioned before, it will take lots of testing for it to be ready for public use. How long that could be is anyone's guess. Google / Waymo already proved level 4 is very difficult even with capable hardware. That's not even debatable at this point. So, all we're left with in terms of doable autonomy in the near future, is L3 or GF L4 which, IMO, is one hell of an accomplishment if Tesla can pull it off.
 
As a professional pilot for 45 years, I love this stuff, and can't wait for it to come to my car, but I will never think of it as anything more than workload reduction software, no matter how sophisticated it becomes. I've been too long in cockpits staying prepared for anything and everything that might go wrong. I don't believe I would ever trust a car anymore than I would an aircraft to operate perfectly 100% of the time, I will always have to monitor them.

That said, I am very impressed with how far things have come in my life. Autoland in aircraft has been around fro a while in airliners, for landing at places like Heathrow in zero visibility and such, but it is monitored closely by the pilots when in use, and the aircraft has to be certified and pilots trained especially for these types off approaches.

Now Garmin has developed a system for small aircraft, that monitors the pilot for incapacitation, as well as can be initiated by a passenger in case of pilot incapacitation. It can find the nearest airport, contact ATC, change power settings, descend, lower flaps and gear, land, stop, all while avoiding terrain. Really impressive stuff.

You guys might enjoy this if you haven't seen it before, it's the promo video from Garmin, but there are separate demo videos on YouTube as well. Pretty cool stuff.




Didn't I see somewhere that Elon said to the pilot....'...your job will be obsolete.' The pilot seemed speechless.
 
  • Like
Reactions: Lindamon
Didn't I see somewhere that Elon said to the pilot....'...your job will be obsolete.' The pilot seemed speechless.
He also said this:
Screen Shot 2020-10-28 at 12.24.33 PM.png
 
As a professional pilot for 45 years, I love this stuff, and can't wait for it to come to my car, but I will never think of it as anything more than workload reduction software, no matter how sophisticated it becomes. I've been too long in cockpits staying prepared for anything and everything that might go wrong. I don't believe I would ever trust a car anymore than I would an aircraft to operate perfectly 100% of the time, I will always have to monitor them.

That said, I am very impressed with how far things have come in my life. Autoland in aircraft has been around fro a while in airliners, for landing at places like Heathrow in zero visibility and such, but it is monitored closely by the pilots when in use, and the aircraft has to be certified and pilots trained especially for these types off approaches.

Now Garmin has developed a system for small aircraft, that monitors the pilot for incapacitation, as well as can be initiated by a passenger in case of pilot incapacitation. It can find the nearest airport, contact ATC, change power settings, descend, lower flaps and gear, land, stop, all while avoiding terrain. Really impressive stuff.

You guys might enjoy this if you haven't seen it before, it's the promo video from Garmin, but there are separate demo videos on YouTube as well. Pretty cool stuff.


Great post and analogy! I totally agree with you.

I think everyone's wants/desires are different but I personally have limited interest in driverless cars or 100% fail-proof auto-pilot. What I AM interested in is something very similar to what you've discussed above (and discussed very well).

It's about redundancy and assistance for me.
 
NOTE 4: ADS-DVs might be operated temporarily by a conventional or remote driver: (1) to manage transient deviations from the ODD, (2) to address a system failure, or (3) while in a marshalling yard before or after being dispatched.

It's not very clear, and you know I don't like the SAE levels in general.

According to what you posted, the remote driver should only intervene in cases where there is system "failure" (I'm assuming this is related to hardware) or the car finds itself outside of its ODD.

If the remote driver has to create a new path for the car because the software is confused, I don't think that satisfies the level 4 remote driver caveats. This isn't a system failure or outside the ODD. The software developers simply haven't solved the necessary edge cases, so the remote driver has to use his human intuition to navigate the car. So no, the car isn't "driverless" imo.
 
It's not very clear, and you know I don't like the SAE levels in general.

According to what you posted, the remote driver should only intervene in cases where there is system "failure" (I'm assuming this is related to hardware) or the car finds itself outside of its ODD.

If the remote driver has to create a new path for the car because the software is confused, I don't think that satisfies the level 4 remote driver caveats. This isn't a system failure or outside the ODD. The software developers simply haven't solved the necessary edge cases, so the remote driver has to use his human intuition to navigate the car. So no, the car isn't "driverless" imo.
The car is driverless just not all the time. I would say they should just address this in the L4 definition since I bet everyone developing L4 is planning to have remote assistance for the foreseeable future.