FSD / Autopilot safety vs liberty

[Dids]

Someday soon... FSD / autopilot will be better than the average driver. It will decrease accidents overall. Should you be allowed to turn it off?

Advertisement

 

[Fabville]

Someday soon... FSD / autopilot will be better than the average driver. It will decrease accidents overall. Should you be allowed to turn it off?

This isn't a direct quote, I'm going off memory on this one, but I remember Elon talking about the Robo Taxi fleet. He compared it to elevator operation back in the day when they used to be manually controlled by human operators. Apparently there were deaths from human error. Sometimes you would have operators that were distracted, impaired, or fatigued (all of which are issues we experience with drivers on the road today). He mentioned one day people won't just want it, but will demand fully autonomous vehicles. He's also mentioned that at some point he expects Tesla will make vehicles without steering wheels, it was a matter of years (not decades) he mentioned when that could start happening. Overall humans are poor drivers, autonomous vehicles will be far superior at some point.

A couple years ago I remember seeing a video that had an animation of how traffic could be improved. It showed cars weaving across an intersection in all directions without stopping. I think autonomous vehicles could do this, but I would have 0% confidence in human drivers. I need to see if I can find that video again...

 

[Fabville]

Here’s an article that has more context from Musk’s comments around elevator operation and FSD. He mentions allowing humans to “operate 2 ton death machines”. He also says human intervention will DECREASE safety.

https://www.zdnet.com/article/elon-...ear-a-human-intervening-will-decrease-safety/

This is a video that shows how traffic can be improved, around the 3min 45s mark it discusses the intersection weave:

 

[Dids]

Here’s an article that has more context from Musk’s comments around elevator operation and FSD. He mentions allowing humans to “operate 2 ton death machines”. He also says human intervention will DECREASE safety.

https://www.zdnet.com/article/elon-...ear-a-human-intervening-will-decrease-safety/

This is a video that shows how traffic can be improved, around the 3min 45s mark it discusses the intersection weave:

I heard that's also true with airplanes. It's the pilots intervention that causes the crash.

 

[lukefrisbee]

My vote:
YOU should not be allowed to drive when FSD becomes a reality.
(Me? I can't figure out why I would want to except one case.... when I want to get there FAST!)
I will so willingly allow the steering wheel to be removed from my cyber, and the front seats faced backwards, and into a bed.
I voted to not be able to turn it off. right now 3 0f the 8 people are thinking intelligently of others and not turning it off.

 

[BillyGee]

I'd be okay with it being on by default, but I still want to be able to drive my car when it needs a human touch or just to launch it on the freeway.

 

[Jhodgesatmb]

Someday soon... FSD / autopilot will be better than the average driver. It will decrease accidents overall. Should you be allowed to turn it off?

I suspect that even when we reach L5 we will still be able to turn it off in most/many situations.

 

[Fabville]

My vote:
YOU should not be allowed to drive when FSD becomes a reality.
(Me? I can't figure out why I would want to except one case.... when I want to get there FAST!)

The Autopilot on my Model X was “not available” several times during a heavy rain storm a couple months ago, it was a quick reminder how much more enjoyable the commute is for me when it’s not all manual. I have a feeling I’ll rely on FSD nearly all the time once it’s available.

 

[Geo]

I don't mind being a passenger, but at times I actually enjoy driving myself.

There are incompetent drivers, competent drivers and then there are talented exciting drivers.

When FSD can drive like Sebastien Loeb, (when he needs to make up time ) I'll be selecting that mode sometimes,
and not because Im running late.
But even then, sometimes I'll want to have a go myself. I'll want to have control myself.

P.S. I'm looking forward to ubiquitous safe FSD, as there will be no reason not to raise the speed limit, when vehicles can communicate with each other and have superhuman performance.

In the name of efficiency, perhaps there will be ramps on freeways that initiate vehicles driving on two wheels, so we can share lanes.
i.e. vehicles forming triangles with wheels in the air leaning on each other

I may let FSD take the wheel on that one . . . . . or I may not . . . . . . . either way I'm definitely taking a selfie !!

 

[VI Tesla]

I agree that for town/city, highway driving, FSD should be left on. But I can think of a few scenarios where it would be nice/a must to be able to disable it, hence I don't see the wheel being removed entirely.

1. Off road driving. For a vehicle designed for the back 40 as well, I would have to assume you could turn it off.
2. FSD also assumes a destination in mind. How about for those Sunday drives or road trip where you just want to explore a bit.
3. pulling into a job site where there is no allocated parking

I guess time will tell.

 

[Kamin]

Tesla's current self driving tech is still all based on visual data. The ever improving algorithms for this will enable FSD to work better than humans in general it lacks the true safety improvements that will come once the cars are networked together and a traffic system helps to coordinate all the data the cars generate. I think that while the software of visual processing will improve steadily some form of lidar and/or radar will be used to augment visual data in truly self driving systems.
I envision that at some point the cars driven by humans will require the addition of a network box to allow the self driving cars to exchange data with the vehicles still driven by people. As more time passes and virtually all vehicles are self driven I see a special licence and equipment being required to actually drive on public roads yourself. The insurance for those people will be a big factor in people deciding the cost effective way to go is not to drive yourself.

 

[scomer]

The Existentialist definition of the good choice is that choice leading to the maximum number of choices. In 2018 Musk warned that AI is more dangerous than nukes, so my hope would be that he would always give humans the option to purchase a vehicle with a steering wheel and the ability to take control of the vehicle.

 

[Dids]

Tesla's current self driving tech is still all based on visual data. The ever improving algorithms for this will enable FSD to work better than humans in general it lacks the true safety improvements that will come once the cars are networked together and a traffic system helps to coordinate all the data the cars generate. I think that while the software of visual processing will improve steadily some form of lidar and/or radar will be used to augment visual data in truly self driving systems.
I envision that at some point the cars driven by humans will require the addition of a network box to allow the self driving cars to exchange data with the vehicles still driven by people. As more time passes and virtually all vehicles are self driven I see a special licence and equipment being required to actually drive on public roads yourself. The insurance for those people will be a big factor in people deciding the cost effective way to go is not to drive yourself.

I hope you are wrong about network box and about lidar. The network would be a common point for failure or attack. I hope if networking between vehicles occurs it is done locally between the vehicles and doesn't require some central control. I also hope lidar will not be needed. It will always fail in a snowstorm, it shouldn't be necessary since a human can drive without it and its just plain ugly.

 

[Sirfun]

Didn't you folks see the movie 2001 space odyssey? Remember Hal? Not good to turn over ALL control. LOL, you get in the vehicle and WHAM all doors lock and you get taken for a ride to the Police Station, warrant we don't need no stinking warrant. I ALWAYS want control of MY vehicle.

 

[rr6013]

Going forward AI will be better trained in most all situations just as ABS braking supercedes human brake control today. Not a single driver challenges the inability to turn-off ABS brake systems on their car. SO incremental implementation, regulation and certification on new AI systems will follow the ABS example.

The question is all but mute for safety systems implemented individually, incrementally and certified by an approved authority. The implication Dids question implies " general" AI (i.e. FSD, HAL, etc...). Artificial general intelligence is a hypothetical intelligent machine learning that has the capacity to understand or learn any intellectual task that a human can. Decades into the future, generalAI and Dids will square off on taking over the whole car, human detrimental reliance and society regulatory mandate over safetyAI.

In the intervening decades, incremental improvements in FSD will begat AI subsystems purpose designed for critical safety componentry to take over systems in cars. I expect subsystems to follow ABS's example. The nexus where human intervention and subsystems cross will blur and either integrate with human behavior (aka ABS) or supercede human interaction and by definition exclude independent human input by design, experience or regulation at some point.

Going forward the argument narrows on evidence based facts, human factors and the points at which AI supercedes.