Lamenting the decline of the car

[ssiguy]

I have always thought that self-driving cars were much ado about nothing and I still do.

Even if they start to perfect the technology in all climatic conditions, the legal barriers are too great to cross. If someone is in an accident that their vehicle causes and yet the "driver" wasnt driving then who is responsible.........if you charge the "driver" that is akin to currently blaming the passenger for a driver error. Good luck getting the automakers to take responsibility. Will people even need to be able to drive or get a license for that matter.

[someone123]

Originally Posted by ssiguy

Even if they start to perfect the technology in all climatic conditions, the legal barriers are too great to cross. If someone is in an accident that their vehicle causes and yet the "driver" wasnt driving then who is responsible.........if you charge the "driver" that is akin to currently blaming the passenger for a driver error. Good luck getting the automakers to take responsibility. Will people even need to be able to drive or get a license for that matter.

If the rate of accidents is lower without human drivers it is rational to switch to automated cars, all else being equal. For insurance purposes and assessing risk it does not really matter if there is a human driver.

The self-driving vehicle debate suffers from being abstract, with the pro-self-driving camp tending to wave their hands and assume all of the problems will obviously be solved and the anti-self-driving camp assuming that the question of adoption comes down to trolley problem type scenarios or philosophical debates around agency and culpability. In reality I suspect it all comes down to costs and reliability. If the technology becomes reliable enough it will be adopted, and until then it won't. Historically AI research has been prone to boom and bust cycles as new breakthroughs in approach or hardware create new unknown ceilings that are hit, and certain niche applications are developed while other basic and obvious applications remain out of reach.

[Architype]

Originally Posted by someone123

If the rate of accidents is lower without human drivers it is rational to switch to automated cars, all else being equal. For insurance purposes and assessing risk it does not really matter if there is a human driver.

The self-driving vehicle debate suffers from being abstract, with the pro-self-driving camp tending to wave their hands and assume all of the problems will obviously be solved and the anti-self-driving camp assuming that the question of adoption comes down to trolley problem type scenarios or philosophical debates around agency and culpability. In reality I suspect it all comes down to costs and reliability. If the technology becomes reliable enough it will be adopted, and until then it won't. Historically AI research has been prone to boom and bust cycles as new breakthroughs in approach or hardware create new unknown ceilings that are hit, and certain niche applications are developed while other basic and obvious applications remain out of reach.

You would have to assume that logic will prevail, and that there will be no further repercussions from the implementation and our further dependence on technology. I think it will eventually adapt, but that doesn't account for the obstinance of human nature, and for example, my own preference of a manual shift over automatic transmission.

[someone123]

Originally Posted by Architype

You would have to assume that logic will prevail, and that there will be no further repercussions from the implementation and our further dependence on technology. I think it will eventually adapt, but that doesn't account for the obstinance of human nature, and for example, my own preference of a manual shift over automatic transmission.

I think a lot of this depends on what the trade-offs really are, and people are often more rational than they get credit for (with pro-technology people or salesmen arguing that others are irrational for not switching to or buying some newfangled thing that provides a nebulous payoff when you look at the full balance sheet). Some people prefer manual transmission but it's pretty low stakes. Do you think they'd pay an extra, say, $2,000 a year to be able to keep manual transmission? My understanding is manuals are a particularly bad example because they have historically been the cheaper option.

It's possible that automated cars will advance to the point where they are significantly less likely to cause accidents than human drivers, so there will be a significant incentive to switch. Right now there is no financial incentive because the self-driving cars are not even operating without back-up drivers in them, and the debates are about hypothetical future technology.

[Architype]

^ You also have to assume that a high level of quality will exist in these vehicles, some of which may come from less than reputable sources where the reliability and consistency of technology is not assured.

[goodgrowth]

I could see there be an insurance premium calculated for each vehicle manufacturer based on accident rate for their self-driving modes. Theoretically that would be more standardized than individual rates.

Maybe the company could track hours driven in either mode on a car and calculate a premium dynamically each month?

[someone123]

Originally Posted by Architype

^ You also have to assume that a high level of quality will exist in these vehicles, some of which may come from less than reputable sources where the reliability and consistency of technology is not assured.

Accidents are public knowledge. If Tesla lies and says their self-driving feature is reliable when it's not it will cause a higher rate of accidents that will be easy to detect. I think safety is actually much harder to fake than, say, emissions.

Right now one of the ways the self-driving cars are being measured for progress in the US is driver interventions per mile. The accidents also get reported.

[logan5]

We are at Atari 2600 level right now. Anybody who thinks this is "much ado about nothing" is a fool. Such a game changer on so many levels.

[acottawa]

Originally Posted by logan5

We are at Atari 2600 level right now. Anybody who thinks this is "much ado about nothing" is a fool. Such a game changer on so many levels.

We could stay at Atari 2600 levels for decades or indefinitely. Lots of technologies never proved feasible for large scale implementation despite being theoretical game changers (supersonic passenger jets, super jumbo jets, rocket packs, flying cars, nuclear powered whatever, domestic robots).

[MonctonRad]

Originally Posted by acottawa

We could stay at Atari 2600 levels for decades or indefinitely. Lots of technologies never proved feasible for large scale implementation despite being theoretical game changers

Agreed. There should always be a capacity for human decision making when it comes to the operation of a personal motor vehicle.

A good example is in conflict resolution in a critical incident situation, for example a no win situation where the three choices are:

1- head on collision with a Mack truck
2- swerving to the left and hitting a small Toyota SUV containing a young family
3 - swerving to the right to and risking hitting a 90 year old lady on the sidewalk.

Which is the proper decision? Doesn't this constitute a personal moral choice? Do you want an AI performing the calculus and deciding on it's own what your choice should be? Maybe there's a fourth choice that the manufacturer had not considered during their programming?

This is a slippery slope. I would certainly contest any choice made on my behalf by an AI when the lawsuits begin to fly. I would only accept liability if the choice made was mine and mine alone.........

[Mister F]

Originally Posted by whatnext

There's a whole bunc of "intuition" that goes into driving in bad conditions, along with skills learned over time. It's one reason I've awlays said occasional rideshare drivers are more dangerous than regular driver.

Here's an article on autonomous driving and poor weather:
[I]
Snow and Ice Pose a Vexing Obstacle for Self-Driving Cars
[/url]

Snow and ice pose a vexing obstacle for human drivers too, intuition or not.

[logan5]

Originally Posted by acottawa

We could stay at Atari 2600 levels for decades or indefinitely. Lots of technologies never proved feasible for large scale implementation despite being theoretical game changers (supersonic passenger jets, super jumbo jets, rocket packs, flying cars, nuclear powered whatever, domestic robots).

The benefits are far too great for this to not progress. And these aren't rocket packs or flying cars, this is something that is real. Waymo is now at the stage where they don't even use a safety driver...

• Video Link

[milomilo]

I would bet money that in 5 years time autonomous vehicles are still not in any way mainstream - as I have said before we are moving backwards, as 5 years ago everybody thought they were just around the corner, and now few have that confidence.

I used to think it was going to be mostly regulation and adaptation that would hold us back, but now I'm not even sure the technology is going to be as easy as it appeared. Turns out it wasn't as easy as they made out.

[lio45]

Originally Posted by MonctonRad

Which is the proper decision? Doesn't this constitute a personal moral choice? Do you want an AI performing the calculus and deciding on it's own what your choice should be? Maybe there's a fourth choice that the manufacturer had not considered during their programming?

This is a slippery slope. I would certainly contest any choice made on my behalf by an AI when the lawsuits begin to fly. I would only accept liability if the choice made was mine and mine alone.........

What's likely at that point is that AIs will be insured by the companies that designed them. So if one is available to drive your car (included in the price that you paid, including liability insurance for decisions made by it), you can still choose to eschew it and drive yourself, but then you will have to pay (likely a really high premium) to get human driver liability insurance.

I'm almost certain that'll be the way it'll be.

[logan5]

Originally Posted by milomilo

I would bet money that in 5 years time autonomous vehicles are still not in any way mainstream - as I have said before we are moving backwards, as 5 years ago everybody thought they were just around the corner, and now few have that confidence.

I used to think it was going to be mostly regulation and adaptation that would hold us back, but now I'm not even sure the technology is going to be as easy as it appeared. Turns out it wasn't as easy as they made out.

Maybe more impressive than the Waymo video is this video from Zoox (Amazon). 1 hour (at 2x speed), of driving through the densest parts of San Francisco with pedestrians and traffic everywhere, with unprotected left turns in heavy traffic... all in autonomous mode.

Does it really look like we are moving backwards?

• Video Link

[esquire]

Originally Posted by MonctonRad

Agreed. There should always be a capacity for human decision making when it comes to the operation of a personal motor vehicle.

A good example is in conflict resolution in a critical incident situation, for example a no win situation where the three choices are:

1- head on collision with a Mack truck
2- swerving to the left and hitting a small Toyota SUV containing a young family
3 - swerving to the right to and risking hitting a 90 year old lady on the sidewalk.

Which is the proper decision? Doesn't this constitute a personal moral choice? Do you want an AI performing the calculus and deciding on it's own what your choice should be? Maybe there's a fourth choice that the manufacturer had not considered during their programming?

This is a slippery slope. I would certainly contest any choice made on my behalf by an AI when the lawsuits begin to fly. I would only accept liability if the choice made was mine and mine alone.........

How ethically appropriate of a decision do you think you could make in the split second you have between becoming aware of the situation and colliding with someone/something?!

[MonctonRad]

Originally Posted by esquire

How ethically appropriate of a decision do you think you could make in the split second you have between becoming aware of the situation and colliding with someone/something?!

At least it would be my decision.

Have you ever seen the movie I Robot with Will Smith.

His character was forever haunted because of an AI decision during a traffic accident based on probabilities. The AI made a decision he could not reconcile with. I can definitely see this happening...……...

[drew]

Originally Posted by MonctonRad

Agreed. There should always be a capacity for human decision making when it comes to the operation of a personal motor vehicle.

A good example is in conflict resolution in a critical incident situation, for example a no win situation where the three choices are:

1- head on collision with a Mack truck
2- swerving to the left and hitting a small Toyota SUV containing a young family
3 - swerving to the right to and risking hitting a 90 year old lady on the sidewalk.

Which is the proper decision? Doesn't this constitute a personal moral choice? Do you want an AI performing the calculus and deciding on it's own what your choice should be? Maybe there's a fourth choice that the manufacturer had not considered during their programming?

This is a slippery slope. I would certainly contest any choice made on my behalf by an AI when the lawsuits begin to fly. I would only accept liability if the choice made was mine and mine alone.........

I think no matter what, statistically speaking, you will almost always swerve to the right - just by instinct.

In an accident such as that, you don't really have enough time to consider the implications after the swerve. You first attempt to avoid the head-on collision, and from there, it's anyone's guess.

I have been in plenty of accidents, and you don't have time to hit the horn, let alone decide who you are or are not going to hit.

[whatnext]

Originally Posted by MonctonRad

At least it would be my decision.

Have you ever seen the movie I Robot with Will Smith.

His character was forever haunted because of an AI decision during a traffic accident based on probabilities. The AI made a decision he could not reconcile with. I can definitely see this happening...……...

Absolutely. Your car is going to have to choose who to kill in an accident situation.

[milomilo]

Originally Posted by logan5

Maybe more impressive than the Waymo video is this video from Zoox (Amazon). 1 hour (at 2x speed), of driving through the densest parts of San Francisco with pedestrians and traffic everywhere, with unprotected left turns in heavy traffic... all in autonomous mode.

Does it really look like we are moving backwards?

What I mean by moving backwards is that despite huge advances in technology, the predictions for when this will become mainstream are further in the future now than they were five years ago. If you listened to people back then, we'd have them available by now. The obvious lies they told about capabilities have been toned down now too, thankfully.

That's not to say the technology isn't amazing, and if I was king of Canada I'd happily set down a decree banning incompetent humans from driving cars - they're objectively terrible at it. But the real world is much more complex, and the technology evidently is much harder to get from 90% to 100% than it was from 10% to 90%.