Ed Self-Driving Cars: Pros, Cons, and Predictions

[mumblethrax]

The autonomous vehicles in use today are already much safer than human drivers, so technically we are past 100%.

This is not a view supported by the evidence.

 

[arthwollipot]

That long?

I had to look it up, but apparently, yeah.

 

[Darat]

And that's just the tip of the iceberg.



The autonomous vehicles in use today are already much safer than human drivers, so technically we are past 100%. The biggest hurdle now is public acceptance.

Recently a health insurance CEO was murdered, and many argued this was a good thing because by denying coverage the company was killing far more. I look forward to those people advocating the same for anyone spreading FUD about this lifesaving vehicle technology.

Are they better? How has that been determined?

 

[acbytesla]

This is not a view supported by the evidence.

That's debatable. Evidence on this is extremely difficult to put together. There is evidence suggesting autonomous vehicles are safer and evidence that suggest they are not yet. Varying conditions and other variables make a definitive conclusion tough. But if it isn't yet, it is damn close.

 

[mumblethrax]

That's debatable. Evidence on this is extremely difficult to put together. There is evidence suggesting autonomous vehicles are safer and evidence that suggest they are not yet. Varying conditions and other variables make a definitive conclusion tough. But if it isn't yet, it is damn close.

Which is why the evidence doesn't support this view. It does support cherry-picking, however.

 

[acbytesla]

Study Finds Self-Driving Cars Are Actually Safer Than Humans in Many (But Not All) Situations

Autonomous cars are understandably held to high safety standards—but it's sometimes forgotten the true baseline is the often dangerous driving of humans.

singularityhub.com

A matched case-control analysis of autonomous vs human-driven vehicle accidents - Nature Communications

Through a matched case-control analysis this study reveals accident risk disparities between autonomous and human-driven vehicles. It suggests that accidents of vehicles equipped with Advanced Driving Systems generally have lower occurrence chance than human-driven ones in most scenarios.

www.nature.com

https://www.iotworldtoday.com/transportation-logistics/self-driving-cars-safer-than-humans-except-in-turns-study-finds

The studies suggest they are. But with caveats.

 

[mumblethrax]

The studies suggest they are. But with caveats.

I would not characterize the conclusions of either of those studies as supporting the view that autonomous vehicles are much safer than human drivers. They both reflect the conventional wisdom that the performance of autonomous vehicles depends on driving conditions. Autonomous vehicles are good at easy mode driving, bad at dealing with edge cases (which includes things like "rain").

I've had some white-knuckle drives in upstate New York in winter where I would have been happy to turn over control to a self-driving system I could trust, but those are precisely the conditions where I wouldn't.

 

[Dabop]

A self driving car would definitely struggle around my place- the (dirt) road looks identical to the dirt bushland around it just (usually) less trees on it lol... (not always though- after the wet season, it isnt unusual to see saplings spring up in the middle of the road until they are driven over enough times to kill them!!!)

And teaching a driverless car to deliberately drive off the edge of the tarred road when I get to it would be 'interesting'- the 'main road' into town is literally only one lane wide... as in a single lane of tar, when you meet oncoming traffic, you both drop the left hand wheels onto the dirt to get past each other (except trucks and road-trains- they stay on the tar and the car goes entirely on the dirt)- safer that way as no 'trailer sway' on the rear trailer/s and no rocks kicked up by the many tyres....

My car trailer is literally almost the entire width of the tar...


Two little cars 'might' just squeeze past each other- with their mirrors almost touching- at a combined closing speed of 200kmh- its safer to drop the wheels off into the gravel...

 

[acbytesla]

I'm not saying they are there yet. But close. I have been in autonomous vehicles on the freeway and in city streets. This obviously wasn't in every possible condition. It's pretty damn impressive.

 

[mumblethrax]

I'm not saying they are there yet. But close. I have been in autonomous vehicles on the freeway and in city streets. This obviously wasn't in every possible condition. It's pretty damn impressive.

I agree it's impressive, but I'm specifically responding to the claim that they're already much safer than human drivers. That's the view that I don't think is supported by the evidence.

 

[Gord_in_Toronto]

The problem is not so much a problem of safety but the perception of safety.

Every time a self-driving car is in an accident, some people will say, "See. Told you so. They are unsafe".

If/when, things reach the point that there are demonstrably, statistically fewer accidents per whatever metric you measure, it will make no difference to such people.

 

[mumblethrax]

The problem is not so much a problem of safety but the perception of safety.

Every time a self-driving car is in an accident, some people will say, "See. Told you so. They are unsafe".

If/when, things reach the point that there are demonstrably, statistically fewer accidents per whatever metric you measure, it will make no difference to such people.

Which means that to gain widespread acceptance, they need to not only be better than human drivers in basically all conditions, but to not introduce errors that humans wouldn't make.

Which is why I think this task is more difficult than most people realize.

 

[acbytesla]

I agree it's impressive, but I'm specifically responding to the claim that they're already much safer than human drivers. That's the view that I don't think is supported by the evidence.

I think they are pretty damn close. If Musk wasn't a turd, I would have bought a Tesla five years ago. And a huge selling point is the autonomous driving. That ability to let the vehicle to take the wheel on long drives is awesome.

 

[Gord_in_Toronto]

Sometime about 1960 the Chairman of BOAC was asked whether he would fly on a completely automated passenger plane flight (from takeoff to landing) and he replied he would if such had been demonstrated safer than a piloted flight.

A good answer then and a good answer now.

(I remember reading this at the time in either Flight or Aeroplane but cannot provide a reference other than my memory)

 

[mumblethrax]

I think they are pretty damn close. If Musk wasn't a turd, I would have bought a Tesla five years ago. And a huge selling point is the autonomous driving. That ability to let the vehicle to take the wheel on long drives is awesome.

We aren't close to autonomous vehicles being much better.

I wouldn't buy a Tesla, because I don't think they're particularly good cars, and because of boneheaded moves like removing/forgoing ultrasonics and LIDAR. And yeah, because Musk is a dildo. Also, because I can get by with public transit, where somebody else already takes the wheel. The last fatality due to a collision on the NYC subway was in 1995.

 

[mumblethrax]

Sometime about 1960 the Chairman of BOAC was asked whether he would fly on a completely automated passenger plane flight (from takeoff to landing) and he replied he would if such had been demonstrated safer than a piloted flight.

A good answer then and a good answer now.

(I remember reading this at the time in either Flight or Aeroplane but cannot provide a reference other than my memory)

It's a rational answer. It's not an answer that will persuade people who don't have entirely rational reasons for opposing autonomous vehicles.

I mean, most of us here know the reasons why nuclear power is preferable to coal, but that has not led ineluctably to the success of nuclear power.

You ◊◊◊◊ one goat....

 

[acbytesla]

The problem is not so much a problem of safety but the perception of safety.

Every time a self-driving car is in an accident, some people will say, "See. Told you so. They are unsafe".

If/when, things reach the point that there are demonstrably, statistically fewer accidents per whatever metric you measure, it will make no difference to such people.

It's not a question of if.

 

[Gord_in_Toronto]

It's a rational answer. It's not an answer that will persuade people who don't have entirely rational reasons for opposing autonomous vehicles.

I mean, most of us here know the reasons why nuclear power is preferable to coal, but that has not led ineluctably to the success of nuclear power.

You ◊◊◊◊ one goat....

And forever after you are known as the Goat ◊◊◊◊◊◊

 

[dirtywick]

Which means that to gain widespread acceptance, they need to not only be better than human drivers in basically all conditions, but to not introduce errors that humans wouldn't make.

Which is why I think this task is more difficult than most people realize.

i agree, and i also think they need to be held to a higher degree of liability in an accident that causes some kind of injury or death. you can't ever get every single person to make good driving decisions, it's impossible. an ai program you can, so to a much greater degree accidents with self driving are much, much more preventable. barely better than the dumbest, most reckless people on earth isn't good enough imo

 

[theprestige]

I could see SDCs being safer in usual conditions, less safe in unusual conditions, and statistically safer overall.