Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

its not evidence without context and the details of variables I mentioned. for all we know it could've driven around a perimeter loop of an unpopulated airfield clocking those miles.


Automatically Appended Next Post:
I dont think mistrust (not fear) of computers is irrational in any sense.

It's a well-known bias dating back to the days of Frankenstein. Earlier than that even, the story of the Golem is a similar precautionary tale.

The fact you somehow believe that companies like Toyota are falsifying their stats in order to get driverless cars into widespread use where they would instantly, inevitably and very publicly be shown to have been developed unsafely by falsified stats, is prima facie evidence of your irrational mindse ton this topic.

I never said anything about falsifying stats, but the fact is that any statistics published by a group with any sort of incentive invested in the results is going to come under scrutiny.
As I said, I'm happy to read other studies from impartial sources, I just cant find any.

 queen_annes_revenge wrote:
I never said anything about falsifying stats, but the fact is that any statistics published by a group with any sort of incentive invested in the results is going to come under scrutiny.
As I said, I'm happy to read other studies from impartial sources, I just cant find any.

Toyota is an impartial source. A car manufacturer has no incentive to care about whether their vehicles are automated or human-driven, they're still selling the same car. They have incentive to develop automated vehicles as long as those automated vehicles continue to look like a superior product that will inevitably take over from human drivers, but if the results show that human drivers are better than the car manufacturer has no incentive to hide that research. Their self-interested action there is to discontinue all work on automated vehicles, publish the research, and use it as a weapon against any competitor who puts automation in their vehicles.


Automatically Appended Next Post:
In fact, given the cost of developing automated vehicles, the last thing a car manufacturer wants to do is hide research that proves that automation is worse than humans. Proof that automation will fail means they can stop dumping R&D money into it, and that proof would be the best possible outcome from the study. If a car manufacturer is publishing research that automated vehicles are safer than humans it's because they've failed to find a way to avoid making that investment and are forced to acknowledge that AI is the future.

Toyots wasnt the company in question. google was.

But thats debatable regardless. you're operating on the assumption that car manufacturers don't want to advance and wish the status quo to remain the same. I'm fairly sure most car manufacturers are realists and wouldnt want to risk not getting involved in the driverless car bubble, and leave any rewards to their competitors.


Automatically Appended Next Post:
Here we go. a scientific study.

https://orfe.princeton.edu/~alaink/SmartDrivingCars/Papers/RAND_TestingAV_HowManyMiles.pdf

'The results show that autonomous vehicles would have
to be driven hundreds of millions of miles and sometimes
hundreds of billions of miles to demonstrate their reliability in
terms of fatalities and injuries. Under even aggressive testing
assumptions, existing fleets would take tens and sometimes
hundreds of years to drive these miles—an impossible proposition
if the aim is to demonstrate their performance prior to
releasing them on the roads. Only crash performance seems
possible to assess through statistical comparisons of this kind,
but this also may take years. Moreover, as autonomous vehicles
improve, it will require many millions of miles of driving to
statistically verify changes in their performance.'


Automatically Appended Next Post:
https://www.csmonitor.com/Business/In-Gear/2016/1014/How-safe-is-Tesla-Autopilot-A-look-at-the-statistics

 queen_annes_revenge wrote:
you're operating on the assumption that car manufacturers don't want to advance and wish the status quo to remain the same.

It's a correct assumption. R&D costs money, if automated cars aren't going to work then the self-interested action from every car manufacturer is to prove it ASAP and stop sinking money into a loss.

I'm fairly sure most car manufacturers are realists and wouldnt want to risk not getting involved in the driverless car bubble, and leave any rewards to their competitors.

Those rewards only exist if they are proven to be safe enough to be allowed into full-scale use. Proving that they aren't safe enough pops the bubble. If a manufacturer is not publishing proof and popping the bubble it's because no such proof exists and they believe that automated vehicles are going to be in mass use.

existing fleets

Here's the key premise of the study. Why limit it to existing fleets? There are 3 trillion miles per year driven in the US, handing even a small percentage of that number over to automated vehicles (preferably in a ramping-up approach with safety checkpoints for each increase) gets you hundreds of millions very quickly with a low risk level (assuming automated vehicles demonstrate sufficient safety parity in smaller-scale tests to even try). This is not an impossible obstacle.

Queen it seems that your issue has actually shifted from the concept of driverless cars to the question of if the technology is ready today.

That's easy - NO - driverless cars are NOT ready today. They might not be ready tomorrow, or the day after. However the point is that there will come a time when the development of them has advanced so far that they ARE ready. In fact that not only are they ready but that they are better than their human counterparts.

The other important fact is that the day they are ready is likely within a span of time that most of us here will see them on the roads.

Certainly I think that we'll see them take over urban and inter-urban travel. Countryside and offroad might take longer or never come, but certainly I can see it happening for many road only networks. Heck considering how many accidents are the result of driver error just having all cars on highways automated could make a huge saving in lives alone. It might also deal with traffic jams at peak times as the AI could adjust it speed to prevent clogging up choke points along the way. Heck consider the amount of time saved at roundabouts when each car knows where the others are heading and doesn't hae to wait or guess (by the direction of the car) where a car Might be heading.



Heck that's one area where signals are not often used and not always clear and where drivers can get confused (very easily) which lane they should be in to come off a roundabout. They are complicated, esp at peak hours and esp when people are already rushed and want in and out of the experience as fast as possible.




I agree driverless cars are not ready today, they are doing well, but there are clear errors and problems with them right now. That said I can see it happening I can see AI being good enough and better than human drivers very readily.

I'm covering many bases. Personally, I hate the idea of autonomous vehicles. Sentimentality as a safe and competent Driver of many different vehicles, possibly.

But, and as a rational thinker, I do try to separate my personal views from the arguments I'm making here, my main point is that, we as humans are not even close to being ready to say with any certainty, that automation/machines are cleverer/safer/more suitable than humans. I believe that is a flawed hypothesis, and quite insulting to the human race in general, for certain people to write us off so casually against computers.

The future, well I believe that is yet to be seen. the leaps in technology even over as small a scale as the past 20 years suggest that some of the points here could be absolutely become a reality at some point, but I don't think they should be being introduced to the roads for public use yet.
In theory these ideas are great, but we all know that the transfer of those theories to reality rarely run smoothly. There are billions of variables and unconsidered real world elements that will affect that process.

The study above shows that using the current statistics in an attempt to prove the safety value of autonomous vehicles can be refuted.

This quote from the scientific american sums up my feelings pretty well.

'It is true that self-driving cars don’t get tired, angry, frustrated or drunk. But neither can they yet react to uncertain and ambiguous situations with the same skill or anticipation of an attentive human driver, which suggests that perhaps the two still need to work together. Nor do purely automated vehicles possess the foresight to avoid potential peril: They largely drive from moment to moment, rather than thinking ahead to possible events literally down the road.'

Waymo driverless cars racked up their first 10 million miles earlier this year.

https://www.digitaltrends.com/cars/waymo-receives-first-permit-to-test-fully-driverless-cars-in-california/

They are now rolling out a fleet of 500 taxis in Arizona. Assuming each taxi did 20 miles each day, that's another 3.5 million trial miles a year.

It won't take long to rack up 100s of millions of miles of live driving.

 Peregrine wrote:

 queen_annes_revenge wrote:
if they will never be programmed to swerve, then deaths will be inevitable.

And if you allow humans to continue to drive cars deaths will be inevitable. It's a simple matter of "is X > Y". Compare the deaths per year from human drivers to the deaths per year from automated cars, whichever kills fewer people is the correct choice regardless of the details of AI programming or whatever.

That is terrible utilitarian reasoning. Generally fewer deaths is preferable yes, but not all deaths are equal. Morality is not a simple formula.

 Peregrine wrote:

the thing is there are things that are, while not universally accepted, and disputable, but generally accepted. Male expendability https://en.wikipedia.org/wiki/Male_expendability

Garbage idea. It shouldn't be accepted, and should receive nothing but contempt. Life is life, gender is irrelevant.

I am sorry. I am afraid society does not agree with you on that. Gender is not irrelevant, it is a pretty big deal.

And the idea that adult males are more expendable than women or children is found worldwide in all societies. It is likely this stems from the traditional roles that adult males perform in human societies (hunter, warrior etc.), which are fraught with risk. Therefore we have been conditioned to expect adult males to be willing to accept risk and death, and see their deaths as less shocking than those of more vulnerable members of society such as women and especially children.

 queen_annes_revenge wrote:
I really dont understand why more people arent putting up resistance to the increased forcing of automation on society.

Because automation has led to a lot of benefits and conveniences and so far not that many drawbacks. It hasn't really been forced, and it won't need to be forced as long as the benefits continue to outweigh the drawbacks. You seem rather suspicious about technology, but fact is that people just want it. They want convenience and safety, and self-driving cars promise to be able to deliver both.

 Overread wrote:
Queen it seems that your issue has actually shifted from the concept of driverless cars to the question of if the technology is ready today.

That's easy - NO - driverless cars are NOT ready today. They might not be ready tomorrow, or the day after. However the point is that there will come a time when the development of them has advanced so far that they ARE ready. In fact that not only are they ready but that they are better than their human counterparts.

The other important fact is that the day they are ready is likely within a span of time that most of us here will see them on the roads.

I agree, but aside from technical capability and morality, cost is also a big issue that is keeping back self-driving cars, and the issue that may in fact prove to be the most stubborn one. Currently, the cost of a self-driving car is far higher than that of a normal one. If self-driving cars are to replace normal ones, they will need to be the same price or even better, cheaper.

Queen - one thought is that you keep swinging back to the idea of the ideal driver. An experienced, confident, calm, rested driver. The problem is that most people are not driving at their best. They drive whilst late; whilst being nagged at by the passenger; whilst half awake; whilst listening to the audiobook/radio; in strong sunlight after rain without their sunglasses; in winter when its dark after only having driven in the light through summer etc...

I've not even touched on illegal things like mobile phones, drinking, drugs and bad car repair or even driving without licence. Even a perfectly normal person driving within the law can be significant danger.

Another aspect is experience, reactions and reaction times are heavily hinged upon personal experience. A person who has never experienced aqua planing or black ice or a car rushing head on at them on the wrong side of the road - all these are experiences. Some we might all have, others we might never have. The point is that until you experience you don't know how to react or what reactions you can even make.


Look at sports, in sports you practice dodging and catching and tackling and such over and over again. The army does likewise with exercises. Practice over and over, repetition and repetition in order to get the muscles and mind used to certain situations. Driving we don't do that - we practice most ideal situations, but we don't practice how to react in a crash or if someone walks out infront of you. And we don't refresh ourselves every week nor every month - in fact driving is a "one test pass" situation where you won't get retrained EVER unless you are forced too. That in itself means a huge chance for building up bad habits. Driving a little faster; being lazier on corners and drifting into the other lane, using the phone whilst driving etc...




Humans are a huge variable lot and many people only learn to drive good enough to get to work and back alive. WE don't push ourselves to be the best drivers ever. As an ambulance driver you've likely had way more training and refresher training which puts you in a much more confident position than the average person. You know already how to react and drive in a way that other people simply never will.


The idea of the self driving car is that, at present, its a child learning. It's in the early stages of being a beginner. In theory it will never forget, never get lazy or take short cuts. So in theory the more experience it gets the better it gets. So eventually it can reach a point where its experiences outnumber that of the average human. That's before we even add on top the huge number of sensors that it can take information from. The car can read everything about its own conditoin, the road condition and what is around it; and it can process that info in a speed no human could.


Automatically Appended Next Post:

 Iron_Captain wrote:
.

I agree, but aside from technical capability and morality, cost is also a big issue that is keeping back self-driving cars, and the issue that may in fact prove to be the most stubborn one. Currently, the cost of a self-driving car is far higher than that of a normal one. If self-driving cars are to replace normal ones, they will need to be the same price or even better, cheaper.

I agree. But I think there are inherent dangers in trying to solve those issues by giving the human less need for concentration. These cars are clearly still going to require a driver paying attention, and having fewer functions to perform seems to impair this function rather than giving them more chance to concentrate, as shown by the above examples of people watching films, napping etc with their autopilot on. People are going to be on their phones, doing all sorts of things in their cars because they think their autopiloted car will keep them from harm and that they don't need to pay attention to the road

That's the idea, the idea is that the car does the driving totally. I agree that it cannot function any other way. Either the autopilot is ON and the driver is not driving or the autopilot is OFF and the driver is driving.

It cannot work half way on the main roads. I fully agree that it would be dangerous to even try. Even a well trained person being fully attentive is going to find it hard to concentrate and not be distracted when they have nothing to do. An untrained general person with no overhead pressures to pay attention is going to be distracted the whole time, even if just because in doing nothing they won't even know really what to focus on.


That's a given and the human driver should only take over at specific moments. Take the film I Robot with Will Smith. On the main highways of the city the car does the driving without any need for intervention; in fact the driver taking over in that situation is a forced moment and almost totally dangerous and the only time he "needs" to take over is when he's being attacked.

Yet when it came to the countryside he was doing the driving and the car wasn't. Two clear lines when the car is in auto and manual mode.

 Kilkrazy wrote:
It's a well-known bias dating back to the days of Frankenstein. Earlier than that even, the story of the Golem is a similar precautionary tale.

The fact you somehow believe that companies like Toyota are falsifying their stats in order to get driverless cars into widespread use where they would instantly, inevitably and very publicly be shown to have been developed unsafely by falsified stats, is prima facie evidence of your irrational mindse ton this topic.

I know. Kind of like how no car company would falsify stats on safety, quality issues, or fuel efficiency performance to further their sales.


Oh, wait...

Volkswagen were caught.

They will be caught evern quicker if their stats on self-driving car safety are falsified.

I was going to mention vw but the debate had moved on.

 Kilkrazy wrote:
Volkswagen were caught.

They will be caught evern quicker if their stats on self-driving car safety are falsified.

Caught invalidates the attempt? How many companies do you think padded results that DIDN'T get caught, on a daily basis I might add?


No one is denying the benefit of being shuttled around like pharaohs with robots feeding us Doritos while they haul us around, what we deny is the competency of current AI programming to accomplish this wholesale safely and without unforeseen catastrophic programming misinterpretations.

Not many companies running safety critical applications, because it's too easy for the truth to out.

For example, do you think there is lots of false data proving that airplane jet engines don't blow up all the time, but actually they do?

Of course not, because it's a stupid idea.

In the same way, it would be useless to falsify data to show that a driverless car doesn't run into pedestrians 1/10 times, because it very soon would be obvious that it did.

In terms of mis-trust and liability and who may or may not be hiding or fudging results?

Insurance. That's what's stopping them.

Consider that at least in the UK, we can still consider 'proximate cause'. Now in many instances it's a sod to prove. Example there is having a crash because another vehicle refused to yield, or did something stupid, but wasn't themself caught in the accident.

But then, there's concertina accidents. That's where there are multiple rear-end shunts. Car at the front is least likely to be responsible. And the more cars you add, the more liability gets muddied. If it's a three car shunt, dead easy. Ask the front driver how many bumps they felt. One? It's the car at the rear (Newton's Cradle). Two? Probably the car in the middle. First hit you, created a hazard, causing the third car to hit them.

Now, introduce the software for a self-driving car. If that does something stupid, and it turns out (and man, that will out) there was a known flaw in the programming? That's a lot of liability lying at the dev's door. Especially if anybody is injured.

And a last fun fact? Someone killed in an accident is quite cheap as insurance goes. Single payout. Someone left with lasting injuries? Guess who's on the hook for the rest of that person's life?

Aye. And I dont think the car companies have figured that out yet, but once they do...

Another question posed by the hypothesis of future highways featuring only autonomous cars, what about those who drive for pleasure? Enjoy the tinkering and maintenance of a classic vehicle or the flashy mods of an old low rider or hot rod? Will they be forbidden from the roads? There will have to be some sort of transition phase, but once that's complete will no human controlled vehicles be allowed at all? That itself takes me back to emergency response drivers. I'm going to throw it out there and say that an automated system will never be able to drive to roadcraft standard. So there will have to still be manually controlled vehicles on the roads.

I've considered that myself. I think what will happen is that such people will go to private car rallies, such as Beaulieu Festival of Speed, where they will be able to drive together in unregulated conditions, separate from the public highway.

The emergency vehicles are another thing. I don't see a problem with human piloted ambulances. They will work fine in conditions where automatic cars are guaranteed to get out of the way in good time. There aren't many accidents involving ambulances anyway.

Police pursuit driving is another thing, of course, but it's hard to see criminals making getaways in autonomic cars which simply stop when ordered by the police.

Of course, part of the logic of autonomous vehicles is that there will be a lot fewer vehicles on the roads at all, which in itself makes everything easier.

Possibly. I can't see the Vatos in East LA being happy taking their low riders to private circuits though.

Chances are once the technology is accepted there will be loads of ways to personalise your car to your own standards. It might mean that people can't do it at home, but then again cars are going that way anyway with a lot of built in computers being very hard to deal with even for many local mechanics.

And yes those who want to drive it themselves might be restricted to country roads or rally events. It's a loss, but at the same time I think if it made the roads a lot safer and more efficient many would accept that loss.



One real barrier might be price. After insurance the next hurdle is getting everyone onboard with the idea to make it work. The problem there is that a new self driving car is not going to be cheap, even if its price is marked down and its mass produced its still going to be a lot of money for people, even more so if their old cars are then devaluing fast because they are not worth resale.
We are already seeing that with cars right now as governments put more pressure on people adapting to electric. It's not that people "don't want to" as much as many just can't afford the cost of a brand new car. There's plenty of 2nd hand dealers who sell cars for dirt cheap and often get then back again when they break, swap over and sell another cheap car etc....

That's why I think we'll see buses and the like take up the tech first. Heck long distance lorry drivers might well take up the tech and then end up with a lorry driver being more of an overseer position for the delivery and possibly driving at any depo that hasn't got a selfdriving guide system setup yet.

 queen_annes_revenge wrote:
Another question posed by the hypothesis of future highways featuring only autonomous cars, what about those who drive for pleasure? Enjoy the tinkering and maintenance of a classic vehicle or the flashy mods of an old low rider or hot rod? Will they be forbidden from the roads? There will have to be some sort of transition phase, but once that's complete will no human controlled vehicles be allowed at all?

Most likely, yes they would be banned from public roads. Private tracks would still exist, and probably grow in number to accommodate the hobby. It's just like any other dangerous hobby, it doesn't matter if you really enjoy guns you still have to take them to a gun range and can't go around shooting stuff in the middle of a major city.

That itself takes me back to emergency response drivers. I'm going to throw it out there and say that an automated system will never be able to drive to roadcraft standard. So there will have to still be manually controlled vehicles on the roads.

First of all, that's a lot of human ego in that claim that they'll never be able to drive to your standard and not a lot of evidence to support it. Second, those few vehicles aren't really relevant. Even if you allow a small number of highly trained professional drivers (perhaps treated like airline pilots are now, in terms of training and experience requirements) to drive emergency vehicles you're still talking about effectively total automation. And that's fine to leave that small exception if it is necessary, professional emergency drivers aren't the accident risk we are worrying about.

Waymo's trial is a taxi service.

Well, it remains to be seen. I believe that there are some things that a computer will never be able to do, or even learn if we follow that folly of allowing them to learn. You praise computer's like some sort of super deity, yet the human brain is so complex that even after all these years of scientific study we know basically nothing about it, despite, ironically, the creation and use of computers.


Automatically Appended Next Post:
You will never be able to code a computer with human instinct.

 queen_annes_revenge wrote:
You will never be able to code a computer with human instinct.

But why do we need to? In the context of automated vehicles "instinct" is just a crude approximation of things like solving the physics problem of how to avoid a collision. The automated vehicle doesn't need instinct because it just makes the correct choice.

No it's not. A blue light driver processes cues from the direct front of the vehicle, to the very visual limit of the road, the autonomous car looks at the vehicle directly in front, rear and sides with sensors and moves where it 'thinks' is right or follows. It can't position itself in the road based on forward observation and judgement of the environment.


Automatically Appended Next Post:
Computers making the right choice, because that happens often... Your borderline pathological worship of something we only invented 70 years ago is very concerning. I'd go so far as to say that of any technical human invention, the computer is probably the most prone to flaws and malfunction.

 queen_annes_revenge wrote:
No it's not. A blue light driver processes cues from the direct front of the vehicle, to the very visual limit of the road, the autonomous car looks at the vehicle directly in front, rear and sides with sensors and moves where it 'thinks' is right or follows. It can't position itself in the road based on forward observation and judgement of the environment.

Uh, what? An autonomous car can look anywhere you feel like putting a sensor. In fact, the autonomous vehicle's sensors are going to be far superior to human vision because it can process everything in full 360* coverage while the human driver is limited to their head's range of motion, obstructions from the vehicle's frame, and the inability to look in multiple directions simultaneously. I have no idea why you would suggest that it can't scan a wider range or position itself in the road based on observation, other than willful ignorance of the subject.

Computers making the right choice, because that happens often... Your borderline pathological worship of something we only invented 70 years ago is very concerning. I'd go so far as to say that of any technical human invention, the computer is probably the most prone to flaws and malfunction.

Computers do often make the right choice. The fact that you're writing this on a computer is proof of it. The only pathological worship here is your bizarre faith in the superiority of humans against all evidence otherwise.