Garages, beach huts, woods etc

martinsurrey

Conrad wrote: »

When ALL cars are self drive I think you're correct to say we'll have fewer accidents, however I'm on about the many years of transition where software (literally atomic level atomic interactions) are trying to cope with Human road users all around them.


You imply a lot of trust in software, when you mention power stations et al, but I find as soon as you get into changing dynamic environments and Human interfacing, software can be anything but smart. Every day, even typing this message I encounter issues.


Horizon did a thing on latest AI, and a chief boffin summed it up as 'we still haven't reached the real world intelligence level of the slug'.


Just because software can crunch chess numbers fast, is really meaningless in terms of real world ability and practical intelligence.


Google browser is still trying to flog me holidays I booked months ago, talk about thick, and it's trying to flog me all those I expressly rejected.

http://www.extremetech.com/extreme/181508-googles-self-driving-car-passes-700000-accident-free-miles-can-now-avoid-cyclists-stop-for-trains

The first real prototype has had 700,000 miles with zero at fault accidents, again its just a prototype but that is a very real 700,000 miles of driving on public roads.

This is going no where, MWPT is right, this is coming, and it will be safer, the financial incentive will be vastly more expensive insurance for self drive as they'll cause 99% of the accidents per average mile driven.

If you needed intelligence to drive our roads would be much much quieter.

CLAPTON

Conrad wrote: »

When ALL cars are self drive I think you're correct to say we'll have fewer accidents, however I'm on about the many years of transition where software (literally atomic level atomic interactions) are trying to cope with Human road users all around them.


You imply a lot of trust in software, when you mention power stations et al, but I find as soon as you get into changing dynamic environments and Human interfacing, software can be anything but smart. Every day, even typing this message I encounter issues.


Horizon did a thing on latest AI, and a chief boffin summed it up as 'we still haven't reached the real world intelligence level of the slug'.


Just because software can crunch chess numbers fast, is really meaningless in terms of real world ability and practical intelligence.


Google browser is still trying to flog me holidays I booked months ago, talk about thick, and it's trying to flog me all those I expressly rejected.


EXAMPLE - I was aware someone was driving right up my aaa£se in icy conditions on country roads. I knew if I had to stop quite hard he could hit me. So I made a point of pumping my brake lights as I approached blind bends.
Sure enough and I kid you not, around one of them was a queue of traffic and I had to stop fast (baring in mind the ice), he ploughed straight off the road to avoid me an into a ditch.


Had I not pumped by brakes (heads up) previously he could well have been even closer to me than he was and not had time to avert hitting me.


No way will any software do this.

why won't s/w know it's icy?
or why won't it know some-one is close to your tail?
why don't it know its a bendy road
why can't it deduce that there might be traffic/animals round the next bend?
why can't it flask the break lights?

martinsurrey

Conrad wrote: »

Sure enough and I kid you not, around one of them was a queue of traffic and I had to stop fast (baring in mind the ice), he ploughed straight off the road to avoid me an into a ditch.

Had I not pumped by brakes (heads up) previously he could well have been even closer to me than he was and not had time to avert hitting me.


No way will any software do this.

software would have made it so you wouldn't have had to brake hard to avoid a collision as it would drive as far as it could see given the road conditions.

westernpromise

martinsurrey wrote: »

you still haven't dealt with the fact that with driverless cars you would not be in the position of deciding to crash into the car or lorry...

No, that is precisely my point.

Software will decide who gets to live and die in the scenario I set out. I don't share your confidence that the software writers will replicate my choices but make them faster than I could.

This software is going to be written by the kind of people who gave us Tinder, Windows and Battlecruiser 3000AD. It's likely to be in their economic interests to write the software so that the fewest people possible get killed in accidents without regard for who's actually at fault. To get it legal it's likely to be necessary to show that road deaths fall. What better way to show that than to ensure that in situations where one car must be written off so the other two survive, the one with the fewest occupants is chosen?

In the above case, that's clearly going to be the driver of the legal car. The occupants of the stolen and illegal vehicles get to live. If you think car theft or bald tyres will become impossible, I'd think again. It was blithely assumed that better car keys would reduce car thefts. Instead professional car thieves just burgle the house, or assault the driver, and steal the keys.

There is no reason at all to think your touchingly trusting assumption is going to happen, but quite a few reasons to think something quite a bit more sinister will.

westernpromise

martinsurrey wrote: »

http://www.extremetech.com/extreme/181508-googles-self-driving-car-passes-700000-accident-free-miles-can-now-avoid-cyclists-stop-for-trains

The first real prototype has had 700,000 miles with zero at fault accidents, again its just a prototype but that is a very real 700,000 miles of driving on public roads.

This is going no where, MWPT is right, this is coming, and it will be safer, the financial incentive will be vastly more expensive insurance for self drive as they'll cause 99% of the accidents per average mile driven.

If you needed intelligence to drive our roads would be much much quieter.

The class actions in the US will be a thing to behold when widows and orphans sue the authors of software that decided to kill their partners and parents to spare car thieves.

It'll make VW's dodgy diesels look like small change.

CLAPTON

westernpromise wrote: »

No, that is precisely my point.

Software will decide who gets to live and die in the scenario I set out. I don't share your confidence that the software writers will replicate my choices but make them faster than I could.

This software is going to be written by the kind of people who gave us Tinder, Windows and Battlecruiser 3000AD. It's likely to be in their economic interests to write the software so that the fewest people possible get killed in accidents without regard for who's actually at fault. To get it legal it's likely to be necessary to show that road deaths fall. What better way to show that than to ensure that in situations where one car must be written off so the other two survive, the one with the fewest occupants is chosen?

In the above case, that's clearly going to be the driver of the legal car. The occupants of the stolen and illegal vehicles get to live. If you think car theft or bald tyres will become impossible, I'd think again. It was blithely assumed that better car keys would reduce car thefts. Instead professional car thieves just burgle the house, or assault the driver, and steal the keys.

There is no reason at all to think your touchingly trusting assumption is going to happen, but quite a few reasons to think something quite a bit more sinister will.

but car thefts did fall.....................

westernpromise

CLAPTON wrote: »

but car thefts did fall.....................

And violent assaults and car jackings rose. Progress, eh?

Someone mentioned insurance. I'm glad someone mentioned insurance because insurance companies are going to be the biggest beneficiaries of this. There'll be fewer claims overall and they needn't bother investigating anyone's driving record any more because whether you're 19 or 90 you will be the same risk. Consequently it will be insurance companies dictating the software's judgment subroutines.

If, now, you want hardware under the bonnet that will enable your car to do 150mph, your insurance company will charge you more than if you go for the 1.0L model. If you're 19 they will quote you a figure you simply can't afford.

They'll look at self-driving car software the same way. They'll charge you less if you drive a car fitted with the version of the software that ensures the fewest people die. It will tend to expend you because you're only on £15k a year, so the payout to dependants will be nil and the lost earnings for injury won't be much either.

They'll charge more to the £ million a year software tycoon who has the version that ensures it's always other people who get killed before him. He pays up for this bigtime. If it's a choice between taking out the £15k a year sap or a whole bus queue, the £15k a year sap isn't going to make it. so he's a cheap risk. The tycoon in the Porsche is going to make it so it's the bus queue that gets taken out and has to compensated.

In a straight choice between the poor guy and the rich guy the rich guy will always be spared because the poor guy can't afford the same software.

I'd be the guy in the Porsche but somehow the idea still makes me uneasy for some reason.

antrobus

westernpromise wrote: »

No, that is precisely my point.
....

Your point appears to be to construct ridiculous scenarios that have no basis in reality.

The fact that I might encounter a robot car on the roads sometime in the near future does not worry me at all. The fact that I might encounter a car driven by you does however give me some concern.

Generali

westernpromise wrote: »

No, that is precisely my point.

Software will decide who gets to live and die in the scenario I set out. I don't share your confidence that the software writers will replicate my choices but make them faster than I could.

This software is going to be written by the kind of people who gave us Tinder, Windows and Battlecruiser 3000AD. It's likely to be in their economic interests to write the software so that the fewest people possible get killed in accidents without regard for who's actually at fault. To get it legal it's likely to be necessary to show that road deaths fall. What better way to show that than to ensure that in situations where one car must be written off so the other two survive, the one with the fewest occupants is chosen?

In the above case, that's clearly going to be the driver of the legal car. The occupants of the stolen and illegal vehicles get to live. If you think car theft or bald tyres will become impossible, I'd think again. It was blithely assumed that better car keys would reduce car thefts. Instead professional car thieves just burgle the house, or assault the driver, and steal the keys.

There is no reason at all to think your touchingly trusting assumption is going to happen, but quite a few reasons to think something quite a bit more sinister will.

Given that 90% of collisions are apparently caused by driver error I'll trust the geeks.

kabayiri

I don't think we will be seeing autonomous self-drive beach huts anytime soon.

There just isn't the demand for it.