Autonomous vehicles

[uppi]

If you can get your software onto the cars decision making hardware, you can turn ALL cars into psycho killing machines, not just the one owned by a psycho killer.

In theory, yes. In practice, this is going to be very hard, because you need to create the software that turns the car into psycho killing machines. It is very unlikely that once you have hacked a car you could just sen the command "enable psycho killer mode" and it will behave as such. Instead you would have to teach the car how to behave on a murderous rampage, how to target fleeing pedestrians. To do that you would have to interact with all its sensors and disable all safety mechanisms, some of which may be implemented in hardware so that they cannot be touched by an over the air update. It would be possible to do that, but it would require a significant amount of effort for development and testing. Someone on the terrorist watch list building a test track, where he has cars trying to run over cardboard figures with images of pedestrians on them, would draw serious attention from law enforcement and other agencies, don't you think?

 

[Samson]

In theory, yes. In practice, this is going to be very hard, because you need to create the software that turns the car into psycho killing machines. It is very unlikely that once you have hacked a car you could just sen the command "enable psycho killer mode" and it will behave as such. Instead you would have to teach the car how to behave on a murderous rampage, how to target fleeing pedestrians. To do that you would have to interact with all its sensors and disable all safety mechanisms, some of which may be implemented in hardware so that they cannot be touched by an over the air update. It would be possible to do that, but it would require a significant amount of effort for development and testing. Someone on the terrorist watch list building a test track, where he has cars trying to run over cardboard figures with images of pedestrians on them, would draw serious attention from law enforcement and other agencies, don't you think?

I guess it depends how it is implemented. If it is some human defined decision tree, like if(pedestrian) then (stop) else if(cyclist) then (go round) else if(car) then (follow) else if(empty tarmac) then (drive into) then you could change it to if(vulnerable object) then (drive into). Given that it is likely to be a convolutional neural network that no-one understands it will be a lot more difficult than that, but still it is likely to be much easier to develop an unsafe AI than a safe one.

One that just turned off the brakes and maxed the throttle and was uploaded to a small percentage of cars could well be enough to turn public opinion away from automated cars for a generation, as well as potentially killing more people than 9/11. The fact that this was still fewer than would have died from manual cars in the time since adoption would not make much difference.

 

[civvver]

CAVs (Connected and Autonomous Vehicles) are coming and they are coming sooner than many of us expect. Just this week we had Waymo (ie Google) cars driving around Phoenix with no one in the front of the car. Cars have driven from the east coast to the west coast of the USA without driver intervention. A truck has delivered a big load of beers from the brewery to the depot 120 miles away also with no human intervention (only overseeing).

Two things are going to ‘drive’ this forward: Costs and safety.
Trucking companies are going to save millions in their wage bills, as will taxi firms. Future CAV taxis will be a lot, lot, lot cheaper per mile if you eliminate the driver’s salary and remember these cars will be able to work virtually 24/7. Insurance will be much, much cheaper too because they will cause far fewer accidents. People reckon these ‘uber-type’ taxis will be so cheap and plentiful that many people will simply give up owning a car.
Most cars sit for 90+% of the time doing nothing except depreciate and many will do the maths and ask themselves why should they own a car when they can just hop into a cheap, easily available ‘uber-type’ taxi.

Something like 95% of accidents are caused by humans. About 100 people are killed on the roads of the US every single day (In he UK it takes us about 17 days for 100 deaths). Apparently 35 per cent of teeenage deaths in the US are from road accidents!
CAVs will not be speeding, drunk, drugged up, asleep, texting, just not concentrating etc. etc. Once CAVs are properly up and running I predict statistics are going to be produced that show, something like: For the last 100 deaths on the road, 95 were caused by human error; 4 from other factors like deer, burst tyres etc, and one caused by a not quite perfect AI.
And this will begin to change people’s minds. Indeed it will make people begin to demand that the boy-racers, the 20mph oldies, the road-ragers, the texters, the drunks and the drugged up are all banned from driving. And the only way to do that is to ban all humans from driving.

And the really good news – that one AI accident I referred to above will be analysed and analysed and (hopefully) a patch will stop it ever happening again – in all cars with that particular software, anyway. Just about nothing however will have been learned by the rest of humanity regarding the 95, and the things that caused those 95 accidents will be destined to be repeated over and over again.

The spread of CAVs will start in the cities, where the safety of pedestrians and cyclists will virtually demand ‘driverless only zones’. And it will spread out from there.

I have seen various estimates that there is going to be a massive drop in car ownership in the USA in the coming years as more people realise the cost savings.

I work in the motor trade and it is going to absolutely decimate it. Most motor dealers will disappear because most carsmodules will be sold direct from the manufacturer to the ‘uber-type’ taxi services. Most CAVs will, almost certainly be electric or hydrogen-electric which will require a lot less maintenance too.

I have been concerned about the future of my industry for a while and then this below by Bob Lutz (a former GM boss) is one of the scariest things I have ever read.
Unfortunately I think he is right. Earlier this year I would have said it was all going to happen much later than he reckons. But the CAV stuff that has been going on in the last few months suggest to me he could be right.


http://www.autonews.com/article/201...1109944/bob-lutz:-kiss-the-good-times-goodbye

Common use of driverless cars are coming and coming soon. I told my daughter this weekend that her children, the first will be born in Feb., will probably never drive a car except as a novelty. I suspect that over the next couple of decades human driven cars will be be more and more restricted. In town/city driving will be the first to be restricted and highway, higher speed the last. Human driven cars will need all kinds of special additional software capabilities so that they can mix with the driverless ones safely. My niece lives in DC and she and her husband recently took uber and ended up in a driverless gull winged Tesla. They passed around photos of the uber driver facing them with no hands on the wheel. As soon as we have fleets of driverless taxis for in town errands, car sales will drop and the need for garages decline. Just like we got over our family attachment to horses and mules and shifted to automobile ownership instead, we will give up our love of cars too.

Investment speculating certainly could be profitable if you are willing to take the risks and do the research. NVIDIA would be a great place to start.

It's indeed a drastically changing time, though I'm not sure if 15-20 years is a quick enough time frame. But it'll happen in my life time. Bad news for me is my parent company is a huge automotive supplier. I suppose they could adapt and start building parts for autonomous vehicles. But we can already see some of our sales dipping as companies move away from internal combustion and into ev. Their main products are combustion engine parts.

Semi bad news is the division I work for our business is all software auto diagnostics. We could transition easily into a new market like this, surely av's will need diagnostic software too, but I worry that our relationships with the big autos won't transition. Like are ford, gm and chrysler account for probably 80% of our sales. If some new company pops up and puts one of those out of business will we be able to get some contracts with them? I'm not sure our sales team has it in them right now to see this future.

Good news is like I said I'm in software. Software will be around until it gets advanced enough that ais self program and eliminate my use. But that won't be during my working lifetime. And if ai's ever get that advanced our entire workforce is in serious trouble and we're going to have to move to some sort of social income when robots take all the jobs.


It is frightening what a drastic reduction in autos produced/sales and all that infrastructure will do to the economy, plus removing all the transport industry jobs. The only solace is that in general technology has always created more jobs than it has eliminated. I just worry people who say oh well this is just like when autos replaces horse and buggys, it's not the same thing because we have a global economy now. Horse and buggy guy lost his job to an auto plant down the road, not some conglomerate in china. Where you could once adapt to the changing industry and re-position, you can't if the jobs are all off shored or replaced by robots.

 

[Birdjaguar]

It's indeed a drastically changing time, though I'm not sure if 15-20 years is a quick enough time frame. But it'll happen in my life time. Bad news for me is my parent company is a huge automotive supplier. I suppose they could adapt and start building parts for autonomous vehicles. But we can already see some of our sales dipping as companies move away from internal combustion and into ev. Their main products are combustion engine parts.

Semi bad news is the division I work for our business is all software auto diagnostics. We could transition easily into a new market like this, surely av's will need diagnostic software too, but I worry that our relationships with the big autos won't transition. Like are ford, gm and chrysler account for probably 80% of our sales. If some new company pops up and puts one of those out of business will we be able to get some contracts with them? I'm not sure our sales team has it in them right now to see this future.

Good news is like I said I'm in software. Software will be around until it gets advanced enough that ais self program and eliminate my use. But that won't be during my working lifetime. And if ai's ever get that advanced our entire workforce is in serious trouble and we're going to have to move to some sort of social income when robots take all the jobs.

It is frightening what a drastic reduction in autos produced/sales and all that infrastructure will do to the economy, plus removing all the transport industry jobs. The only solace is that in general technology has always created more jobs than it has eliminated. I just worry people who say oh well this is just like when autos replaces horse and buggys, it's not the same thing because we have a global economy now. Horse and buggy guy lost his job to an auto plant down the road, not some conglomerate in china. Where you could once adapt to the changing industry and re-position, you can't if the jobs are all off shored or replaced by robots.

‘In 50 years, every street in London will be buried under 9 feet of manure.” With this 1894 prediction, the London Times warned that the era’s primary source of transportation energy—the horse—would soon create an environmental crisis.

In New York City, about 100,000 working horses produced roughly 2.5 million pounds of manure a day. Residents were exposed not only to the stench but to biohazards like anthrax. One commentator estimated in 1908 that roughly 20,000 New Yorkers died each year from diseases related to horse waste.

But the deluge of dung predicted by the Times never arrived. Instead the free market solved the problem in roughly 25 years, while creating new goods and industries that transformed society.

The social benefits of self driving cars are too great to be delayed. Now is the time to be planning changes. I think that it won't be long before fuel cell engines will replace electric battery vehicles as the go to technology. That may be the place to investigate. Battery are a transition tech.

 

[rah]

The average/max life of an auto will play an important role. Current cars will always be grandfathered since it will be unrealistic for the government to expect people to just give up the cars they have purchased.

 

[Birdjaguar]

The average/max life of an auto will play an important role. Current cars will always be grandfathered since it will be unrealistic for the government to expect people to just give up the cars they have purchased.

Probably so, but the tide will turn quickly for all of those people who aren't enamored with cars and who see them as mostly a big expense. As soon as a viable alternative for getting around quickly and easily without delay appears, they will join that band wagon.

I love cars and having one, but we would probably drop one easily; then hold out a bit longer before we get rid of the second.

 

[OldBoomer]

I think we will see self-driving vehicles for public transportation long before they ever become commonplace for private use and this would depend on how affordable these self driving vehicles are. It may be more economically viable to keep paying the bus driver who's only making $30,000 per year driving the city bus that's 15 years old and doesn't require much maintenance.

For private use I personally don't see the appeal. You'd put your life in the hands of a computer and of course everyone who's ever worked with electronics knows that they never malfunction. Not to mention that you'd be tracked everywhere you go and just imagine the giant pain in the ass it would be trying to get the stupid thing to turn around and go back to your house because you forgot something. Repairs would also cost a fortune and need to be done at a dealership. No thanks.

 

[Birdjaguar]

I think we will see self-driving vehicles for public transportation long before they ever become commonplace for private use and this would depend on how affordable these self driving vehicles are. It may be more economically viable to keep paying the bus driver who's only making $30,000 per year driving the city bus that's 15 years old and doesn't require much maintenance.

For private use I personally don't see the appeal. You'd put your life in the hands of a computer and of course everyone who's ever worked with electronics knows that they never malfunction. Not to mention that you'd be tracked everywhere you go and just imagine the giant pain in the ass it would be trying to get the stupid thing to turn around and go back to your house because you forgot something. Repairs would also cost a fortune and need to be done at a dealership. No thanks.

Buses and mass transit will continue. Computers will be better drivers than people. To begin, they won't drink, text or run red lights. They won't smoke dope either. You are likely being tracked already. Do you use a smart phone? A credit card? Use the internet at home? To go get a forgotten item, I'm pretty sure you will only have to speak up and tell the car to do so. Isn't that what happens now? If your wife is in the car with you and says "I forgot my purse", what do you do? You turn the car around and go home. All the aging baby boomers are an impediment to change; it's time for us to die off not live longer.

 

[OldBoomer]

Buses and mass transit will continue. Computers will be better drivers than people. To begin, they won't drink, text or run red lights. They won't smoke dope either.

Do you really think there will be no malfunctions that get people killed? There's been countless parts that have malfunctioned in cars that are not self driving which have gotten people killed.

You are likely being tracked already

I know I am. At least when I have my phone's GPS on and everything I've ever saw, searched, or downloaded on the internet; a record of it has been collected and stored in a massive NSA data center in Arizona. The NSA even knows what my favorite genre of porno is.

To go get a forgotten item, I'm pretty sure you will only have to speak up and tell the car to do so. Isn't that what happens now?

No it's not because I just turn around in the nearest driveway and go home. A GPS will most likely reroute you around the nearest block. I can only imagine the dumb crap these things would pull rerouting in the middle of a downtown core with a bunch of one-way streets in heavy traffic. I've been burned by Google maps on countless occasions.

 

[OldBoomer]

Buses and mass transit will continue.

I meant there would be self-driving pubic transit vehicles long before anything private use became mainstream and even then it could still be a long time before public use is economically viable.

 

[Birdjaguar]

Do you really think there will be no malfunctions that get people killed? There's been countless parts that have malfunctioned in cars that are not self driving which have gotten people killed.

I doubt that there will be more system failures than drunk, texting, speeding, stupid, distracted drivers that we have now.

I know I am. At least when I have my phone's GPS on and everything I've ever saw, searched, or downloaded on the internet; a record of it has been collected and stored in a massive NSA data center in Arizona. The NSA even knows what my favorite genre of porno is.

No it's not because I just turn around in the nearest driveway and go home. A GPS will most likely reroute you around the nearest block. I can only imagine the dumb crap these things would pull rerouting in the middle of a downtown core with a bunch of one-way streets in heavy traffic. I've been burned by Google maps on countless occasions.

How many times a month do you forget things and have to go home?

 

[OldBoomer]

I doubt that there will be more system failures than drunk, texting, speeding, stupid, distracted drivers that we have now.

Most of those situations aren't fatal though and a human driver always has the opportunity to go, "oh horsehocky!' and correct, or lessen the damage of dumb actions. A computer just keeps speeding up and smashes into that utility pole because it never occurs to the machine that it's doing something wrong.

How many times a month do you forget things and have to go home?

Enough that it would be a nuisance.

 

[hobbsyoyo]

The technology for self driving cars is developing very quickly. We're counting down the days before this becomes mainstream but I think it will be decades (if ever) before manually-driven cars are banned.

The legislative framework for how this will work (be it insurance, traffic laws, how to program an AI to handle the trolley situation) will fall in place basically as soon as the technology is released into the wild. I'm not saying the legislation will be perfect on the first go but this is the kind of issue that doesn't have a huge partisan aspect to it but has huge economic, social and health implications and thus legislatures will move very quickly to put a framework in place when it is appropriate to do so.

The arguments that somehow AI drivers will be worse than people is silly to be honest to the point where I can't even understand the arguments against AI drivers.

 

[rah]

Most people are being tracked if you use tollways and have an electronic pass. And many highways now have license scanners.
And yeah a lifetime in technology doesn't exactly fill me with confidence. But it will probably perform better than the other distractions that can happen to regular drivers. My biggest concern is when the system has to decide who dies. Whoever loses that one will probably lead to some nasty lawsuits.

 

[hobbsyoyo]

Last I heard the strategy was going to be that the manufacturers were going to carry the liability in these situations. I also don't really see how this is different from people driving cars. It's not like when you're going down the road at 50 miles an hour you have the time to make a rational decision in an impending impact. You may react but there's very little chance it will be the 'correct' course of action. At least a computer will have the time to make a decision based on inputs, which I believe is better than a person taking random action (if any). And in any case, even if the computer can make a decision, that doesn't mean it can meaningfully act on it - momentum is a thing and there are only so many things you can do going 50 miles an hour to avoid a crash that won't end up causing another crash.

 

[rah]

While I generally agree, people expect other people to not be able to make a perfect decision but the expectation for a computer may be different. Someone is going to think that the programmer made the decision. Probably not a concern in the long run but initially it may cause a problem.

 

[hobbsyoyo]

To an extent the programmer will make the decision. To be fair, things get muddy because the resulting fault trees set up to decide what course of action to take will grow beyond human comprehension very quickly. Programmers will end up doing their best effort at programming the fault tree and then run simulations to see how things play out statistically, then tweak the algorithms and try again. That's my guess anyways.

In any case the results will be far better than what a human can manage in the same situation because these simulations can be run millions of times with no one getting hurt and the software update concurrently. You can't really re-program human drivers in the same way. And all of that will happen before the cars are actually released to the public. Once that happens there will be even more tweaking in rapid order to fix missed problems.

 

[rah]

As a past developer and now a director of them, I understand, but fear there are others not as familiar will think it differently.

 

[Manfred Belheim]

To an extent the programmer will make the decision. To be fair, things get muddy because the resulting fault trees set up to decide what course of action to take will grow beyond human comprehension very quickly. Programmers will end up doing their best effort at programming the fault tree and then run simulations to see how things play out statistically, then tweak the algorithms and try again. That's my guess anyways.

In any case the results will be far better than what a human can manage in the same situation because these simulations can be run millions of times with no one getting hurt and the software update concurrently. You can't really re-program human drivers in the same way. And all of that will happen before the cars are actually released to the public. Once that happens there will be even more tweaking in rapid order to fix missed problems.

Well it's going to be a deep learning thing surely, rather than manual programming.

 

[Birdjaguar]

Most of those situations aren't fatal though and a human driver always has the opportunity to go, "oh ****!' and correct, or lessen the damage of dumb actions. A computer just keeps speeding up and smashes into that utility pole because it never occurs to the machine that it's doing something wrong.

You need to study up on how driverless cars respond to problems. So, non fatal accidents where people are crippled, but left alive are OK? Today's WSJ had a really interesting article on the future of e-cars. It was in the third section. You should seek it out.

Enough that it would be a nuisance.

Leaving what you need at home is a different problem. You need to solve that one first, then the "going back" issue goes away.