Autonomous Land Rovers by end of 2017

[CwazyWabbit]

I reckon the mainstream cars that us normal people can afford will start to come with more intelligent cruise control, then we'll end up with cars that are driverless on motorways and fast A roads. After that they'll gradually add capabilities for more complex situations where the option of killing larger numbers of people (especially pedestrians with a death wish) are more plentiful.

It won't be a day one release for a consumer car to be driverless in all situations, those being tested are pretty much concept cars aren't they?

I'm waiting for riderless motorbikes ......

[GW8IZR]

I drove a Hyundai i40 hire car for a week or so, it had lane control and distance sensing cruise control.

Now I admit I didn't know it was fitted or turned on ( I just pressed a  few buttons while bored ) and I thought there was something wrong with the steering at first as it sort of fought me to stay in the middle of the lane as against me wandering about.

It wasn't perfect and it was better at night time for some reason but I drove it hundreds of miles just occasionally correcting it, overiding it to overtake and change lane on the motorway etc.. it basically pretty much drove itself without too much help.

Now I know that's not an autonomous car nor is it intended to be but it shows the technology is coming.

[geoffbeaumont]

14 hours ago, mad_pete said:

How often does that really come up in driving ? If you are near a bus queue it's a 30, passenger should survive that. Other than that it's so rare how much is it going to matter based on the fact a human will randomly pick on or other or both. 

It is, however, a good example of the ethical questions that autonomous vehicles are throwing up - because these value judgements need to be made in order to programme the vehicles (they will ultimately do what they've been programmed to do - autonomy in IT/robotics doesn't imply free will, just the ability to operate to a predefined set of rules without human supervision). Currently, if you manage to get into a situation while driving where you can't avoid a crash but have a choice of outcomes, you'll make a decision based on your ethical/philosophical framework (everyone has one - basically how you view the world) that aims for the best possible outcome in the circumstances. Autonomous cars must do exactly the same - the controversy comes from two things:

1. We don't like to talk about a lot of these value judgements (e.g. you're going to hit a white woman or an Asian woman - which do you choose? You probably do have a bias - for all sorts of reasons which could include attempting to avoid appearing to have a bias... - but in many parts of society it's not socially and/or legally acceptable to differentiate). But when programming the computers which control an autonomous vehicle, those choices are explicit and public.
2. You may not agree with the priorities your vehicle (or, for that matter, someone else's) has been programmed with. That partly stems from 1 (legislation and corporate decisions are likely to insist on equality in some cases where you wouldn't give it - even if you wouldn't admit that to anyone else - or prioritise one group over another where you wouldn't). A good example - the vehicle may be programmed to put equal value on the lives of all children. But one of the kids that's just run into the road is your son...

Of course, you could argue that forcing the decisions to be made up front reduces hypocrisy - the vehicles are likely to be programmed with whatever priorities have been deemed most socially acceptable (i.e. what the majority of people say their priorities are), rather than what we really think but don't admit to.

[Bowie69]

So in other words, totalitarianism is invading our world via computer programming? 

Oh how I look forward to the future(!)

 

[geoffbeaumont]

4 minutes ago, Bowie69 said:

So in other words, totalitarianism is invading our world via computer programming? 

Oh how I look forward to the future(!)

You'll do as you're told.  Your non-compliant views are being monitored and corrective action will be taken if necessary...

[Davo]

8 minutes ago, Bowie69 said:

So in other words, totalitarianism is invading our world via computer programming? 

Oh how I look forward to the future(!)

 

Yes, me too (not). Every time I see an article about some new robot meant to save the world I simply have an urge to destroy it in the most violent manner possible. (It: "Hello sir, I am the Helpbot 2000." Me: "Oh good. Please take this hand grenade.") But then I feel that way about drones. 

[Cynic-al]

I wonder if they'll come with an Isis setting, park outside busy place and blow up? War would become like a computer game. 

I do think people get their priorities wrong with these things, yesterday I was reading an article about a £12,000 robotic woman, she was programmed to do two things and one of them was talk. I mean what sort of a person makes talk their second biggest priority in a robotic woman? You could've gone for fetch beer, cut the grass, cook a curry, put the loo roll in the freezer for after the curry...

[CwazyWabbit]

Geoff, perhaps the simple solution is to have cars that randomly choose who dies in one of these no win situations? Then there is no ethical argument and purely a burden to prove that the PRNG (pseudo random number generator) has no bias, this can be analysed mathematically.

[geoffbeaumont]

Ah, but at what point do you become random? We'd probably all agree that it's better to run over a cardboard box than a human. Dog or human? You've already got some people arguing for the dog to be on at least an equal footing at that point. Adult or child? 8 year old child versus 2 year old child?

There's an article about some of the work that's going on around this gone up on The Register today:

http://www.theregister.co.uk/2017/07/05/selfdriving_algorithms_that_make_ethical_decisions_based_on_impulse/

[Bowie69]

Fundamentally, people should be making those decisions based on their own set of beliefs, not one set of beliefs that have been dictated to them by a higher power.

Hmmm, reminds me of some of the stories that came out of Russia in the 60s and 70s...

[GW8IZR]

or.. maybe the safety systems and precautions preclude ( maybe reduce to an insignificant number ) that decision from ever needing to be made..

Most times people have to make such decisions when driving is because they have been failing to take account of the conditions or their driving is impaired  drunk.. tired .. distracted .. angry with wife..angry with bloke who just cut them up..

So driverless cars drive at the speed limit or slower and leave sensible gaps so there is less bunching and thus less congestion, improves ability to stop in time etc. This reduces number of accidents overall so the occasional bad decision is statistically insignificant.

 

[mad_pete]

First if you are in a driving state where you have to choose between various amounts of people to hit you have most likely made very poor / illegal driving choices that have led you to that point. Second in a car accident you will most likely be in full on 'lizard brain' mode making at best reflex actions rather than carrying out complex life choice evaluations.

 Cars can be programmed to choose well and quickly. If there are 2 people go into the one you will hit the slowest and deploy the bonnet airbag before you hit them. 

A lot of people get excited about this use case but how often is a speed limit abiding, knows it's stopping distance at all time driverless car going to encounter this ?  As opposed to drunk drivers racing each other through town centres.

Which is what the post above says a lot better than this one. :-)

 

[CwazyWabbit]

The autonomous vehicle may well be far less likely to have made the poor choices of a human driver in the lead up to a death or death decision I agree, but what if the situation is caused by one of us human types doing something stupid in an old vehicle (or on foot) that forces the auto into the situation? I'd view that as the most likely reason for this state of affairs to occur.

It's an interesting area but I have to concur that it is probably a statistical insignificance that will sell lots of tabloid newspapers.

[FridgeFreezer]

I'm more concerned about entirely random stuff confusing the AI, have a look at the pictures that it's absolutely sure are a cat or a photocopier or a school bus and wonder how many ways that could go wrong or be exploited:

https://www.wired.com/2015/01/simple-pictures-state-art-ai-still-cant-recognize/

That's my issue with all this machine learning - it might work 99.99% of the time... then one day a bit of graffiti on a wall will "spook" your car like a nervous horse and cause a pile-up, and no-one will be able to work out exactly why or how to fix it.

[GW8IZR]

47 minutes ago, CwazyWabbit said:

It's an interesting area but I have to concur that it is probably a statistical insignificance that will sell lots of tabloid newspapers.

yep .. that's the truth

28 minutes ago, FridgeFreezer said:

That's my issue with all this machine learning - it might work 99.99% of the time... then one day a bit of graffiti on a wall will "spook" your car like a nervous horse and cause a pile-up, and no-one will be able to work out exactly why or how to fix it.

More likely spook the car and bring the car to a safe stop, all the other cars will react to the safe stop and everyone is safe, a quick re assess and everyone moves on.. I bet we'd hardly look up from the tabloid newspaper reading about how dangerous it all is.

Of course thats just speculation on my part but I think more likely than a death race 2000 style inferno

 

[Davo]

I'm sure if this spooked robot car stopped suddenly on a busy highway in front of a huge robot truck it would be flattened before it had time to think "BSOD!" 

[FridgeFreezer]

Another interesting article from Bosch:

https://www.theregister.co.uk/2017/07/07/bosch_autonomous_car/

[geoffbeaumont]

9 hours ago, Davo said:

I'm sure if this spooked robot car stopped suddenly on a busy highway in front of a huge robot truck it would be flattened before it had time to think "BSOD!" 

Except that the robot truck wouldn't be following too close to stop...

[Davo]

13 hours ago, geoffbeaumont said:

Except that the robot truck wouldn't be following too close to stop...

Aha! I thought of that, (of course), but what if our robo-car had changed lanes to avoid, say, an old Land-Rover going too slowly? How does it speed up in time? Or does it wait until the traffic is clear? (If that's possible in some places.) Just how much distance will this robo-truck leave for emergencies? And if it does that, how does that affect the traffic flow? Are these things able to detect the road surface to allow for longer braking time for different weather? 

Of course people will try to design these things to allow for all of the above, but that's an awful lot at stake for someone in an office to get right before anything actually happens. 

[CwazyWabbit]

45 minutes ago, Davo said:

Aha! I thought of that, (of course), but what if our robo-car had changed lanes to avoid, say, an old Land-Rover going too slowly? How does it speed up in time? Or does it wait until the traffic is clear? (If that's possible in some places.) Just how much distance will this robo-truck leave for emergencies? And if it does that, how does that affect the traffic flow? Are these things able to detect the road surface to allow for longer braking time for different weather? 

Of course people will try to design these things to allow for all of the above, but that's an awful lot at stake for someone in an office to get right before anything actually happens. 

Which is why they will ban human driven cars because they are too unpredictable and lead to autonomous vehicles crashing ..........

[Davo]

Yes, I can see a time when the robots have priority over the rest of us! (At least, they'll probably get their own lanes . . . )

[Bowie69]

You could build an entirely separate network for freight traffic, have physical guides on the ground, electrify the lanes with overhead lines, so the battery/power storage issue is gone, and easily run them at a fixed distance from one another. 

Oh hang, we had that, it was called a rail network...

[Davo]

I say old man, what a brilliant idea! 

[GW8IZR]

7 hours ago, Bowie69 said:

You could build an entirely separate network for freight traffic, have physical guides on the ground, electrify the lanes with overhead lines, so the battery/power storage issue is gone, and easily run them at a fixed distance from one another. 

Oh hang, we had that, it was called a rail network...

 

don't forget canals as well, two good freight superhighways there and to be fair the canals are probably as fast as the A55 on a Friday afternoon

[smallfry]

Its coming...........of that there is no doubt. Why anyone would want one mystifies me, but that's just my opinion because I believe there are too many variables involved.

Just at random, will it be able to recognise a child from every conceivable angle and size through any amount of texture and type of clothing ?

As it approached a horse, will it recognise that the horse is getting skittish, like we humans can.

And as has been said, just WHO is liable ? I dont imagine any parent accepting that the death of their child "Is just one of those things" and letting it go.

No, I see it all being fine and lovely for a while, and then there will be some catastrophic accident caused by the failure of the sensors and software to recognise some circumstance.

Then what ?