Self-driving cars- the future or just another shiny toy?

[nrubenstein]

I think the big issue is the idea of programming morality into these cars. Often derived from the trolley problem thought experiment.

https://en.wikipedia.org/wiki/Trolley_problem

If a person needs to make a decision between hitting a granny with a puppy or two kids on bikes, assuming there was no other choices that a reasonable person could have made then we don't generally hold the person to a decision that was made under less than ideal circumstances.

But with self driving cars, this decision needs to be programmed in. So assuming that the hazards are identified correctly (another issue, false positives...) , who gets to decide who dies? the other scenario that comes up in discussions is whether the vehicle protects the occupants or protects people outside the car. Ie if the decision is between hitting a wall and killing the occupants or hitting a pedestrian and killing the pedestrian with the occupants surviving. The computer in the car has (or will have) the ability to identify these risks, and it needs to make a preprogrammed decision. And will some companies prioritise the occupants, and other companies priorities the pedestrian? will one company advertise "we prioritise the occupants in our software" ?

I've no doubt the technology will be here and affordable before we know it, but technology is the easy bit.

The trolley problem is massively, massively overblown. It’s a non-issue that people obsess over. They are going to program these things to minimize collisions as much as possible, and they are going to avoid getting into a logic loop that advocates killing the passengers. Not least because a vehicle programmed to kill the passengers in certain circumstances is going to do it erroneously.

The real issue is that we are a LONG way from self driving AI being able to effectively read the road / road conditions / other vehicles / and especially stationary objects correctly.

[varg]

Heady philosophy and thought experiments aside, self driving cars are currently a joke and a long way from being a widespread thing. People really underestimate the complexity of driving and overestimate how bad people are at it. Driving is not just the simple preprogrammed obstacle course task but something dynamic that requires observation and assessment of intent of drivers, pedestrians etc. to do well. You have to start from scratch to make a good self driving car, there is nothing common sense to a computer, there is nothing it knows how to do already. Even the perception of the information the sensors provide has to be sorted, no assumptions can be made. There does not currently exist any system that can look at a common scene and identify things happening in it like a person can.

Furthermore, things like Tesla "Autopilot" are a crude gimmick that is a liability more than a useful thing, and it's baffling to me that the company chose to market it under the name "Autopilot" given the implications of that term. Name it "Autopilot" and people will treat it as such, and they do, and bad things have happened because of it. With Tesla in particular (and other Musk operations to some extent) here seems to be a cavalier attitude to the tasks at hand behind the operation of the company. It's inevitable that given Tesla's software developer style approach of testing in situ and regular automatic updating to complex and dangerous tasks with real world consequences that some serious problem that would have been avoidable through extensive real world testing instead of using the end user as a guinea pig will occur. Maybe the resultant lawsuit will humble the cocksure silicon valley attitude, since it will hit them where it hurt$.

[george graves]

self driving cars are currently a joke and a long way from being a widespread thing.

Yea, and that arpanet never went anywhere! :)

I think if you had followed the work that Stanford did during the DARPA challenge(Sebastian Thrun) and realise where things are today, you would see things in a different light. That work was started in 2004 - FIFTEEN YEARS AGO - in technology terms, that's a forever ago. The state of things today is well beyond what you paint it to be frank.

[dessutom]

I think the big issue is the idea of programming morality into these cars. Often derived from the trolley problem thought experiment.

https://en.wikipedia.org/wiki/Trolley_problem

If a person needs to make a decision between hitting a granny with a puppy or two kids on bikes, assuming there was no other choices that a reasonable person could have made then we don't generally hold the person to a decision that was made under less than ideal circumstances.

But with self driving cars, this decision needs to be programmed in. So assuming that the hazards are identified correctly (another issue, false positives...) , who gets to decide who dies? the other scenario that comes up in discussions is whether the vehicle protects the occupants or protects people outside the car. Ie if the decision is between hitting a wall and killing the occupants or hitting a pedestrian and killing the pedestrian with the occupants surviving. The computer in the car has (or will have) the ability to identify these risks, and it needs to make a preprogrammed decision. And will some companies prioritise the occupants, and other companies priorities the pedestrian? will one company advertise "we prioritise the occupants in our software" ?

I've no doubt the technology will be here and affordable before we know it, but technology is the easy bit.

I totally agree. The ethical considerations are a side of this technology that is not shown in the marketing campaigns and it's no wonder why - it's a tricky question that nobody can give an unambiguous answer to. I'm also worried about the security shortcomings of autonomous vehicles. I hope automotive producers address this as a priority. Whom to prioritize in the case of an inevitable accident is also quite a valid point which they might want to agree upon.

[e30davie]

Years ago i remember seeing an interview with an engineer working on autonomous cars, and he was basically saying he can program it to do whatever we want, but its not the engineers decision as to what the car does in these decisions. It would be very interesting to see how this is handled these days in the cars we have. Are these hard decisions sitting at the CEO level? is Elon Musk signing off on the decision matrix for his cars? Probably. Should we be leaving it up to one guy to make decisions like this? probably not.

[AndrewBird]

I'm all for the type of autonomous driving that is basically cruise control with steering and braking as well. Would make commuting and stop and go traffic much nicer.

If we do go to completely autonomous, like, no steering wheel, I think they should be more like cabs, not personally owned. You call one to pick you up and it takes you where you want to go. Then the liability is on the company, not the passenger.

Being able to have all the cars talk to each other would be a great idea, even if not autonomous. Imagine being able to have the car avoid a crash because of info it got from all the other cars around it.

[Twistytee]

I work in the industry and autonomous vehicles are closer to reality than most of these forum post lead you to believe. I love wrenching and driving my non- automated vehicles as much as the majority of R3vlimited forum members but don’t be naive as to think this technology won’t displace our preferred method of transportation in the long haul. The legal/common welfare issues will eventually get sorted. It’s not as sci-fi as some of these posts would imply.

[george graves]

I'm all for the type of autonomous driving that is basically cruise control with steering and braking as well. Would make commuting and stop and go traffic much nicer.

Drafting at 90mph, on a train of 4-5 cars 3 feet off each other's bumpers - I wonder what kind of gas/power that would save.

Apparently it works for NASCAR :)

[325e '87]

I'm all for the type of autonomous driving that is basically cruise control with steering and braking as well. Would make commuting and stop and go traffic much nicer.

I know its off topic, but high volume traffic is a trigger for me. The general population just cant do it right and it would be a huge boon for everyone if there was a way for vehicles to navigate clogged interstates on their own. With the near zero presence of cars with manual transmissions the majority of people cant coast or slow down without showing brake lights. Every time brake lights show the problem turns into the maddening stop, go, go too fast, brake hard and stop.

[MrAdam]

I work at a Tesla certified body shop
I fix these self driving cars just as much as the human operated vehicles.

I think there is a lot more work that needs to be done before we are in the Jetsons era

[ForcedFirebird]

Coincidence this subject came up here when I just read this last week:

https://electrek.co/2019/07/08/tesla-will-stop-selling-cars-full-self-driving-elon-musk/

[e30strokr]

I think the big issue is the idea of programming morality into these cars. Often derived from the trolley problem thought experiment.

https://en.wikipedia.org/wiki/Trolley_problem

.

trying to put feelings into a car wont help.. it will eventually be a numbers game if its inevitable... quote morpheus here.....


granny vs 3 kids..
3 kids Vs 2 adults
1 granny with a puppy Vs 2 adults


where waffleswaffleswaffles s it stop?


we should have more trains...

[varg]

Yea, and that arpanet never went anywhere! :)

I think if you had followed the work that Stanford did during the DARPA challenge(Sebastian Thrun) and realise where things are today, you would see things in a different light. That work was started in 2004 - FIFTEEN YEARS AGO - in technology terms, that's a forever ago. The state of things today is well beyond what you paint it to be frank.

Sarcasm doesn't change the fact that they are far from being common. People are vastly oversimplifying the complexity of the task at hand, DARPA challenge stuff was a mere oddity and barely worked even though it was done on static courses with little to no unpredictable factors. Self driving cars will have to deal almost entirely in unpredictable factors, teslas on autopilot still have issues holding lane and detecting stopped traffic and they're crunching millions of miles worth of data and constantly updating things. Current state: not commercially viable by a longshot, will be lucky to even be deployed on more than a limited test basis in the next 5 years let alone reach common status in the next 10.

we should have more trains...

We don't all live in urban areas.

[george graves]

they are far from being common.

I don't look at it as a percentage of cars on the road to get data, I look at it as miles driven. Of course you don't need 50% of the car on the road to be self-driving to get data That's absurd.

And there's been more miles driven on a self driving car then you or I have driven in our lifetimes. So - yea.

From a computer science perspective, the tech is evolving at a hockey stick type graph.

Real world and anecdotally, as i said before - there will be accidents. And a lot of this depends on how the media covers it. So far it's really been cruel to Tesla (and no, I don't own stock, or the car, or plan to ever)

I'll just leave this here....you're driving on a country road, you and the car approaching are both doing 60 mph. The other driver reaches for a cheetos she got at walmart, and it dropped down her blouse, cause she missed her big fat mouth and swerves into your lane. That's a 120 mph impact. You die cause she wants a snack, and she missed her mouth.

I want a computer driving her car. Hell yes I do.

[dessutom]

Years ago i remember seeing an interview with an engineer working on autonomous cars, and he was basically saying he can program it to do whatever we want, but its not the engineers decision as to what the car does in these decisions. It would be very interesting to see how this is handled these days in the cars we have. Are these hard decisions sitting at the CEO level? is Elon Musk signing off on the decision matrix for his cars? Probably. Should we be leaving it up to one guy to make decisions like this? probably not.

You are right that we shouldn't let one person or one company to make those decisions for everyone, this is exactly what I want to draw attention to, primarily the automotive producers' attention because corporations wrongly assume that the fact they are influential and have resources, gives them the right to do whatever they please. Well, history shows that this has happened, it is still happening but it's certainly not right, at least from my point of view. Engineers actually DO have the responsibility to say "no" if a corporation wants to create something that is detrimental for the society. History again shows most didn't, that's why Facebook and Twitter and such have become so destructive. I hope we can avoid that with autonomous vehicles because they certainly have the potential to work in people's favour.