Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Amperzand]

PTW

[Dozebôm Lolumzalìs]

I suppose that people don't like the idea of "those robots" having control over cars. They imagine that a robot uprising could occur, and then self-driving cars would wreak havoc on the world. (Or they could be hacked, etc.)

There are a few issues with that:

1. People could do that, too. Your neighbor Bob could go insane and drive his car through your house. But that's so unlikely that we don't worry about it. Similarly, with decent security, we could make the odds of a self-driving car driving itself directly into a pedestrian very low. (And since it's not really an AI...)

2. XKCD notes that cars already have computers. All this would do is allow them to see and steer. Bad, but not that bad.

There's also the potential reason of "I want people driving cars, so we have someone to hold responsible for an accident!" That's a decent objection.

[Max™]

Driving has always seemed to me like one of the things most obviously best suited for a computer.  Faster reaction times, no variation in awareness levels, 100% adherence to rules, etc.

It's always seemed like that to me as well, most of the people around me seem to disagree though. They seem to think that humans are inherently better at absolutely everything.

I noticed it tends to alarm people when I inform them how little I think about driving. It's all essentially an automated subroutine I run when I've got pedals and a wheel available. I think when I'm racing along at 200 mph jostling for position and managing fuel and tires. I keep an eye out for unexpected obstacles and potential problems and basically let my body handle the rest when driving. I don't even really think about the destination unless it looks like the current route is less than ideal.

I know that there is too much going on to fully divert my attention doing something stupid like texting, taking selfies, putting on makeup, or any o the other stupid things people do when driving... but I wouldn't call it beyond automation in any sense.

[Emma]

There's also the potential reason of "I want people driving cars, so we have someone to hold responsible for an accident!" That's a decent objection.

But if an automated car has a crash you still have someone to hold accountable, the designer or manufacturer of the vehicle. They might not be physically driving the car but if there's a crash because of something that they did they can still be held responsible and punished accordingly.

[SalmonGod]

There's also the potential reason of "I want people driving cars, so we have someone to hold responsible for an accident!" That's a decent objection.

On that note... do you need insurance if you yourself never directly pilot your vehicle?  You didn't cause the accident.  The car's programming did.  It makes more sense to hold the company responsible for that.  So what if you told people "Hey self-driving vehicle = save money because you don't have to buy insurance."

And haven't there been very, very few accidents involving self-driven vehicles, over millions of miles of testing?  And all were caused by other human drivers on the road?  So I don't actually see the companies developing self-driving vehicles having a problem with that liability, especially as the technology takes off and human drivers on the road become less of a risk.

[redwallzyl]

There's also the potential reason of "I want people driving cars, so we have someone to hold responsible for an accident!" That's a decent objection.

On that note... do you need insurance if you yourself never directly pilot your vehicle?  You didn't cause the accident.  The car's programming did.  It makes more sense to hold the company responsible for that.  So what if you told people "Hey self-driving vehicle = save money because you don't have to buy insurance."

And haven't there been very, very few accidents involving self-driven vehicles, over millions of miles of testing?  And all were caused by other human drivers on the road?  So I don't actually see the companies developing self-driving vehicles having a problem with that liability, especially as the technology takes off and human drivers on the road become less of a risk.

mandatory no fault insurance?

[Max™]

I'd just like to point out this little anecdote.


My parents own a rather new car; can't recall the brand. The car's cruise control allows for car-length spacing between you and the person ahead (you can set it to be  1 car-length behind, 2, etc), and steers itself to stay between the road lines. Only thing it doesn't really handle well are bends and when you actually need to make a turn or change lanes to reach your destination.

We've been approaching self-driving cars for a while now, by slow incremental improvements.

There are much more involved cruise control forms now.

http://www.porsche.com/international/models/panamera/panamera-4s/assistance-systems/porsche-innodrive/
http://www.porsche.com/international/models/panamera/panamera-4s/assistance-systems/lane-change-assistant/
http://www.porsche.com/international/models/panamera/panamera-4s/assistance-systems/park-assistant/
http://www.porsche.com/international/models/panamera/panamera-4s/assistance-systems/lane-departure-warning/

These are tuned to be a more fun cruise control along with the overhead surround view for parking, warnings if you drift from your lane, and systems to keep track of nearby cars. It'll keep track of the corner radius ahead and alter the throttle/braking to remain within limits you set, so you steer and it does the rest.

Stuff like this will probably filter into more and more cars and bit by bit more aspects of driving will be moved out of the hands of the driver. Manual shifting is damn near a theft deterrent system at this point, various cars can be lined up and told to parallel park, and just do it nowadays. Keeping pace with traffic, altering for cornering, street signs, bit by bit it's moving towards a driver only being present for emergency input, if even that.

[Starver]

An automated vehicle within a controlled environment (say robot forklifts servicing between the stacks of an automated-only warehouse floor) is going to be advantageous compared to the very same system with entirely human controllers. But the safeguards to prevent bad man/machine interactions should probably go so far as to dissallow any pedestrian access down the inter-stack aisles unless the system were specially asked to disallow robo-forklift transit (and possibly sufficiently preventative barriers set up, to maybe both signal that revocation of access and physically gst in the way).

Our roads, however, are not so tightly controlled as the automated warehouse. The problems an automated vehicle has to overcome may start with other autonomous vehicles (inter-vehicle notifications and negotiation might be 'predesigned' into a properly specced vehicle with a wifi-like beacon perhaps) but also passes through manually-controlled vehicles (beacons for which, if they exist, cannot speak for the driver's actual intentions, only its current position and velocity), pedestrians (assumed intelligently aware, for the most part, but requires fully un-negotiated detection), animals (may be intelligent/aware to a lesser degree, but that'd be effectively an exception to the norm) and environment, to name just a few.

Right now, the 'open road' is too open to guarantee that hypersmart and (mostly!) all-seeing vehicles will not be fooled by a swirl of sutumn leaves on the road or somesuch. If not to cause a crash, then to slow down and halt as the vehicle's lidar/camera system inputs are taken as out-of-range taking common-sense beyond mere caution. Almost ensuring zero potential incidents, but also disrupting traffic flow.

But, about the all-seeing: there was the Tesla crash in June (IIRC). The visual systems of the vehicle being allowed to drive 'unattended'¹ did not observe a big-rig turning across the road, blending the upper body in with the sky, apparently, so ploughed straight on and under the rear. A person driving with such visual impairment might (admitedly, knowing humans, not as often as they should...) at least have the common sense to know that not being able to see things does not mean that there are no things to see. Maybe the algorithm needs tweaking for additional caution. Maybe all artic trailers need big zebra stripes on them to be seen better in all conditions (except in herds, obviously!). Maybe all trucks should be autonomous such that no road-crossing truck is ever not negotiating its traversal with all traffic within transceiver contact distance?

But, getting from the current human-fallible situation to the machine-perfect future is frought with difficulty in the transition, and probably only asymptotic in the achievement, at least whilst Zion keeps sending out its hovercraft patrols with their pesky EMPs.

¹ The owner/nominal-driver, a proponent and enthusiast of the self-driving car was apparently watching a Harry Potter movie and not monitoring his car's 'routine' self-driving behaviour, as Tesla were still recomending its customers do.

[Reelya]

That's all good theory, but the trials of driverless cars seem to be involved in significantly less accidents per mile than human drivers. So for all that e.g. a computer could be fooled by ambiguous sensory data "swirl of autumn leaves", it seems that on the balance, humans are distracted by more things than the robots.

So the logical error here is to compare any problems a robot car would have to some idealized "perfect driver" who never makes mistakes. That's clearly not a valid argument when it boils down to "human drivers or machine drivers". You need to compare accident statistics, not highlight anecdotal possible accidents that could happen. For all the "swirl of autumn leaves" problems, it's clear that humans get that worse. Maybe not the exact same stimuli, but statistically they cause more accidents.

[TempAcc]

Additionaly, the events that did cause an AI driver to fuck up showed to also be completely capable of causing a human driver to fuck up. One of the most notorious ones involving a google car was apparently caused by a passing truck reflecting a whole lot of light onto the car's sensors, essentialy blinding it and making the sensors relay incorrect information to its processing unit. On a statistical levels, AI and human drivers are very similar, to the point that considering the current cutting edge AI driven cars unsafe implies one also considers humans to be unsafe drivers.

[BorkBorkGoesTheCode]

Driverless cars need to become hacker/assassination-proof before general use. Please don't blindly support them because they appear to be a winning front in the battle between Progress and Reaction.

[TempAcc]

"Hackers" has almost become a buzzword in any discussion regarding any new tech. Any sort of new tech that shows up that makes some meager use of the internet usualy gets plastered by fearmongering involving hackers. Sure, hackers are a very real danger, but its not like hackers aren't already a significant danger, and that hasn't stopped google from, well, making google? I mean, if we're gonna refuse to use any new tech because hackers might one day try to hack it, we might as well give up on the internet and go back to sending paper letters.

Plus, we may be on the cusp of new ways of information encryption using neural nets (and maybe quantum computing, depending on how acessible quantum computing becomes), which will present hackers with new, bigger challenges. Caution is one thing, fearmongering is another.

[Emma]

Driverless cars need to become hacker/assassination-proof before general use. Please don't blindly support them because they appear to be a winning front in the battle between Progress and Reaction.

How is it going to be possible to make them hacker-proof? There's always going to be a possibility for automated vehicles to be hacked, reducing that possibility is in no way a bad thing but, as TempAcc has already mentioned, not using new technology just because it can be hacked isn't entirely logical.

[BorkBorkGoesTheCode]

This conversation will mainly involve Enthusiasts and Fearmongers.

[martinuzz]

Overall, I agree that fear of hacking should not stop new technology from being used. But automated cars are not just some new automation thingy. A coordinated hack during rush hour could kill millions. So at the very least, anti-hacking prevention should be amongst the top priority design goals for automated cars.