5 ms·

I'd also like unicorns towing magic carpets and appreciate the sarcasm, self driving cars are not the 'unicorn' anymore. Uber is actively investing in prepping

by ktta 10y ago

I'd also like unicorns towing magic carpets and appreciate the sarcasm, self driving cars are not the 'unicorn' anymore. Uber is actively investing in prepping for a self-driving world.[0][1] https://newsroom.uber.com/mapping-ubers-future/ https://newsroom.uber.com/mapping-ubers-future/ http://www.theverge.com/2016/7/31/12338268/uber-maps-investment-500-million http://www.theverge.com/2016/7/31/12338268/uber-maps-investm...

brokenmachine 10y ago
When a ~$100k Tesla can't successfully avoid a stationary guardrail or a turning semi, even under the limited conditions that it's autopilot actually attempts to run under, I'd say it's massively premature to suggest that there's anything like a driverless future coming in even the medium-term for affordable vehicles. There's just too many variables on our roads and the stakes are too high. If they are coming, it's more than 25 years away IMO. Remember 25 years ago, cars weren't super different from today. ABS was first in a production car in 1971. I think the average car on the road now is 11 years old, so there's bound to be unpredictable manually-driven cars still on the road for a long time that any driverless cars would still have to contend with as well. Self driving cars are pretty much unicorns.
- JoeAltmaier 10y ago
  It isn't about single accidents; its about statistics. Autodriven cars don't get tired, or drunk, or fool around with the playlist. There is so much low-hanging fruit that its entirely conceivable that autodrive is already safer (statistically) than human drivers. Add to this, that each time a special case comes along, the software can be patched and it won't happen again. Something humans are terrible at - learning from others' mistakes.
  - tedmiston 10y ago
    I think parent comment was suggesting that human-driven cars that behave erratically are better handled by opposing humans rather than software (that has not been exposed to dealing with the novel situation). (Not that auto-driven cars don't produce less accidents when other human-driven cars behave expectedly.)
    - brokenmachine 10y ago
      I'm imagining that driverless cars could work well on very predictable roads that are exclusively for driverless cars. But on real roads with questionable maintenance and other unpredictable cars on it, I just can't see it happening unless there's some major change to the roads infrastructure.
  - brokenmachine 10y ago
    So when an autodriven car maims little Johnny, before they patch the bug to not run over future infants, then who's responsible and has to pay? What if other cars maim more infants before they get around to patching what is now a known bug? You can't blame the driver, they weren't driving. The car manufacturers can't accept blame, as it costs a lot if you have to factor in the cost of a lifetime of support for a maimed infant. Even ignoring the fact that it only works on highways and not normal surface roads, Tesla's "beta" Autopilot feature takes no responsibility because the driver is still supposed to be holding the wheel. Also as any software developer would tell you, software updates would also be a problem in themselves. If you propose that cars be auto-updated automatically, I wonder how insurance would work when the car is running different software to the software it was originally insured with. Critical saftey systems can't get auto-updated without a lot of testing. What if the auto-update mechanism gets hacked, or maliciously changed by a disgruntled self-driving car employee? What if a legitimate update causes another bug where it maims the elderly now instead of infants? Do they roll back to the infant-maiming software now? I'm being silly but updates on critical safety stuff isn't so simple, especially with the legal ramifications. I think legal issues will be a problem for self-driving cars, maybe even a larger problem than the actual designing a working self-driving car problem. Someone has to be responsible for inevitable accidents (bugs?). After a century of being able to blame the driver, car companies won't want to change that status quo. Maybe insurance will decide that they'll cover self-driving cars if they prove themselves to be safe, but we'll see.
    - ktta 10y ago
      This is the biggest problem that's really holding back self driving cars right now, and it isn't really new. It is called the trolley problem [0]. Wired has an article or two about this [1] [0]https://en.wikipedia.org/wiki/Trolley_problem https://en.wikipedia.org/wiki/Trolley_problem [1] https://www.wired.com/2016/06/people-want-self-driving-cars-save-lives-especially/ https://www.wired.com/2016/06/people-want-self-driving-cars-...
  - Spooky23 10y ago
    No, it's about liability and risk. Uber is amazing, but offloads lots of risk and capital expense to the saps^H drivers and their judgement. I associate self-driving Ubers with unicorns and flying carpets not because of self-driving car technology, but because of the total shift in business model for a future Uber. Self driving cabs require, at a minimum: - Owning or leasing capital assets (robot cars) - Maintenance (from mechanical repair to removing puke and vandalism) - Provide local infrastructure for them (real estate for storage, maintenance and charging) - Own liability for all of the various mishaps that will naturally happen. (accidents, human-prompted mischief or criminal acts (Uber drive by shootings, prostitution, narcotics transport, etc)
- tedmiston 10y ago
  > so there's bound to be unpredictable manually-driven cars still on the road for a long time that any driverless cars would still have to contend with as well. Luckily unicorn-driven magic carpets don't face this problem.
- ktta 10y ago
  >Tesla can't successfully avoid a stationary guardrail or a turning semi It can. The problem wasn't that. The problem was that it wasn't able to differentiate the color properly and thought nothing was there. There's more info in Tesla's blog post. And it's not like it is an unsurmountable problem that is leading to accidents everywhere. It was an edge case and will probably be sorted out. >massively premature to suggest that there's anything like a driverless future I really think that's a very harsh opinion. And you're asserting this opinion on many intelligent people working towards a difficult solution. Thousands of people have used auto pilot without any problems and many lives were probably saved (consider if they were driving normal cars instead of self driving ones?) >There's just too many variables on our roads and the stakes are too high. Nobody is disagreeing with you on that. That's the reason it already took so long. Self driving cars first appeared in 2005 if not earlier [0]. And remember all of self driving is mostly a software problem, which greatly reduced iteration time for new developments unlike electronic or mechanical revisions. >25 years away IMO Funny how people predict things. Ray Kurzweil predicts AGI (general Artificial Intelligence) to arrive before the year 2045, 4 years after you say we get self driving cars. I'm not going to argue with you on that number since we probably have different opinions on what constitute self driving cars. Just so you know, I predict we'll have fully self driving cars on the road (with humans in them) in 5 years. >Remember 25 years ago, cars weren't super different from today Please tell me you're kidding. [1][2] >Self driving cars are pretty much unicorns At least those unicorns are achievable and people are trying towards achieving it rather than give up. [0]https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2005_Grand_Challenge https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2005_Gra... [1]http://www.popularmechanics.com/cars/a11201/why-cars-are-safer-than-theyve-ever-been-17194116/ http://www.popularmechanics.com/cars/a11201/why-cars-are-saf... [2]http://www.iihs.org/iihs/sr/statusreport/article/50/1/1 http://www.iihs.org/iihs/sr/statusreport/article/50/1/1
  - brokenmachine 10y ago
    >>Tesla can't successfully avoid a stationary guardrail or a turning semi >It can. The problem wasn't that. The problem was that it wasn't able to differentiate the color properly and thought nothing was there. Really? You're arguing that it can avoid a guardrail if it was actually able to tell a guardrail was there?? lol. >Funny how people predict things. Ray Kurzweil predicts AGI (general Artificial Intelligence) to arrive before the year 2045, 4 years after you say we get self driving cars. Interesting. Who knows. I've heard people say that it will accelerate exponentially after the first AI is developed as the AI will be able to iterate itself. >>Remember 25 years ago, cars weren't super different from today >Please tell me you're kidding. Sure, they're safer because of improvements in crumple zones/better brakes/steering/etc. I meant the basic technology of an internal combustion engine for power and a human in control. There have been incremental improvements but nothing like an autonomous car. >>Self driving cars are pretty much unicorns >At least those unicorns are achievable and people are trying towards achieving it rather than give up. I think they're technically achievable with dedicated road infrastructure but on normal badly-maintained roads with non-autonomous cars to dodge, they're unicorns. Legal issues will be the biggest hurdle IMO.

by ktta 10y ago

brokenmachine 10y ago

When a ~$100k Tesla can't successfully avoid a stationary guardrail or a turning semi, even under the limited conditions that it's autopilot actually attempts to run under, I'd say it's massively premature to suggest that there's anything like a driverless future coming in even the medium-term for affordable vehicles. There's just too many variables on our roads and the stakes are too high. If they are coming, it's more than 25 years away IMO. Remember 25 years ago, cars weren't super different from today. ABS was first in a production car in 1971. I think the average car on the road now is 11 years old, so there's bound to be unpredictable manually-driven cars still on the road for a long time that any driverless cars would still have to contend with as well. Self driving cars are pretty much unicorns.

JoeAltmaier 10y ago
It isn't about single accidents; its about statistics. Autodriven cars don't get tired, or drunk, or fool around with the playlist. There is so much low-hanging fruit that its entirely conceivable that autodrive is already safer (statistically) than human drivers. Add to this, that each time a special case comes along, the software can be patched and it won't happen again. Something humans are terrible at - learning from others' mistakes.
- tedmiston 10y ago
  I think parent comment was suggesting that human-driven cars that behave erratically are better handled by opposing humans rather than software (that has not been exposed to dealing with the novel situation). (Not that auto-driven cars don't produce less accidents when other human-driven cars behave expectedly.)
  - brokenmachine 10y ago
    I'm imagining that driverless cars could work well on very predictable roads that are exclusively for driverless cars. But on real roads with questionable maintenance and other unpredictable cars on it, I just can't see it happening unless there's some major change to the roads infrastructure.
- brokenmachine 10y ago
  So when an autodriven car maims little Johnny, before they patch the bug to not run over future infants, then who's responsible and has to pay? What if other cars maim more infants before they get around to patching what is now a known bug? You can't blame the driver, they weren't driving. The car manufacturers can't accept blame, as it costs a lot if you have to factor in the cost of a lifetime of support for a maimed infant. Even ignoring the fact that it only works on highways and not normal surface roads, Tesla's "beta" Autopilot feature takes no responsibility because the driver is still supposed to be holding the wheel. Also as any software developer would tell you, software updates would also be a problem in themselves. If you propose that cars be auto-updated automatically, I wonder how insurance would work when the car is running different software to the software it was originally insured with. Critical saftey systems can't get auto-updated without a lot of testing. What if the auto-update mechanism gets hacked, or maliciously changed by a disgruntled self-driving car employee? What if a legitimate update causes another bug where it maims the elderly now instead of infants? Do they roll back to the infant-maiming software now? I'm being silly but updates on critical safety stuff isn't so simple, especially with the legal ramifications. I think legal issues will be a problem for self-driving cars, maybe even a larger problem than the actual designing a working self-driving car problem. Someone has to be responsible for inevitable accidents (bugs?). After a century of being able to blame the driver, car companies won't want to change that status quo. Maybe insurance will decide that they'll cover self-driving cars if they prove themselves to be safe, but we'll see.
  - ktta 10y ago
    This is the biggest problem that's really holding back self driving cars right now, and it isn't really new. It is called the trolley problem [0]. Wired has an article or two about this [1] [0]https://en.wikipedia.org/wiki/Trolley_problem https://en.wikipedia.org/wiki/Trolley_problem [1] https://www.wired.com/2016/06/people-want-self-driving-cars-save-lives-especially/ https://www.wired.com/2016/06/people-want-self-driving-cars-...
- Spooky23 10y ago
  No, it's about liability and risk. Uber is amazing, but offloads lots of risk and capital expense to the saps^H drivers and their judgement. I associate self-driving Ubers with unicorns and flying carpets not because of self-driving car technology, but because of the total shift in business model for a future Uber. Self driving cabs require, at a minimum: - Owning or leasing capital assets (robot cars) - Maintenance (from mechanical repair to removing puke and vandalism) - Provide local infrastructure for them (real estate for storage, maintenance and charging) - Own liability for all of the various mishaps that will naturally happen. (accidents, human-prompted mischief or criminal acts (Uber drive by shootings, prostitution, narcotics transport, etc)
tedmiston 10y ago
> so there's bound to be unpredictable manually-driven cars still on the road for a long time that any driverless cars would still have to contend with as well. Luckily unicorn-driven magic carpets don't face this problem.
ktta 10y ago
>Tesla can't successfully avoid a stationary guardrail or a turning semi It can. The problem wasn't that. The problem was that it wasn't able to differentiate the color properly and thought nothing was there. There's more info in Tesla's blog post. And it's not like it is an unsurmountable problem that is leading to accidents everywhere. It was an edge case and will probably be sorted out. >massively premature to suggest that there's anything like a driverless future I really think that's a very harsh opinion. And you're asserting this opinion on many intelligent people working towards a difficult solution. Thousands of people have used auto pilot without any problems and many lives were probably saved (consider if they were driving normal cars instead of self driving ones?) >There's just too many variables on our roads and the stakes are too high. Nobody is disagreeing with you on that. That's the reason it already took so long. Self driving cars first appeared in 2005 if not earlier [0]. And remember all of self driving is mostly a software problem, which greatly reduced iteration time for new developments unlike electronic or mechanical revisions. >25 years away IMO Funny how people predict things. Ray Kurzweil predicts AGI (general Artificial Intelligence) to arrive before the year 2045, 4 years after you say we get self driving cars. I'm not going to argue with you on that number since we probably have different opinions on what constitute self driving cars. Just so you know, I predict we'll have fully self driving cars on the road (with humans in them) in 5 years. >Remember 25 years ago, cars weren't super different from today Please tell me you're kidding. [1][2] >Self driving cars are pretty much unicorns At least those unicorns are achievable and people are trying towards achieving it rather than give up. [0]https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2005_Grand_Challenge https://en.wikipedia.org/wiki/DARPA_Grand_Challenge#2005_Gra... [1]http://www.popularmechanics.com/cars/a11201/why-cars-are-safer-than-theyve-ever-been-17194116/ http://www.popularmechanics.com/cars/a11201/why-cars-are-saf... [2]http://www.iihs.org/iihs/sr/statusreport/article/50/1/1 http://www.iihs.org/iihs/sr/statusreport/article/50/1/1
- brokenmachine 10y ago
  >>Tesla can't successfully avoid a stationary guardrail or a turning semi >It can. The problem wasn't that. The problem was that it wasn't able to differentiate the color properly and thought nothing was there. Really? You're arguing that it can avoid a guardrail if it was actually able to tell a guardrail was there?? lol. >Funny how people predict things. Ray Kurzweil predicts AGI (general Artificial Intelligence) to arrive before the year 2045, 4 years after you say we get self driving cars. Interesting. Who knows. I've heard people say that it will accelerate exponentially after the first AI is developed as the AI will be able to iterate itself. >>Remember 25 years ago, cars weren't super different from today >Please tell me you're kidding. Sure, they're safer because of improvements in crumple zones/better brakes/steering/etc. I meant the basic technology of an internal combustion engine for power and a human in control. There have been incremental improvements but nothing like an autonomous car. >>Self driving cars are pretty much unicorns >At least those unicorns are achievable and people are trying towards achieving it rather than give up. I think they're technically achievable with dedicated road infrastructure but on normal badly-maintained roads with non-autonomous cars to dodge, they're unicorns. Legal issues will be the biggest hurdle IMO.