Self-driving test cars with remote control backup could hit California roads in April

2»

Comments

  • Reply 21 of 33

    cgWerks said:
    When one thinks of the nuances of what one does with a car -- such as where to park it or pick up someone (rather than the typical driving mode):
    ...  For example, a person who normally parks at the far end of the parking lot to protect the car from nicks and to get exercise decides to drive around the lot looking for the closest spot because its cold outside and they forgot a jacket....

    Some form of human control is required...

    If that control is remote, then it is possible to hack it and drive the car off a cliff rather than find the closest parking spot.  So, if anybody is going to be remotely controlling a car that I'm in, I would want it to be Apple.   Would anybody trust Microsoft or Google?
    The big problem is that proponents of AI look at this as a quantitative problem. i.e.: sensors aren't good enough, processors aren't fast enough, etc. and that once this is overcome, the capacity will be there for the system to 'think' and figure this stuff out. However, it's actually a qualitative problem. AI doesn't think, and never will. The 'databases' of the decision tree will get built in more detail... programmers will make adjustments to how the car should respond... sensors and processors will get faster, allowing more detailed input and ability to crunch the data more quickly, but the fundamental issue remains. There will ***ALWAYS*** be unique, unpredictable situations that occur. When that happens, the AI will fail (hence remote control so a human can solve the situation).

    While I trust Apple (currently) more in terms of data privacy, I don't trust any of them in regard to AI, because the proponents aren't even thinking correctly about it. They are living in a sci-fi dream-world.... and IMO, not really thinking of the downsides as much as the potential upsides and $$$.
    This is an opinion obviously. You dismiss completely the way humans thinking by applying a “special” filter to the way we make descions. AI may kill us I’ll grant You that... but there’s nothing about our thinking that we can’t or won’t achieve artificial. Leave your god nonsense out of it. We will program us the same way we’re programmed... there is no other way
  • Reply 22 of 33
    cgWerkscgWerks Posts: 2,840member
    bradford_kirby said:
    This is an opinion obviously. You dismiss completely the way humans thinking by applying a “special” filter to the way we make descions. AI may kill us I’ll grant You that... but there’s nothing about our thinking that we can’t or won’t achieve artificial. Leave your god nonsense out of it. We will program us the same way we’re programmed... there is no other way
    And, this is your opinion. The question is... which is more grounded in reality?

    AI is accomplishing amazing things, no doubt. But, there is a difference between reality and sci-fi. Even the pioneers of AI are now beginning to question the foundations. And, the truth is that we don't have a clue how real intelligence and sentience actually work (via science).

    The basic idea behind automated vehicles is that eventually enough real-world data will be fed in, such that *enough* situations are covered so as to make them safer (on average) than human drivers. It's a statistics game. But, the autonomous vehicle, when faced with a unique situation (where enough info isn't available to make the decision via the programming) is going to fail. That's because it's artificial (the 'A' aspect) and not really thinking.

    And, unfortunately, a lot of smart people like Kurzweil, Musk, etc. have bought into the sci-fi aspect, as has most of the media. AI (as useful as it is) is being way overhyped and misunderstood.

    As for my 'god nonsense' ... not quite sure where that came from... but I'll put my worldview up against your worldview any day. Probably a bit OT for here though, but if you're interested, I'll give you a way to get in touch. It might make good fodder for my blog or upcoming podcast.
    GeorgeBMaclolliver
  • Reply 23 of 33
    GeorgeBMacGeorgeBMac Posts: 11,421member

    I’d like to point out yet another technology (company/companies/industry) that is completely oblivious to the world outside the ideological (and climatological) bubble that is Southern California. So I’m supposed to scrape snow and ice off of 40 different sensors if I want to drive? And how’s the system going to work if it can’t see the road lines thanks to… snow and ice? Even rain seems to be a problem. “So they’ll put them all under a single glass dome.” Okay, so I still have to get the dome clean (which is easier), but leaving streaks of water/ice will bend the light entering the sensors and alter their perception. And electric cars: batteries don’t like the cold. Ah, well. I refuse to use a self-driver in the first place, so the point is moot for me. Just not for anyone trying to sell them to the northern (and soon to be southern) states.

    Since California is the most litigious place on Earth it’s hard to believe the savings from remote operation would offset the damages a jury would award to the first accident victim. 
    And when computers are given human rights, some of the damages will have to be given to the car that was hit for medical expenses (body restoration) or a funeral (cubificaiton at a junkyard).

    Don’t laugh. I’m not joking. There are people trying to grant human rights to AI. Meaning restrictions on workdays. And voting rights.
    Good points on snow, ice & rain effects on sensors.
    But, Uber has been running a fleet of self driving cars in Pittsburgh throughout the winter.   Admittedly, there isn't a lot of reporting being done on them so I don't know how well they did through the rather nasty winter we've had -- but I haven't heard of any failures.  But, in any case, they are being tested and used in hostile, winter weather conditions.
    No in fact none of them are good points. LIDAR sees through snow and ice and can be tuned to various weather related phenomena. As for your laziness and upkeep of your automobile... I can’t say... I’d think if you paid 80k for a Cadillac CTS that you’d clean snow off it... Tesla is one to worry about Elon won’t use lidar and it’s a problem
    I see... You don't live in a snow belt where you come out of a movie and your car is coated with a 1/4" of ice...
    cgWerks
  • Reply 24 of 33
    cgWerkscgWerks Posts: 2,840member
    GeorgeBMac said:
    I see... You don't live in a snow belt where you come out of a movie and your car is coated with a 1/4" of ice...
    Heh, yea... it's more how they deal with this kind of stuff... (the ice was accumulating as I was driving, even though the car was fully warmed up).


  • Reply 25 of 33
    GeorgeBMacGeorgeBMac Posts: 11,421member
    cgWerks said:
    GeorgeBMac said:
    I see... You don't live in a snow belt where you come out of a movie and your car is coated with a 1/4" of ice...
    Heh, yea... it's more how they deal with this kind of stuff... (the ice was accumulating as I was driving, even though the car was fully warmed up).


    Oh, I see!   I never realized you could just wait for the snow & freezing rain to stop, the temperatures to rise, and the sun to come out before driving!   That'll work for all those sensors on the self-drive cars!
  • Reply 26 of 33
    I’d like to point out yet another technology (company/companies/industry) that is completely oblivious to the world outside the ideological (and climatological) bubble that is Southern California. So I’m supposed to scrape snow and ice off of 40 different sensors if I want to drive? And how’s the system going to work if it can’t see the road lines thanks to… snow and ice? Even rain seems to be a problem. “So they’ll put them all under a single glass dome.” Okay, so I still have to get the dome clean (which is easier), but leaving streaks of water/ice will bend the light entering the sensors and alter their perception. And electric cars: batteries don’t like the cold. Ah, well. I refuse to use a self-driver in the first place, so the point is moot for me. Just not for anyone trying to sell them to the northern (and soon to be southern) states.

    As someone who grew up in Western NY, where winter can start in September or October, and last until April or May (I've seen snow in June), I learned long ago to clean the entire car of snow and ice before attempting to drive.  Few things are worse than having snow or ice fly back to the windshield, or fly off of someone else's car into mine.

    Of course, that said, where I live now, nobody seems to know about how to do that, so while your point is well founded, I suspect there will have to be safeguards in place to prevent autonomous operation when the sensors aren't clear.  If there aren't such safeguards in place now, they will be the first time something happens.

  • Reply 27 of 33
    Regarding snow and ice - the big difference most people are missing is that these are not cars that are sold and are sitting in your driveway or a car park in the snow. All the companies mentioned here; Waymo, Zoox, Uber are building autonomous car services. Just like you don’t clean the bus and train before you ride it you will not be cleaning the sensors on these cars. They will be stored in a hub where they will be cleaned/serviced waiting to be called on-demand arriving to you clean and ready to go.

    Regarding AI not being able to make decisions in the edge case, I would much rather trust a computer that has access to millions of previous scenarios and can calculate the most appropriate response based on the outcomes of those past scenarios than a distracted 16-year old with a bunch a friends in the car making the same decision for the first time after potentially never seeing a similar scenario before, or a drunk driver making that decision intoxicated, or ...

    There are 1.25 MILLION deaths due to cars every year. The goal of autonomous cars and AI is not to go to zero deaths but to very significantly reduce those numbers. The people that say the an autonomous AI car might have a crash are missing the point.
    gatorguy
  • Reply 28 of 33
    GeorgeBMacGeorgeBMac Posts: 11,421member
    sams said:
    ... All the companies mentioned here; Waymo, Zoox, Uber are building autonomous car services. Just like you don’t clean the bus and train before you ride it you will not be cleaning the sensors on these cars. They will be stored in a hub where they will be cleaned/serviced waiting to be called on-demand arriving to you clean and ready to go.

    ...

    There are 1.25 MILLION deaths due to cars every year. The goal of autonomous cars and AI is not to go to zero deaths but to very significantly reduce those numbers. The people that say the an autonomous AI car might have a crash are missing the point.
    For your first point:   For these cars to become viable transportation in northern climates, they will need to be capable of operating safely in ALL weather.  People are not going to wait until the snow stops the ice melts or is cleared from the cars.  Because, unlike, the southern climates, life goes on unless or until the snow surpasses 2-3 feet (not inches).

    As for your last point, I agree.
  • Reply 29 of 33
    cgWerkscgWerks Posts: 2,840member
    sams said:
    Regarding AI not being able to make decisions in the edge case, I would much rather trust a computer that has access to millions of previous scenarios and can calculate the most appropriate response based on the outcomes of those past scenarios than a distracted 16-year old with a bunch a friends in the car making the same decision for the first time after potentially never seeing a similar scenario before, or a drunk driver making that decision intoxicated, or ...

    There are 1.25 MILLION deaths due to cars every year. The goal of autonomous cars and AI is not to go to zero deaths but to very significantly reduce those numbers. The people that say the an autonomous AI car might have a crash are missing the point.
    Why are we always comparing AI cars to the worst of human behavior? If we really cared about all those deaths, we'd take the license away from those people, and/or have much, much more stringent training. This isn't really about saving lives, it's about corporate profits and some pie-in-the-sky utopian ideas of 'futurists.'

    And, even if the overall statistics were to drop (which I don't really see happening until all cars are AI... the people/AI mix is the biggest problem aspect), it may sound a bit selfish, but I care about my family and each vehicle we might possibly interact with. I don't want to be the 'oops' statistic when the Tesla semi doesn't detect my car, even if the overall accident rate is lower.

    Lowering the overall isn't a good enough tradeoff for reckless. If they want to lower that rate, as I said above, there are tons and tons of non-AI solutions that have been available for decades. Cars are safer than ever, yet the accident rates have been going up due to degradation of society. I'm not for just trying to plaster over that degradation with some technology in an area here or there... we should fix the degradation issue.
  • Reply 30 of 33
    gatorguygatorguy Posts: 23,399member
    Here's a great video posted that shows the current state of autonomy, what the car "sees" and a view from the cabin.


  • Reply 31 of 33
    cgWerkscgWerks Posts: 2,840member
    gatorguy said:
    Here's a great video posted that shows the current state of autonomy, what the car "sees" and a view from the cabin.
    It's quite the marketing spiel, for sure.
  • Reply 32 of 33
    gatorguygatorguy Posts: 23,399member
    cgWerks said:
    gatorguy said:
    Here's a great video posted that shows the current state of autonomy, what the car "sees" and a view from the cabin.
    It's quite the marketing spiel, for sure.
    Well of course there's a marketing intent. :) It does not mean it isn't accurately framed. 
  • Reply 33 of 33
    cgWerkscgWerks Posts: 2,840member
    gatorguy said:
    Well of course there's a marketing intent. :) It does not mean it isn't accurately framed. 
    It's more of an 'in-concept' take, highlighting the positives under ideal conditions. If we can't get a dancing paperclip to help us write or a talking cylinder to do anything useful, how do we expect the same technology (fundamentally), to guide thousands of pounds of steel and plastic around our streets? Maybe it would be nice to see a working example of AI in a non-life and death situation, working properly, first.

    You do know why YouTube changed their rules for paid creators, right? They needed to reduce the number of accounts to levels human curators can handle, because the 'AI' attempts failed.

    AI has some great uses and will be an interesting tool, for sure. But, in the media and public perception, it is highly overhyped. It is flawed at a fundamental level, and simply won't be able to do some of the things people are hoping it will.
Sign In or Register to comment.