Autonomous driving

What's the status on this g, is it achieved yet?

Shopping Cart Returner Shirt $21.68

DMT Has Friends For Me Shirt $21.68

Shopping Cart Returner Shirt $21.68

  1. 2 weeks ago
    Anonymous

    Sure. Except all the data goes to the manufacturer instead of the other cars

    • 2 weeks ago
      Anonymous

      Way to miss the point completely. No one cares about your pol bullshit.

  2. 2 weeks ago
    Anonymous

    Look at the tens of thousands of youtube videos of Tesla FSD 12 videos. There are also handful of Waymo videos too.

    There are ~1 million Tesla FSD cars running on the road now. They're able to do ~90% of the drives without any intervention. Often driving hours at a time in some of the videos without any human input (other than destinations or certain personal choices)

    • 2 weeks ago
      Anonymous

      So what's missing for this last 10%?is it a situation where it gets exponentially more difficult to iron those last bits out? Or is the tech held back by legal issues?

      • 2 weeks ago
        Anonymous

        For Tesla, whats missing is compute and data. That is being addressed by Tesla as they tripled their GPU training cluster from ~10K to ~35K in the year or so. Supposedly their next version (12.4) will have 10X improvements over the current V12.3. And the version after that, another 10X. Even if its only 2X improvement, it means the improvements are happening rapidly from their own internal statistical projections.

        And supposedly in ~3 months they will unveil their ride sharing fleet for public/general use. IDK if thats heavy puffing, but they seem to want to push the limits more and more towards general end users outside of Tesla drivers.

        For Waymo, they are bit limited in scale due to expensive hardware and they need to buy their fleet of cars from some other vendor. I dont know if they're able to scale, but they are certainly able to do what Tesla wants to do in the next few years. They're just limited to few dozens of cars in 2(might be 3) cities I believe.

        Ultimately I think we'll see some self driving cars on the road widescale within 3-5 years absolutely for general usage in almost every city in the US.

        • 2 weeks ago
          Anonymous

          They also have their Dojo printing training chips but they've been hush hush about it and prob wont say until their next event.

        • 2 weeks ago
          Anonymous

          >buying Tesla
          But I’m not gay though.

        • 2 weeks ago
          Anonymous

          Thank you. How the hell is Waymo more advanced than Tesla, what happened?

          • 2 weeks ago
            Anonymous

            Doesn't Waymo use LIDAR? Tesla is just using cameras and proximity sensors.

          • 2 weeks ago
            Anonymous

            LIDAR is garbage. Thats why Tesla isn’t using it.

          • 2 weeks ago
            Anonymous

            Tesla does not use lidar because they want to save money. Some redundancy on challenging conditions would be a no brainer for most other companies

          • 2 weeks ago
            Anonymous

            Its nonsense. If you want to use lidar on "some conditions" then you still need to work on it so that it works 100% on those challenging conditions such that it also works on other conditions. Which also means you now have 2 systems that you need to invest in and work on and have to power and have to compute and so on. Its inefficiency all around. Might as pursue one horse to bet it all but devote all the resources to that horse.

          • 2 weeks ago
            Anonymous

            Wut, lidar and cameras complement each other. Even the chink designer of my robot vacuum managed to get them to work together. Are you fricking moronic? Have you never heard of redundancy? You need one (1) sensor to fail and your car is completely blind in one direction. Or the sensor does not fail in an obvious way and misinterprets the size / distance of an object. Or just gets dirty. Would be cool if we could do something to mitigate this risk. Sry bois, lidar is too hard.

            Might as well put your NAS in raid 0, then you don't have to worry about restoring your data when one drive fails. Also saves on computing time to start from scratch.

          • 2 weeks ago
            Anonymous

            The all system design has trade offs.

            1) power consumption
            2) resource limitation in training
            3) compute inefficiencies
            4) latency increases
            5) noise
            6) reliance complexity
            etc etc

            The leading self driving car company choose one way and thats the reason why they're leading the race.

          • 2 weeks ago
            Anonymous

            Further, if

            Wut, lidar and cameras complement each other. Even the chink designer of my robot vacuum managed to get them to work together. Are you fricking moronic? Have you never heard of redundancy? You need one (1) sensor to fail and your car is completely blind in one direction. Or the sensor does not fail in an obvious way and misinterprets the size / distance of an object. Or just gets dirty. Would be cool if we could do something to mitigate this risk. Sry bois, lidar is too hard.

            Might as well put your NAS in raid 0, then you don't have to worry about restoring your data when one drive fails. Also saves on computing time to start from scratch.

            can prove out that its better to do with multiple sensors, you're free to create a company to create the system you believe is better. Or work for the company that you believe is doing something wrong.

            All you gotta do is show your hard work rather than just guesstimates.

          • 2 weeks ago
            Anonymous

            Dont worry, many small startups like Mercedes Benz, BMW, Audi, Ford and Volvo are already doing it just like this.

          • 2 weeks ago
            Anonymous

            Yeah. Did you see this?

            Mercedes drivers assist that is ranked "higher" than Tesla (according to consumer reports) vs Tesla.

          • 2 weeks ago
            Anonymous

            Most recent Tesla vs Waymo comparison. For comparison between the two leading companies. Similar performances.

          • 2 weeks ago
            Anonymous

            >41 interventions

            German technology lol

          • 2 weeks ago
            Anonymous

            In Germany we are fricked. They will never allow Tesla FSD until German cars can use it.

          • 2 weeks ago
            Anonymous

            Its not just German technology, but rather thats how the competition is. If you look at Ford's Cruise, it cant handle slight curves on highways. Same with GM's whatever the frick they're doing.

          • 2 weeks ago
            Anonymous

            is this fake?

          • 2 weeks ago
            Anonymous

            Its a guy trying to test Mercedes system that consumer reports rates as 10/10 and the company advertises as having "Level 3" while Tesla does "Level 2".

            It cant keep itself in lane, let alone avoid obstacles

          • 2 weeks ago
            Anonymous

            note that the "drive assist" they tested is level 2;
            the one advertised as level 3 is "drive pilot".
            I'm not sure their comparison is legitimate

          • 2 weeks ago
            Anonymous

            Well, the comparison is on real world usage. No matter what the "labels" are or the "ratings" are.

            Thats the real legitimate comaparison, not the irrelevant ratings.

          • 2 weeks ago
            Anonymous

            What is the leading self driving company? The one that can drive on its own in limited scenarios or the one that always needs constant babysitting?

          • 2 weeks ago
            Anonymous

            Both are leading edge. But of different constraints and scope. Limiting scope means you dont need to baby sit as much. Babysitting means you have a wide open world to experiment with in which something wrong might happen for whatever reason the user chooses to put the car through in whatever conditions.

            Those dont takeaways from either company but rather show the current limits of the technology. And Tesla seems to be reducing the babysitting with the next update, as well as removing the babysitting all together with the public ride share release in few months.

          • 2 weeks ago
            Anonymous

            Waymo uses LIDAR + Camera + Radar + some other sensor fusions.

            Tesla originally used camera + radar fusion, but they didnt see the use of radar and it simplifies training, reduces cost, reduces noise, increased accuracy, increase reliability, reduce wear/tear, reduce power usage, etc. So they went with camera alone. They have ~9 or so cameras total, and each one prob costs ~$2-3 each totaling maybe ~$30.

          • 2 weeks ago
            Anonymous

            Doesn't Waymo use LIDAR? Tesla is just using cameras and proximity sensors.

            Also they dont use proximity sensors anymore. Radar used to function as a way to gauge distance/speed of cars ahead but now their neuralnet predicts all that from cameras. Range, speed, dimension, etc.

          • 2 weeks ago
            Anonymous

            Waymo isn't "more advanced" per se, but its way more restricted and thus the possible scope is limited by factor of 100. This allows Waymo to drive coherently in their bounded realms. The cities and the corridors that Waymo operates are very limited with limited interactions.

            Sort of like a 5 lb fish in a 1 cubic meter fish tank. Its the king of that fish tank but if you release that fish in the ocean, it might get eaten by the next bigger fish. But if you dont release it from the fish tank, its king of the fish tank.

            Tesla is the 5 lb fish that can cross the ocean, rivers, any fish tank, and have a high 90% survival rate.

          • 2 weeks ago
            Anonymous

            Also if Tesla's robotaxi fleets limit the mapping to easy roads (avoid strange curves/roads with multiple confusing lanes/many underbridge pillars in inner city/etc) they should be easily able to surpass Waymo's capability.

        • 2 weeks ago
          Anonymous

          ETA on when I can safely and legally get in my car, punch a GPS destination 2 hours away in, take a nap in the back, and wake up on arrival?

          • 2 weeks ago
            Anonymous

            imo, anywhere from 1-2 years.

          • 2 weeks ago
            Anonymous
          • 2 weeks ago
            Anonymous

            Same as always; five to ten years.

        • 2 weeks ago
          Anonymous

          There's no way you actually believe this lel. FSD is always 3 years away.

        • 2 weeks ago
          Anonymous

          sounds like you swallowed the marketing with a smile on your face, anon. self-driving cars have been "five years away" for a long time now.

          • 2 weeks ago
            Anonymous

            All you gotta do is open your eyes wide to mounting evidence from actual users. If you dont want to change your stance even with mounting evidence, the problem is you.

          • 2 weeks ago
            Anonymous

            call me in five years then

          • 2 weeks ago
            Anonymous

            Why? Its already happening in front of your eyes. You sound more like those AI truthers

          • 2 weeks ago
            Anonymous

            don't call me then, it'll help you save face

    • 2 weeks ago
      Anonymous

      >hey! just look at all the videos where cars are driving in highly trained straight roads with little to no variables!
      Sounds like future to me!

    • 2 weeks ago
      Anonymous

      I also want to add to OpenPilot to the mix for the current state of the art length as well.

      Its more advanced than any other non-Tesla and non-Waymo capability.

      It costs ~$1K($2K?) or something and most are able to mod their cars easily. They also support many of the newer cars from the past 6-7 years.

  3. 2 weeks ago
    Anonymous

    >nyoooom

  4. 2 weeks ago
    Anonymous

    member, change of cancer goes up 500% whenever you partake in car traffic.

  5. 2 weeks ago
    Anonymous

    https://x.com/GoingBallistic5/status/1794936716156903931

    Small video with analysis captions of the Tesla FSD's dynamic decision making process. Its wild.

    • 2 weeks ago
      Anonymous

      They are training their shit end to end with human input. When it produces behavior that is humanlike, some people think it has a robust understanding of the scenario. Just like some people thought LLMs think and reason because they produce humanlike text and can answer many questions.

      Even when it goes below human deaths per mile driven, it still won't be a good deal for the people who aren't in the habit of doing dumb shit while driving. The death rate for attentive non moron drivers is low. The dumb fricks constantly trying to exit the gene pool significantly affect the average rate.

      • 2 weeks ago
        Anonymous

        If you think of AI as a second driver sitting in your seat, thats all you need to act as. Trust but verify.

        But really, it depends on scope of the drive in which you're willing to allow yourself to trust the AI. If your destination is small town neighborhood road where the destination is only ~3-5 miles and the speed is ~20-30 miles. I'll absolutely take the robot taxi (after a year or two of it being on the road and testing) and trust it enough. The risk is very low anyway in these conditions. Even if a 30mph impact happens, its still only 30mph impact. Ironically, its much safer on the highway due to limited scope of driving making it much more reliable. But the statistical accidents/error damage is much higher.

      • 2 weeks ago
        Anonymous

        People overestimate their own reliability and attentive non moron drivers aren't that common

  6. 2 weeks ago
    Anonymous

    >Use your autonomous car
    >get cooked by micro waves on the road from other cars
    This is how they're dealing with overpopulation, isn't it?

  7. 2 weeks ago
    Anonymous

    Look up trains

  8. 2 weeks ago
    Anonymous

    This is not going to be a thing, until governments adopt some sort of a standard and we get smart roads and cars that are capable of making use of these smart roads. Autonomous driving in a dense urban environment with fricked up roads and people running around, is never going to be 100% safe, while highways that work like railways and you can 99,% percent trust your car and sleep while it drives you, should not be too hard to achieve. Just that, we need a system where all the autonomous cars are on the same page.

    • 2 weeks ago
      Anonymous

      ~20K humans die each year in accidents in US.
      ~5 million car accidents happen each year in US

      1) There is no 100% safe system
      2) Government never solves problems, it always creates problems
      3) innovation doesn't happen with centralized government mandates, it happens when people pursue their own self-interest/hobbies/goals/etc and work with others to create their vision in mind into reality

      • 2 weeks ago
        Anonymous

        >2) Government never solves problems, it always creates problems
        Thanks for giving me a hearty laugh.

  9. 2 weeks ago
    Anonymous

    Tesla FSD v12.3 has been a huge leap apparently. Let's see what 12.4 brings. Elon claims that it is 5 to 10x better.

    Also 8/8 Tesla has an event presenting the Robotaxi

  10. 2 weeks ago
    Anonymous

    I
    HECKIN
    LOVE
    TRAINS
    WITH FOUR WHEELS
    YAY SLAY QUEEEEEEEEEEN

  11. 2 weeks ago
    Anonymous

    They are focusing on it wrong. The future is bot carriages.

    • 2 weeks ago
      Anonymous

      Real horses are FSD already.

      • 2 weeks ago
        Anonymous

        Especially if their name is Roach.

  12. 2 weeks ago
    Anonymous

    Department of Transportation commissioners will shut it down when some angsty kiddo takes control. Expect a recall on your self driving targeted assassination machine. Don't worry, nothing for you to do ... it will self drive to the impound before too many people get Anne Heche'd.

  13. 2 weeks ago
    Anonymous

    Can we just have a good public transportation system please?

    • 2 weeks ago
      Anonymous

      N

    • 2 weeks ago
      Anonymous

      Your city taxes, your choice. Won't be useful in my rural area.

    • 2 weeks ago
      Anonymous

      If it is segregated.

    • 2 weeks ago
      Anonymous

      self driving guys are the ultimate in public transportation
      >direct A to B
      >don't have to share your air with undesirables
      >can bring more shit with you at once
      >at will convenience rather than tethered to a schedule
      The ultimate goal is basically the taxi / uber system with a self driving fleet all communicating with each other to optimize traffic. you open your app / plan a destination and a car shows up very soon after (or immediately if pre-scheduled)

      • 2 weeks ago
        Anonymous

        >skiddies detect target hired ride
        >disable vehicle safety limits
        >disable logging, clear ram
        >accelerate to 160mph into telephone pole
        >oopsie musta been stuck throttle

        • 2 weeks ago
          Anonymous

          as opposed to them doing that with the subway you're on? or the bus driver having a schizophrenic break? It's PUBLIC transportation, not private

          • 2 weeks ago
            Anonymous

            Yes all of those are risks too. I moved far away from cities to avoid that cancer, homeless bums trying to excrete fluids on me, crazy people starting fights that I am not allowed to stop. Thankyou fren.

  14. 2 weeks ago
    Anonymous

    >HONK-HONK
    >BEEP-BEEP
    >HONK HO-HONK
    SEATBELTS EVERYONE!
    >Please don't let this be a Tesla Trip...
    WITH THE MUSK?! NO WAY!!
    >OOOhhhh!!
    (Crashes and burns, no survivors)

Your email address will not be published. Required fields are marked *