Polestar crash

[Diehard]

also why would the road go there if it wasn’t?

You underestimate the power of focus, determination and undying commitment to the goal some have. Growing up Wile E was always my hero. He is like Elon for kids.

 

[Crissa]

Also the system double checks with the mapping data that there is actually a tunnel there (also why would the road go there if it wasn’t?).

...Because tunnels open faster than maps are updated?

My car still thinks I'm taking a shortcut over the mountain through the Tom Lantos Tunnels.

-Crissa

 

[LoPro]

...Because tunnels open faster than maps are updated?

My car still thinks I'm taking a shortcut over the mountain through the Tom Lantos Tunnels.

-Crissa

I know that. I’m apparently driving through a mountainous woodland wilderness on my way to the cabin too and the autopilot begrudgingly accepts the markings and the cameras in that I am in fact driving on a new highway, but it freaks out now and again based on map data of old roads which used to cross the highway. The new highway has several tunnels by the way.

There’s not usually an open highway going straight to a mountain face meant to be tunneled later though. In my experience there’s an actual older and mapped road to follow until the tunnel opens. But when it opens the car may freak out half thinking you’ve left the road (“half thinking” as in at least taking it into account, not that map data is a dominant variable at all), especially if there are crossing mapped roads now missing in my experience.

 

[carpedatum]

It's not even necessary to use multiple cameras to achieve high dynamic range. The same camera can take successive images at different ISO's by adjusting sensor gain instantly between frames. The human eye cannot do that. So the result would be a video of interleaved frames at different ISO's. The images would be processed to create all frames with super high dynamic range.

I tested the autopilot on a curvy country road on a dark night and was super-impressed. When I turned the headlights off it could see the road edges better than I could. The forward cameras can also see better than I can when driving into a blinding sun. I keep my lenses/windshield pretty clean because this is not the case when there is roadgrime in front of the lenses.

Yes, and consider that there is no particular reason for a "camera" to simply produce a human-viewable image. It isn't that expensive to go somewhat beyond the spectra visible to humans, and also supply a neural net with rich contextual data points like time of day, date, lat/long, simple meteorological data, etc. There is no particular reason that a 'net can't anticipate shadows, understand the probable temperature (e.g. range of IR signatures) of an obstruction vs an optical illusion, etc. With a sufficiently well-rounded data set, even absent maps, a camera-based machine might do better than a human at "seeing" things it should avoid, particularly at night.

 

[HaulingAss]

I am not sure about Polestar but it makes me wonder how a Tesla FSD that works based on vision/light alone deals with shadows. In a time of the day like this and sky like this as clouds move in and out, horizontal shadows keep appearing and disappearing. There must be a contrast threshold that FSD uses to decide if the shadow is an object or not. I never gave this stuff any thoughts before but it is fascinating. If any of you have worked on the stuff feel free to dumb it down for me and break it down to it’s pieces (how computer vision works in this context).

I recommend that anyone wanting a better understanding how machine learning and machine vision could possibly work, and work well, learn the basics of the subject. As you gain an understanding off what machine learning really is, and how it mimics human learning, you will understand how it will change the world within most of our lifetimes.

Lars of "The Best of Tesla" Youtube channel just realeased a very good introductory video on exactly this:

How AI learns & because of that Tesla has already WON | Tesla is playing the END GAME - YouTube

This is a great primer on the subject.

 

[FutureBoy]

...Which will have the same problem any human viewer would have.

Heck, there's record of humans driving their car into said paintings on walls.
-Crissa

Quick search gave me the story below. But that is the only one I found right away.

Real-Life Wile E. Coyote Drives Fiat Smack Into Painting Of Tunnel

Jason Torchinsky
3/18/16 2:46PM

This is so achingly beautiful, it makes me happy just to be alive in the same world where this exists. It seems a graffiti artist painted one of those Coyote v. Road Runner-type murals of a tunnel entrance on a solid wall, and someone actually drove into it. The only way it could be better is if the car was found to be full of ACME products.

This seems to have happened back in December, and seems to be first documented on an Imgur post by user rburn, with the succinct caption “so this happened.” The three images show the wall with the tunnel mural, a red Fiat parked nearby, then a close-up of the smashed front of the red Fiat, then a team painting over the mural.

There was even a Road Runner painted next to the mural.

This clearly isn’t from the US, and seems to be from Brazil, thanks to what some informed commenters tell me. There hasn’t been much more information about exactly what happened, with most stories about the event simply referencing the Imgur post. I haven’t found much more information either, to get details or prove this is what actually happened, or anything,

But I’m not sure how much I care. As long as no one was hurt, I sort of want this to be true. The picture told in these stories is one of a glorious, messy intersection between the rules of the cartoon world and the rules of reality.

How the hell could someone have done this? I mean, sure, a half-starved coyote, mind already addled from so many rocket-assisted concussions, sure, I get how he could mistake a painted-on tunnel for the real thing when approaching at high speed, but a human capable of driving a car?

Certain sources I found have suggested that the driver did suffer at least some injuries to their spine, which caused the spine to take on a springy, accordian-like shape, resulting in a series of musical-sounding wheezing noises emitted with every step walked.

 

[Bond007]

Also the system double checks with the mapping data that there is actually a tunnel there (also why would the road go there if it wasn’t?). If the tunnel is closed for some reason the car can get real-time traffic data and see the warning lights and signs.

The issue highlighted was a real problem as mentioned above.
It would absolutely make sense for the system to double check these with mapping and real time vehicle data. That would make such pranks useless (or will only target human drivers).

 

[HaulingAss]

I think that case can be resolved without radar with enough light and resolution. The same way we can tell the difference. With a camera on right and one on the left you can see if any pixel inside of the tunnel change position with respect to outside pixels as you get closer. Daylight or vehicle headlight should give enough light to make enough of inside visible.

Well, I certainly hope it will be able to distinguish a well-painted tunnel from a real tunnel! Because, depending upon the route I take, I drive by 3 or 5 realistically painted tunnels every day! One of them always almost causes me to drive through it (even though I have no idea where it would go if it were actually real).

 

[Diehard]

Certain sources I found have suggested that the driver did suffer at least some injuries to their spine, which caused the spine to take on a springy, accordian-like shape, resulting in a series of musical-sounding wheezing noises emitted with every step walked.

My mom told me never take pleasure from someone else's misfortune but my dad said, if they have already had the misfortune, it would be such a waste not to take pleasure from it. The world would be a lot less miserable place as a whole if we all do. That is when I became an activist trying to make the world a better place by looking for people in trouble ( Dad style )