Guy whose cars run into stopped fire trucks thinks he’s an expert on computer vision.
Someone died because one of those cars thought the broadside of a white semi trailer was the sky and drove under it
Speaking of fire trucks has anyone here ever read the emergency response procedures for teslas in severe accidents? When I was a volunteer we gave it a look over.
If I remember right, Depending on the model they recommend up to 8,000 gallons (~30k liters) to keep an overheating battery’s temp stable in case of fire or exposure to high heat. I’ll link the resource page here.
Our engine holds 700 gallons (5.2k liters) and the typical tanker in our area holds 2,000 (7.5k liters)
That’s a house fire level response for a single electric vehicle. Just getting that much water moved to a scene would be challenging. We have tankers, but how many city departments can move that much water? You don’t see hydrants on highways. And foam is not effective like it is for normal car fires. The future will be interesting for firefighters.
I found a link on how the Austrian fire workeres handle this. The fire is extinguished first, then the remainders of the car are put into a special roll-off container (Abrollbehälter, AB) and driven to a gravel pit, where the container will be flooded with 21000 litres of water.
That’s interesting. Tesla says the cars shouldn’t be submerged but I wonder if there’s any serious consequence if you did?
30,000 liters is 30m^3, which is a back yard swimming pool full of water.
Now imagine a house on fire with a tesla in the garage or multiple vehicle accidents. Now you need that much more
It is especially important to understand that Tesla’s struggles with navigation are entirely a result of Elon refusing to equip them with LiDAR. This isn’t some “The tech is really new and really complicated and we’re still figuring it out” problem. There is a very good solution to most collision avoidance scenarios, but Elon refuses to let them use it because he’s an idiot.
They also sometimes lock people inside and burn them to death.
https://www.mirror.co.uk/news/world-news/terrified-friends-burned-death-tesla-34087725
For those doing the maths at home:
An F35 who obligingly flies top-towards-you (not exactly something you can do, but hey, maybe they’re turning) is all of 10m tall.
An AIM-120C can very comfortably hit a target at 100km.
At that range, the F-35 takes up 26 arcseconds, or 0.007 degrees. That’s roughly about the size of this period, at a distance of 3 meters away.
[ . ]
Good luck spotting that in a sky of roughly the same colour, full of other objects.
You can place cameras anywhere, they don’t need to be right next to what is being targeted. Nearer ranges will allow AI to misidentify at much higher rates than max standoff ranges of an AIM-120C.
Pffffffff
I can see that bright white dot against the dark mode background on my maximum brightness screen with ease! Therefore your argument is invalid!
Yeah but what about the AI? Have you thought about the AI that would be running it, which never misses, and would totally be a useful existing thing? 😉
Just for reference: JWST has an optical resolution of 0.07 arcseconds. It’s a mirror 22 feet in diameter though, not something you’d put inside a missile guidance package.
JWST operates in space, i.e. there is no atmosphere blurr to take into account.
Oh yeah I’m not suggesting we make a missile with JWST mounted on the front!
Sir, our air defence is down!
Is it hackers?
No sir, it’s cloudy.
If a fighter jet is within visual range of a camera, it’s already too late. And that’s if there aren’t any clouds.
your not thinking like a musk, not if the government pays the subscription and contract for his early warning camera drone balloon swarm thing or something something they could run on ketamine or something.
Part of the reason air defences mostly rely on radar and other parts of the electromagnetic spectrum from at least 3 locations using triangulation to build a precise map of objects in the sky, but just like cameras that doesn’t work when the objects in question are too high or hidden behind objects. From there you can send countermeasures to intercept coordinates and then arm them to search for nearby objects via infrared.
Using the visible part of the electromagnetic spectrum is pretty much useless in modern weapons. I remember seeing even a Tank operator’s display being totally jank because they don’t use normal cameras either, perhaps because they wanted data to train machines to do it instead of human operators? Idk, didn’t make sense to me.
His fucking obsession with computer vision. He’s so convinced he’s right he forgot that clouds exist… and his cars plow straight into obstacles.
Yeah, the “lidar is useless” guy whose cars are consistently crashing into things when visibility is bad is telling us that he can do the same thing with missile targeting systems… Sounds like a great idea
And that a plane at altitude is too small for wide field cameras which means scanning the sky with narrow fov detectors.
And F-35s are really fast. By the time you recognize and can target it, it’ll fly behind a cloud or something. So not only do you need to make a really fast rocket w/ vision-based AI integrated, it also needs to be able to detect said plane at great distances, as well as maneuver well enough to see it as it exits clouds and whatnot. That’s a lot more complicated than slapping radar on something with heat tracking at close distances.
My friend bought a red car specifically so it could be seen by Tesla’s cameras.
No no. See guessing objects from flat images is much better than using math and lidar. Especially if you may have a flawed llm model.
Given how advanced our math and knowledge of radar is, it is literally stupid to use them.
See, those, radar, lidar and math give you a 3d objects.
Oh, wait. It is the other way around.
He’s not, otherwise he would know that “low light sensitivity” cameras aren’t “sensitive in low-light conditions” but “with lower than normal light sensitivity”.
In an imaginary world where cameras are way more expensive, he’d absolutely be pushing LiDAR in cars. The metrics he cares about are cost and marketability (cool factor), or money for short.
