Почалося, росіяни хочуть обдурити ударні бпла СОУ.
While many of you are shouting “Go Ukraine” or “Go Russia”, I am shouting “Go, the return of dazzle camouflage!”
[tweaking the algorithm to target black and white boxes]
The limeys did a bunch of shit like this on ships in World War II too confused bombers or whatever. It’s actually really cool, I saw it on Reddit I can’t find it now maybe I could I’m not going to though.
That was referred to as “dazzle camouflage” and the idea was to throw off the aim of German U-boats by making it more difficult to determine the size and orientation of the ship from brief periscope observation. They would them not be able to compute the correct fire control solution for their torpedos.
To add to this, not just size and orientation! It also made it more difficult to determine the speed (and sometimes it was hard to tell if it was moving at all)
Is this what capthchas have been training AI for?
Wait, I’ve seen that one before!
it worked well back then, there are no documented destructions of warships by autonomous drones during world war 2!
Do birds count as AI? Avian Intelligence did sink some ships
birds aren’t real, or so i’ve heard! ^^
Did you ignore the V2 (rocket) because it didn’t destroy a warship (I didn’t fact check this, but it seems likely), or is your definition of an autonomous drone very restrictive?
Torpedo also should count as an autonomous drone (some were wire guided and not autonomous) and they were sinking warships well before WWII.
I would rather say that your definition of “autonomous” appears rather wide? I wouldn’t call e.g. a heat-seeking or radar-guided missile “autonomous”. I would say that “autonomous” implies the drone/missile/torpedo/etc. can somehow adapt to changing conditions rather than follow a simple pre-programmed path.
You can have very simple “autonomous” systems, for example “Move towards this GPS coordinate, unless you lose connection, in which case fall back to inertial guidance until you either reach the target or re-acquire connection”. Another example could be “Home in on the heat signature, unless you lose the signal, in which case execute some planned manoeuvres to re-acquire the signature”. Yet another could be “Move towards target, but if you detect that you’re being tracked by radar, or detect an AA launch, conduct evasive manoeuvres / move closer to the ground, etc.”
All the above imply that the drone / missile / torpedo / whatever is capable of responding to changes in the surrounding environment. To my knowledge, there were no WWII-era weapons capable of doing that. A typical WWII-torpedo was “Start the propeller and keep going at maximum speed until the detonator is triggered or you run out of fuel and sink.” I can’t really see how that would count as autonomous.
“Start the propeller and keep going at maximum speed until the detonator is triggered or you run out of fuel and sink.”
That is a very simple program. Heat sinking is modifying behavior.
I wouldn’t say things like proximity triggers or primitive homing tech like heat seekers are autonomous. Heat seekers don’t adapt to changing situations, but follow a completely mechanical “go towards the warmest spot” path. Being autonomous would mean the could react to the “warmest spot” either disappearing or moving in an unphysical way (suddenly appearing somewhere it’s shouldn’t be, as can happen with e.g. flares).
Basically, if the weapon has a single thing it can do (move towards hot thing), and no way of adapting if that thing doesn’t work as expected, I have a hard time calling it autonomous.
I wouldn’t call a simple robot-vacuum autonomous either for that matter. If the instruction set is “go forward until you hit an obstacle, then rotate 15 degrees clockwise and repeat”, I don’t really see that as “adapting to changing circumstances”.
The behavior stays the same during heat seeking (go towards the heat). The direction might change as it keeps to it’s behavior (since the heat moves)
Autonomous means having the ability carry out task and adapt to new information. You are ignoring the second part.
fair enough, it was meant as a joke, but your definition of autonomous definitely qualifies ^^
Torpedoes really revolutionized Naval Warfare. I remember reading about the Japanese figuring out to use oxygen as the propellant and got way better results right before World War ii.
I believe the Russians have a nuclear torpedo. What the fucking good that is I don’t know suicide mission for the ship.
Like Dazzle camouflage from WW1 and WW2: https://en.wikipedia.org/wiki/Dazzle_camouflage
I wonder if it works…
Dazzle paint. Everything old is new again.
Does it work? Is this like drawing a mustache on a kid? I want to believe it couldn’t work, but if they’re using AI to detect trucks then who the hell knows?
I remember when facial recognition technology began to be mass deployed, it was possible to trick those algorithms with special makeup. The idea of this camouflage is to trick ai drones in the same way. But it will not work with modern ai technologies, or if drone uses other targeting system like thermal cameras, or if drone simply being operated by person.
Around like the zeros to 2010 at least the facial recognition was pretty lacking. A band aid on the face could defeat it.
Nowadays though it’s gotten better and they do other stuff by like analyzing your gait, obviously the eyes if you get close enough, speech patterns.
Basically we just have to stop these people that control the information because they have too much of it. But that is another story.
But not long after 2010 I seen recall a bunch of fugitives getting caught and I am willing to bet it was because of facial recognition.
I have no idea, but I have some supposition: no it won’t work.
Like others have said warships used to do something similar.
The naive idea is a WWII concept: that the hard lines make it more difficult to see your heading from a distance.
The problem is drones don’t “see” the way people do.
We could trick the drones, they would be reprogrammed to be untripped and maybe a couple months or less I’m sure.
Also beyond not seeing the way people do, the drones are constantly doing range finding and tracking, so they can adjust their trajectory on the fly. The WWI shells and torpedos didn’t do that.
Looks like two different attempts could be testing designs.
It depends largely on which sensors the AI was trained on. Could be a combination of RGB+I, LiDAR etc. If only RGB, this is problematic for object detection.
Most of these longer range drones utilize infrared vision. The camo is not gonna do shit. There are generally a lot of misconceptions around these AI drones in the Russian army, the main one being the alleged face tracking capabilities from that one video.
First turtle tanks, now zebra trucks.
Next: Chameleon cars. Or maybe not. Russia doesn’t have the money or tech for it.
With Buckyballs it could be possible to make an invisibility cloak, but prohibitively expensive, and russia treats their soldiers like cannon fodder they don’t give a fuck.
But this type of magnetic carbon they can make, referred to as Buckyballs sometimes, is really interesting, it has so much potential. You could make a suit to reinforce your muscles and give you superhuman strength, you could make it to read the lights on one side and recreate it on the other to be invisible. Really amazing stuff. I don’t know if there’s been any developments on this in the past 20 years or whatever.
Nice of Russia to barcode their inventory for better identification. Wouldn’t the hard black and white edges make it easier for digital identification?
Maybe not, if the drones aren’t looking for big zebras. Just depends how the AI is identifying targets.
