How far away are we from small autonomous combat drones? Considering the pace of development of civilian rotor drones and of navigating AI seems like it should be too long: cram the brains/sensors of a self-driving car into a quadcopter and stick a gun on it.
Seems like the ISIS problem, for example, would be pretty quickly sorted with technology like that.
>>28086477
We're a long way away from functional AI even on the most powerful supercomputers. Autonomous drones are a fucking dream right now.
Fuck off with your shitty thread.
>>28086574
You don't need an "AI" that can pass the touring test. You just need one that can navigate around without running into stuff, and identify stuff and people (so it knows what to shoot and what not to shoot).
Seems self-driving cars are pretty much there already. It's a matter of miniaturization at this point.
>>28086672
The problem is that a self driving car doesn't have to decide who dies. The navigation systems and such are well developed, but I'm not sure how well they can function without God uplink since that can be jammed (though probably not by Isis at the moment)
Autonomous drones would have to rely on far far better target identification tech than we have now, and at least in the US, they probably would have to check with operators before doing anything.
>>28086672
>identify stuff and people
That's the big problem, self-driving cars can't do that, even people have problems judging who to shoot.
...if you think about it, there's nothing particularly complex about it given modern tech.
Re: navigation and environmental awareness, these self driving cars (and also e.g. the Microsoft Kinect) have 3D scanners that give them a 3D model of their surroundings in real time.
Re: identifying stuff, with enough data to 'neural network' mine (which we have because of the internet) computers can become pretty good at picking out what stuff is. Feed the system thousands of photos of jihadi militants and it will develop a good sense of what one looks like (ditto for every other object it might encounter). The final kill order would probably still be left to a remote human operator.
>>28086732
>Re:
Just go back to reddit already.
>>28086704
I don't see that as a big impediment. If the drone sees what it thinks might be some militants (because they are in a suspicious area, are holding guns, are simply a bunch of youngish men together, etc.) it would alert an operator who could say "yeah, that looks like our targets, kill em". Or that information could be passed to high altitude drones with bombs or whatever.
But having camera drones whizzing around the battle area at all times of the day and night would seemingly make it impossible for a guerrilla army to operate.
>>28086477
I can see bomb/assassin drones within the next 5 years.
>>28086768
That's true, but that's a different application than what the OP said.
I would say it's more semiautonomous than autonomous.
Also I feel like camera drones would have to be really cheap if you used them. People don't like being watched even if they've done nothing wrong, so civilians will likely throw stones.
Unless you can make them extremely unobtrusive .
Also it would have to be very accurate to keep the gurrila army from growing
>>28086871
Missiles.
Those are called missiles.
>>28086909
I think he means drone's that would actually loiter for long periods of time before detonating. Missiles can't really do that.
>>28086895
>People don't like being watched even if they've done nothing wrong, so civilians will likely throw stones.
When they start getting shot because they're assumed to be in league with the enemy because they're attacking the drones the civilians would stop doing that.
>>28086574
Simple AI is pretty simple.
You give it a transform, Which is the locations you want it to patrol.
Give it a simple if statement such as if anything moves shoot it.
AI
>>28087071
So now we're shooting people for throwing rocks at robots?
>>28087090
>if anything moves, shoot it
This can only end well.
A small spy/combat rotor drone wouldn't and couldn't be flying all the time (it wouldn't have the power, and it'd make itself a big target). It'd just find a spot with good sight lines and hang out for long periods not flying.
>>28087096
When wouldn't you be shot for throwing rocks at military equipment / installations in a warzone?
>>28087103
>>28087090
Remember kids, He said simple. Unsimple would be give a list of checks.
Is this person my team? Yes. Then don't shoot them.
Is this person my team? No. Then shoot them.
I can't identify if they are on my team, fire warning shot near them and stand ground while calling for backup.
>>28087128
>>28087136
Its pretty simple to detect if something is a rock, so that would be easy to program out. you would then give it a heat sensor along with a few other ones to check if its living or electrical.
Threat checks to see if its a grenade or bomb, etc.
That's the most basic, but your assuming thats the final model.
>>28087128
Hearts and minds man. Hearts and minds.
>>28086704
Actually, that makes me wonder what self-driving cars would do if they were in a situation where an accident was inevitable, but depending on how they maneuver, only one is likely to survive the crash.
>>28087128
Pretty much all the time.
Or else kids would get absolutely massacred 24/7 all up and down the Gaza strip and during the Iraq occupation.
You seriously think it's standard procedure to slot everyone that tosses rocks?
>>28087189
Activate enough built in thermite charges so that the vehicle couldn't be identified and they don't get sued.
>>28086975
>>28086909
Drones and explosives are cheap(ish). Guerrilla forces, cartels, murder for hire, shit like that would be more probable I think.
Those muslims in LA made a nigger rigged RCXD for shits sake.
>>28087189
There is a lot of debate on that right now. The best solution at the moment is to have all the cars on the road communicating to avoid a situation like that happening. (To my knowledge anyway)
Without that network, the Google cars are extremely cautious and polite drivers
If it's a pedestrian getting hit, the car will do its best to keep the pedestrian safe.
Much cheaper and probably as or more effective would be to build 'sensor mines' and litter them all over the place from the air. Cover them in (ir) cameras all around and 3d (triangulating) microphones. You could build them out of commodity phone hardware for next to nothing.
It'd give you insane situation awareness: you could essentially bug/cctv entire countries for the air on the cheap in real time.
Targets identified could be taken out from the air by conventional means (drones, attack aircraft, cruise missiles, etc.)
>>28087246
I heard of a concept the military had for something even better (or more sinister, depending on your perspective). They called it smart dust, basically sensors so tiny they were almost like sand that covers everything and gets into structures.
>>28087303
That's obviously a LONG LONG ways away in the nano-bot future. Something like >>28087246 you could pump out for under $200 a pop most likely.
>>28087345
>under $200 a pop
>IR cameras
wow no.
>>28087345
I think it's a lot closer than that. The sensors in question weren't ACTUALLY the size of grains of sand, more like 4-6mm across. On their own, they couldn't see shit, but together they could get an idea of what was going on. Kind of the same way insect eyes work.
>>28087362
Any cheap phone camera sensor can do near IR. You can get IR security cameras for $10 these days. Although I think you need an IR light source for them to be useful, or pointed at something really hot.
Quality real thermal cameras are coming quickly down in price, and that's with pretty limited market demand. What was +$5000 a few years ago is now $100-$200. With a volume order you could get those prices way way down I'd imagine.
>>28086477
Combat drones will never happen outside of bombardment platforms like the Predator.
The problem is lag time. No matter how advanced communications get, there will always be some lag when communicating wireless over long distances. So drones are only really viable as surveillance tools or precision strike weapons. You couldn't dogfight with a drone, it's just not responsive enough.
>>28086477
How would an autonomous drone distinguish between friend and foe?
>>28087555
Are we talking cameras that can see IR light or are we talking thermal imaging?
There's a huge difference and you seem to be using them interchangeably.
>>28087662
If you're talking about in an area with friendly combat troops, that's super simple: just have your troops wear "friendly" transponders that drones will pick up.
The much much harder part is having them distinguish between enemies and civilians.
>>28086704
Would IFF transponders for friendly troops/vehicles be a good solution to this?
We would need very, very reliable transponders, and constantly drill the troops on making sure they have them on their person at all times. Also there would be a threat of enemies taking the transponders, possibly have them be able to be remote deactivated, or rely on periodic "check ins"?
Obviously this would be vulnerable to good EW, if fighting against say China, but I think it would be a viable solution for initial occupation of relatively primitive areas (such as Mexico), where you don't want to have natives remaining alive to start insurgencies.
>>28087695
I'm not using them interchangeably (although technically thermal imagine is just very sensitive IR imaging). It would be nice to have real thermal cameras, but they're expensive and low-rez (but like I said improving a lot in both regards).
Still, especially with computers doing the interpretation you can do a lot with simple light amplification / near-infared 'night vision'. Enough anyway that with some moon / ambient light the cameras would still be useful at night.
>>28086975
RC bomb car.
Best idea a 13 year old ever had.
