Hacker News Reader

Comment by freediddy | Hacker News Reader

The biggest security threat any country has is if an adversary sends 1-10 million drones at once, each with a small grenade on it, and overwhelms a city. They could literally target individual politicians or weak spots on infrastructure like buildings or bombs and almost nothing could stop it except possibly an EMP.

I'm not sure what anyone can do about that but that to me is my biggest fear about the future of all this technology.

View on HN

1-10 million drones. At $400 a pop that’s $400,000,000-$4,000,000,000. A lot to throw at a single attack when you don’t actually know what defenses are in place. Maybe there is an EMP. Are you willing to spend a billion dollars to find out, while also murdering hundreds of thousands of civilians?

And these are either autonomous drones (more expensive?), or fpv with the fiber optic line out the back - either way you have to get them in range without being detected somehow.

In short, i think this is an unrealistic scenario - fun to imagine as a horror-sci-fi idea but unlikely to be deployed. Just one opinion.

> At $400 a pop that’s $400,000,000-$4,000,000,000. A lot to throw at a single attack when you don’t actually know what defenses are in place.

Have you seen the price tag on some of the US jets? Are they not doing just this?

Jets aren’t single-use. A better comparison would be something like a tomahawk missile, which costs ~$2mil each (not counting r&d costs, launcher costs, getting them there costs, etc).

The US spent $11.3b in the first six days of israel’s war with iran. So not an unprecedented amount of money, just a lot to put into a single attack that could fail, and that mostly kills humans, and that requires a shit ton of logistics to make happen.

> Jets aren’t single-use.

They are sometimes.

This isn't a game of Command and Conquer with fog of war turned on. Of course they would have intel on exactly what they are going to attack. One cruise missile is about $4 million.

China already has created a UAV that is designed to launch at least 100 drones. If they can make that 1000 drones and then fly out 1000 of these motherships at one time, that's already 1 million.

And yes the drones would be autonomous, there's no reason for any person to be controlling them in the age of AI.

So then my $400/each price tag is wrong - it’ll be much higher. If nothing else, the high price of gpus and ram might be saving us from an attack like this lol.

But also, 1000 carrier drones is a lot easier to shoot down than 1mil drones.

An attack like this might not be today, but in 10-20 years? If a country can develop this technology where multiple cities or military bases are completely overwhelmed by a swarm of drones, and all the politicians and generals are taken out by AI-controlled drones, why wouldn't they invest in it? As far as I can tell, unless some sort of localized EMP is developed, I don't know how an attack like this can be stopped. Maybe some sort of RF beam but I'm sure the devices can be shielded from that.

It's much more effective than a cruise missile because you can just blow up weak points on bridges, buildings, take out entire military bases, etc. Even 1000 of these drones would be extremely effective but 1 million would be devastating.

> all the politicians and generals are taken out by AI-controlled drones, why wouldn't they invest in it?

Because that’s an abhorrent thing to do. Because then you have to occupy people who don’t want to be occupied. Because the nation you just destroyed has allies who have the same weapons. Because that nation has the same weapons. Because killing “all the politicians and all the generals” will become impossible immediately after the first time this is used - assuming it’s not already impossible today.

Right now one of the limits is just controlling that many let along sourcing it. Putting that many actively controlled drones in one area at once and you'll swamp the bandwidth.

The drones in Ukraine largely use miles-long spools of fiber optic cable, so while I agree that sourcing is a problem, bandwidth likely isn't. If you want to be creative there's probably a bunch of hybrid wireless/wired/semi-autonomous configurations that would allow for minimizing bandwidth requirements in practice, but it would still fall apart as soon as a reasonably powerful jammer is turned on.

Feels like a massed attack using fiber optics would have lots of problems with drones getting tangled in the cables of those in front of/above them or cutting the cable. The trouble with making them autonomous in a city is that means more sensors and compute on board which raises the cost. It's a hard problem.

They would just be autonomous. Setting a GPS (or alternate system) coordinate is pretty simple enough. Individual targets could just be AI controlled at this point, or 10 years in the future.

GPSing your way through a city to targets is a big ask.

Aren't drones quite short-ranged, especially when carrying a payload?

>"1-10 million drones at once"

get real please.

Reply on news.ycombinator.com