Hacker News Reader

Comment by 01100011 | Hacker News Reader

Still need to solve the power problem. When drones can fly for 8+ hours straight or charge themselves using the electrical infrastructure then I'll worry about "Stabby the Robot".

AI isn't really the limiting factor as far as I can tell. We have good enough pattern recognition and flight software already.

View on HN

There are much higher density energy sources than lithium batteries which militaries can use, I’m not sure that is at all insurmountable. In other words, time to worry :) https://newatlas.com/nuclear-uav/22041/

Stabby could be sent to wait on the ground until the microphone detects footsteps.

In a warzone, AI is a big factor. GPS and control signal jamming have proven to be somewhat effective countermeasures.

But what happens when the drone is navigating with a camera and selecting its own target based upon training?

I'm not worried about battlefields in the "stabby the robot" scenario. I'm worried about unprepared civilians being exsanguinated en masse by a fleet of drones.

The scary scenario as far as I'm concerned, is mass genocide based on some characteristic. 10k drones wiping out 100k civilians who are identifiable based on location, skin color, clothing or some other measure. It's a scenario that has seemed vaguely plausible to me for several years now.

I refer to it as "stabby the robot" because drones wielding blades or other means of exsanguination do not run out of ammo and have little limitations on how much they can achieve their goals(other than powering themselves).

Exactly, and imagine them in the hands of terrorists. Some pretty awful outcomes are becoming possible.

Simple drone that goes boom on contact seems much easier to build that one that can repeatedly attack with a melee weapon.

Right but the melee weapon can potentially disable 10x as many targets as one that explodes.

Use another AI system to strobe a laser at the cameras of all the drones in the area and see if you can overwhelm their optics.

If you can accurately strobe it with a laser, why not use the same targeting system to shoot it down? But good systems like that with sufficient capacity won't be cheap. And they'd be prime targets for artillery during an offensive operation.

Then again, from watching the publicly available videos, it seems like these drones have been most useful for defensive operations on the battlefield.

> If you can accurately strobe it with a laser, why not use the same targeting system to shoot it down?

There will always be a chance that the targeting system misses so strobbing a laser (or bidirectional jamming) will minimize the risk of damaging the surrounding area (people, animals, etc.) vs shooting it down with bullets/nets.

Also a laser that emits enough light to blind a sensor may not be classified as a munition, whereas something that emits projectiles is almost always classified as a munition.

Or just install some cheap steel cable mesh awnings...

Dazzle camo and decoys seem like the obvious answer, and anything flying is still gonna have to contend with AA weapons platforms. It doesn't seem like it would be the hardest problem in the world to honeypot an autonomous drone into a kill box.

Those sound like captchas for drones to me. We've already started to see that bots can defeat some of those better than humans can.

Decoys are likely the best answer, but not necessarily practical for offensive operations. And given the low cost of drones, likely limited in usefulness anyway.

Reply on news.ycombinator.com