Hacker News Reader

Comment by specialk | Hacker News Reader

Does the US not have regulations about UAVs automatically firing of weaponry? Don't all current UAVs have human operators. The so-called "man in the loop". I'm having trouble finding evidence one way or the other but does the man in the loop have to pull the trigger or can he override a pre-programmed fire order. Allowing robots to automatically fire ordnance seems something congress would legislate against -- or at least I hope so.

I'm a little concerned with a swarm of drones controlled by one human. Humans have limited attention span -- they just can't watch 30 video feeds. Could this pre-programmed approach with limited human control lead to greater civilian causalities?

View on HN

Disclaimer: I am not a UAV operator. I am basing the following on information from mostly documentaries, but various other sources as well.

I believe that is correct, current UAVs (Unmanned Aerial Vehicles) used by the US military always have a human operator that is controlling the vehicle. If I understand correctly they do not control the flight of the vehicle in real time, but rather they give commands along the lines of "circle above this area" or "track/follow this particular target on the ground". The vehicles is semi-autonomous in that it figures out what it needs to do to fulfill the operator's command (ie. adjust yaw, pitch, roll, thrust, etc). I believe to actually engage a target the operator must manually trigger it.

However, this is not the same as the new swarming drones this article covers. Most do not consider UAVs currently used by the military to actually be drones (which implies a high degree of autonomy), instead they are typically thought of as remotely controlled. In contrast, it sounds like these drones are fired once with a target specified ahead of time and these drones are completely autonomous from that point on, swarming, swarming around and firing on a target all on their own. These drones are probably small and they probably dont have a video feed for each one for people to monitor.

Disclaimer: I was a Shadow 200 TUAV pilot in the US Army from 2001-2005, and yes, I spent a year in Iraq (OIF II circa 2003-2004) where I got 480+ combat flight hours.

You basically wrote the exact same comment I was about to write with striking similarity. For reference, this is a 18 year old me in the back of a Shadow GCS at Ft Huachuca, AZ, where all Shadow pilots are trained: http://en.wikipedia.org/wiki/Ground_control_station

The military never uses the word "drone", but prefers UAV aka Unmanned Aerial Vehicle, just as you mentioned. For what it is worth, this new navy thing isn't really that novel. The latest generation of Tomahawk cruise missiles do something very similar. You launch several of them into a battle theater and they circle around sending / receiving target information and telemetry between each other and the control stations. See: https://medium.com/war-is-boring/u-s-marines-can-now-call-in...

So in this case, what makes these drones any different than a cruise missile? You give them coordinates to destroy a target, the weapon then deploys to the target and uses technology to navigate without a pilot directly "steering" it or deciding when it goes "boom".

For what it's worth, the US has already deployed weapons that pick their targets autonomously. [1] These artillery-launched UAVs just seem like a fancier version of the same idea. Whoever gives the go-ahead to fire it will just have to weigh the importance of the target against the extent of the collateral damage.

[1] http://en.wikipedia.org/wiki/CBU-97_Sensor_Fuzed_Weapon

It's been happening since the 70s with cruise missiles, which are a basic drone that just flies into a target and explodes.

Just because there must be a human in the target designation loop, does not imply that there must be a human in the target elimination loop. After the targets are designated an entirely autonomous swarm may indeed be the perfect way to eliminate them.

Comes with its own risks, of course, but efficiency dictates that this is going to happen regardless. Best that the safeguards remain in place where they don't compromise that efficiency rather than pretending that we can just agree as a species not to pursue that efficiency gain.

    Does the US not have regulations about UAVs automatically firing of weaponry?

Even if we do have such regulations, I doubt they'll last long. We'll have to respond in kind once China, Russia, Iran, etc start developing fully automated weapons systems. When Russia can send 1,000 aircraft into a theater because they only require 10% the personnel and we can only send 100 aircraft because we require human operators, the pressure is going to be pretty intense be fully automated. And, even if automation doesn't increase capacity, automation will reduce reaction times, thereby providing other benefits.

    Allowing robots to automatically fire ordnance seems something congress would
    legislate against -- or at least I hope so.

Victory in [symmetric] warfare is already largely dependent on economic advantage, but future war will be entirely determined by economic strength: assuming similar levels of technology, my coalition wins, for example, if we can field 10,000 automated aircraft and your coalition can only field 8,000. I would expect Congress to legislate against this; when China develops fully automated systems and starts deploying them, I would expect Congress to gut those regulations or to let the DoD proceed quietly with developing a to-be-held-in-reserve software update for full automation.

I agree this is not a good outcome, but I'm not sure how the world can avoid it.

Do supervised drones make sense in a symmetric war? Missiles cost roughly an order of magnitude less than a drone (and can be launched in large numbers from the ground), so I wonder what the value of sending drones into a well defended airspace would be.

edit: added 'supervised', because a missile is a sort of drone.

I think this is more likely to be used in battlefield contexts where you have a very clearly defined military opponent. Navies don't do counter-terrorism, they are for conventional conflicts.

for the most part, yes, but there have been a few navies involved in suppressing pirate and possibly terror activity (piracy around the horn of africa, among other places.

That's a good point, I had forgotten about that. Thanks for the correction.

There will not be autonomous strikes. Too much potential for collateral damage. Also one person never watches 30 feeds. It's usually 1 or 2 analysts per feed.

I imagine there will only be specific use cases for this system.

>It's usually 1 or 2 analysts per feed.

there is no way they can put 1000 analysts on a small operation like kicking rebels out of a village :)

Another example - Russian anti-ship missiles attack in a "pack" where they automatically identify and assign targets between them.

The point is one way or another the "human in the loop" is just temporarily and very limited configuration, and it will be gone once we start to see real mass application of drones, especially when by multiple actors, as like in CIWS systems humans just can be fast enough and wouldn't be able to manage an encounter between for example hundreds of drones.

What happens when the target is not human?

Reply on news.ycombinator.com