[some_random]

There's an interesting similarity between autonomous weapon systems and ransomware discussions. Both have just come to the attention of the mainstream audience (or audiences outside their usual niche) and discussion in those spaces is about where it was 5 years ago.

Also I'd really expect HN of all places to understand why banning what is basically computer vision based targeting systems is absurd.

[ethbr0]

Technologists tend to underappreciated relative utility as a driving popularity metric -- does it accomplish a job I need, better / cheaper / faster?

My thought is that ransomware is seeing a resurgence because corporations are increasingly internetworked and hybridized with the global internet, because they have to be. So ransomware has begun to outweigh espionage as a method for extracting money from corporations.

Drones (as used so far, in the remote suicide version) solve a key problem many states have: lack of effective tactical ISR. The US and major powers do not have this problem.

Autonomous drones solve... what problem?

My inclination is that they'll deliver the same value as current-gen ML systems: mass, automated intelligence collection & basic analysis. As well as continuous monitoring and triggering on a pre-configured event (e.g. truck leaves this building, launch missile).

Killing something, once you know what and where it is, is not a problem most militaries have.

Training and paying and allocating large numbers of soldiers to do basic intelligence trawling is absolutely a problem most militaries have.

I think you see this in the US drone approach evolution. Shifting from a single vehicle approach to a survivable system with attritable assets communicating back through stealth communication hubs. Because persistence is the real value.

[whatshisface]

>Killing something, once you know what and where it is, is not a problem most militaries have.

It would be if there was a real war with a lot of things at once. That statement is only true for peacetime "executive action" wars with a 100:1 force ratio. It is an auxiliary role that has little to do with how things would go if there was a war between somewhat equal powers.

The lessons of enforcing foreign policy on ten people who live in a country with neutral diplomatic status will probably not apply if you change the number ten or the neutral diplomatic status.

[ethbr0]

Debatable. Armament stocks are ridiculously large, relative to value targets.

Which makes sense, because cruise missiles cost less than radar systems.

And I question whether "build one more drone" is an economically superior answer vs "build one more smart bomb."

[Arrath]

There are never enough bombs to go around, especially smarter versions. Classic every problem is a nail and I have a hammer situation for commanders and strike planners once they have access to such munitions.

This bit NATO forces in Libya[1], for example.

Building drones may well be more scalable and economically viable than smart bombs. Consider a multiuse drone that could be utilized for battlefield recon with its camera, or the user snaps something akin to a BLU-108[2] smart munition onto the bottom of the chassis and now they have a remote controlled smart anti-tank weapon. The drone being multi use could increase the use cases, increasing the order size, decreasing the unit cost.

[1]https://www.washingtonpost.com/world/nato-runs-short-on-some...

[2]https://en.wikipedia.org/wiki/BLU-108

[ethbr0]

From [1], it seems like the European stocks are a compatibility issue (can't mount US munitions) and a funding issue (don't purchase and store enough for a sustained campaign).

> Building drones may well be more scalable and economically viable than smart bombs

That's the heart of the matter. Is it, or isn't it? And for what? The commentary usually glosses over this.

E.g. what's the difference between a JSOW-ER deploying CBU-103s / BLU-108s or a "drone"?

I think the more important future distinctions are going to be between attritable/simple vs expensive/capable platforms, temporary vs persistent platforms, and organic autonomy vs network-reliant. Where some combination of the qualities is optimal for any particular mission.

As an example: Is a mid-air refuelable X-61 without organic sensors but carrying smart munitions a cruise missile or a drone?

https://en.m.wikipedia.org/wiki/Dynetics_X-61_Gremlins

[Arrath]

> E.g. what's the difference between a JSOW-ER deploying CBU-103s / BLU-108s or a "drone"?

For those on the receiving end, I can't imagine its anything more than an academic distinction.

For the forces using such items, it could come down to doctrine, or funding fights. It could be that lawmakers approving budgets will find it more palatable to approve a big purchase of 'drones' instead of 'smart cluster munitions'.

I expect we'll see a profusion of new devices marketed as either, or both, depending on the use case and the audience. It may be that the definition will never really be nailed down.

> I think the more important future distinctions are going to be between attritable/simple vs expensive/capable platforms, temporary vs persistent platforms, and organic autonomy vs network-reliant. Where some combination of the qualities is optimal for any particular mission.

I absolutely agree.

[whatshisface]

Weapons caches are valuable targets, so unless all the world's major powers are keeping all of their missiles in a number of locations much smaller than the number of missiles their opponents have, I don't see how armament stocks can be large compared to the number of targets. That's even assuming that the only targets will be military targets, which wasn't even true in WWII.

[ardit33]

Um... you are way off the mark here. You just missed the latest war of Azerbaijan vs Armenia, where the azeri forces had these drones, that just hovered in the battle field for hours, and when the Armeni forces turned their radars off their would just home on them and destroy them immediately.

They were acting like moving air-landmines. They key part of this is that they are super cheap to make (cost of a car), and operate, and work autonomously. Thousands of them launched, can wreck havock even on a sophisticated military like the US of Russia.

That's why there is a return of modernized cheaper ww2 type of anti aircraft guns.

China’s New Air Defense System Looks Similar To American ‘Stryker’ SHORAD https://eurasiantimes.com/chinas-new-air-defense-system-look...

[ethbr0]

>> Drones (as used so far, in the remote suicide version) solve a key problem many states have: lack of effective tactical ISR. The US and major powers do not have this problem

That describes Azerbaijan and Armenia's war precisely.

> work autonomously

The TB2 [0] is not capable of autonomous engagement. It's remotely operated for weapons release. And half its key components rely on foreign technology (regardless of what Turkey claims).

The Israeli Harop platform [1] does have autonomous engagement, but intended for the S/DEAD role and is in a different class of technical sophistication and price.

Calling Russian export military equipment "sophisticated" is a bit of a stretch, considering modern delivered models are circa-1995 designs, only recently realized.

Maybe 30 years ago, but they haven't been able to even afford to modernize their own forces, and basically lost 20 years of progress to post-Soviet upheaval and kleptocracy.

We'll see what happens when drones go up against a military keeping pace with the times.

[0] https://baykardefence.com/uav-15.html

[1] https://www.iai.co.il/p/harop

[x86_64Ubuntu]

>...Also I'd really expect HN of all places to understand why banning what is basically computer vision based targeting systems is absurd.

On the contrary, I would expect a site with a usersbase that screams incessantly about "privacy" and has an intense hatred of data-mining dark patterns, dislikes opaque ML models, would be against the lack of privacy and data-mining culminating in an execution by the state.

[nonameiguess]

Whether they kill you with a remote-controlled drone, an autonomous drone, some manned platform, or just a man is orthogonal to whether the decision to kill you is based on privacy invasion and inscrutable threat detection algorithms.

[x86_64Ubuntu]

It's not orthogonal when the entire killchain is automated.

[tclancy]

I would hope (but definitely not expect) HN commenters to see past this and realize the problem isn't computer vision or firepower. It's tying those two together, writing some code and cutting it loose on humanity. Put it this way: if you wrote the code for an autonomous machine gun platform, would you park it in your house? I'm twenty years into this and I wouldn't trust my code that well.

[stfp]

Just like with chemical weapons, banning them makes total sense to avoid overwhelming mass production and ubiquitous use.

You'll still see them used in targeted assassinations and some countries will have illegal military stockpiles, but that's very different from these things being treated like regular guns/ammo, being commercially available or legal to build.

[hn8788]

Chemical weapons are banned because they kill indiscriminately, and it causes undue suffering compared to conventional weapons, not because you can mass produce them. Drones wouldn't have either of those issues.

[throwaway316943]

Flying land mines can be equally indiscriminate. Especially if you set the targeting to an ethnic group. These are explosives so they won’t always kill their target, they’ll do a lot of maiming and anyone standing near the target will be on the receiving end too.