[palata]

The problem I have with the idea of subsidising small drones as a proxy for defense is that they solve very different problems: Making a small quadcopter that flies is now entirely solved: you take an open source autopilot, put it on some open source autopilot board, and that's it.

If you go further than that, successfully producing delivery drones means that they need to carry a payload safely to some destination, deliver the payload nicely (as in, smoothly leave a parcel on the ground), come back and be reusable. The drone flies by GPS, but doesn't really need a radio signal (ideally there is no operator, the drone just goes, delivers and comes back).

Killer drones are "one-way". They are defined by a lifetime of like 25min, ending up violently in a place where the operators care about maximising damage. They fly in war zones. Nobody really cares if some percentage of the drones falls from the sky or doesn't explode upon contact. They need to fly in GPS-denied mode, and they probably need a radio for the operator to select the target when the times comes. This has to be a military-grade radio that works in the presence of jamming to some extent.

Those are very different projects. Feels a bit like saying that subsidising personal cars is good for the tank business.

[toss1]

>>Feels a bit like saying that subsidising personal cars is good for the tank business.

Funny you should say that. The US had in 1938 a grand total of about 38 tanks. WWII started a few years later, and after converting prewar automobile factories to tank factories, the USA built more tanks than every other nation combined.

Pretty much the same thing happened for airplanes, as mentioned in the article.

US industrial production was literally the arsenal of democracy.

It is a LOT easier to convert commercial manufacturing base to military purposes than to start from scratch. So, yes, subsidizing commercial production to stay in-country is definitely good for mil readiness (and ultimately, the tank business).

[CrimsonCape]

I have the impression from reading history books that the workforce at the time of World War 2 was uniquely specialized and widely available. There were many machinists that had special knowledge and experience of how to run their lathes, presses, etc. This workforce was involved in the assembly line of passenger cars, so you had expert machinists involved in producing passenger cars which made expertise widely available. Because of their knowledge, they could easily pivot to an armored vehicle (for example).

In today's world the assembly line itself is derived from CAD, robot CNC machines, and the workforce is not specialized. The workforce consists of "assemblers" and machine operators, moreso than "machinists" or "machine designers"

This difference between workforces is a potentially profound difference.

[toss1]

Good points, although a nit that I'd characterize the workforce as more specialized today rather than less. Didn't the old-school machinists have more knowledge over the full range of production processes, vs a CAD drafter vs a Fanuc CNC operator, vs an assembler?

That said, I'd still say having one capability today still makes a far shorter path to convert from Civ-to-Mil output. I run a carbon-fiber composites shop that does everything from design through materials, CAD, CAM, moldmaking, forming with multiple technologies, CNC machining, and assembly. It would be a straightforward task to setup for new Mil products (and not just because we already do some Mil work), especially compared to not having it at all.

[nradov]

Military vehicles had much more in common with their civilian counterparts in the WWII era. Technologies have almost entirely diverged since then. An M4 Sherman tank had a gasoline piston engine and steel armor. An M1 Abrams tank has a turbine engine, and uranium and ceramic composite armor. To convert a factory from one to the other you'll have to rip out almost everything and start over.

[toss1]

Yes, mil tech has diverged, and much of modern manufacturing requires highly specialized tooling that requires long lead times to get into production.

That is an excellent reason to subsidize maintaining convertible or dual-use tech in the civilian arena. e.g., make sure turbines are used in more civilian uses. Stockpile tech that is really civilian incompatible such as the depleted uranium armor.

Turbines are a good example of how a civilian tech could have gone differently. In the 1960s several turbine-powered cars were in development for street use and a turbine race car qualified and lead most of the 1967 Indianapolis 500 race. But then the USAC effectively disqualified it [0], and civilian development stopped for other reasons. But it arguably might have continued had turbine power been allowed to race and dominate.

Yet, turbines are used both for aircraft and for natural gas power, both stationary and portable, and there are many small turbines. So, of course, we would not go to an ICE engine builder but to the builders of aircraft and gas power plants. There are also manufacturers of small-scale turbines that might ramp up.

On the other hand, we can also look at how modern warfare has changed over the last three years. multi-million dollar tanks are being reliably destroyed by $800 drones. And drone tech is highly fungible. Many common computer chips and boards can be used to control it, many common lightweight motors will work, and composites or lightweight metals can make the bodies. All of these technologies are highly configurable, so it would be a short lead time to make new factories to turn out pretty much whatever shape drone we wanted, whether it is flying, rolling, or swimming.

[0] https://en.wikipedia.org/wiki/STP-Paxton_Turbocar

[nradov]

There was never any possibility of turbine engines becoming widely used for civilian street autos. They are less fuel efficient and have slower throttle response compared to reciprocating internal combustion engines. This is inherent in the technology, not something that could have been fixed with further development.

[cutemonster]

Bomb and kamikaze drones based on civilian drones are already a reality though, Ukraine uses to defend itself. Don't know why you're talking as if that wasn't possible, when it's happening already.

[palata]

Hmm maybe I'm not being very clear, I didn't want to write a 20 pages essay :-). I was saying that I don't think it's a particularly efficient way to approach defense.

My point was that Ukraine doesn't buy 2 millions civilian drones and use them as killer drones. Ukraine is actually producing killer drones.

If you are good at producing civilian drones, it doesn't mean that you are good at producing killer drones because the specs are pretty different. If you subsidise heavily a civilian company making survey drones, for instance, and then try to attach a bomb to those and send them in a war zone, they won't do much today. In the end you will have subsidised work that went into making a drone that can make hundreds or thousands of flights during its lifetime, never fall from the sky, lands smoothly, doesn't make too much noise, follows drone regulations in civilian spaces, etc. But none of that work is useful for a killer drone (that has a lifetime of 25min in a war zone). On the other hand, your civilian drones will not have the ability to lock a target and crash into it, fly in GPS-denied environments and a jamming-resistant radio.

[hkpack]

Ukraine is absolutely buying all the civilian drones it can get, especially the larger ones with good optics.

One of the previous defense minister was skeptical of their utility too and called them “wedding drones”, and now you can see very frequently in war footages mentions how they are using “wedding drones” in this or that reconnaissance or surveillance operation.

You absolutely need tens of thousands of drones in the air all the time to support modern warfare.

And drones are being hunted by other drones too, so they don’t last very long.

“Millitary grade” digital communication and encryption is not that important as the scale itself.

[umbra07]

Yes, but that is simply because they don't have a better choice.

Given the choice between a $200 DJI and a $100 homebuilt "killer" drone, you would probably want 2x of the killer drones. However, if your bottleneck is your manufacturing capabilities instead of your money, then you might be forced to use the DJI drones instead of the custom killers.

[hkpack]

It is more complicated than that as the roles are very different and you need both. You just cannot substitute one for the other.

DJIs with their high zoom ratios and quality stabilised cameras just allow for wide area monitoring which killer drones relies on.

Video from surveillance drones are usually streamed to a teams of analytics far away from front lines for analysis of situation change. People analysing the video data is a significant chunk of the total personnel in this war.

Without having that, killer drones are not effective, since they are very short-lived, have very poor cameras and power characteristics. It is very difficult to find enemy with self-made drones.

So I argue that you can in fact have a civilian drone manufacturing which can be repurposed quickly into a cheap mass produced war-time surveillance drone with minimal effort.

The same goes for software - both sides use civilian service for video streaming and communication which works better than anything "military grade".

[palata]

Hmm, I feel like you get back to "Ukraine needs everything it can get". Sure, but that's not my point.

For your wide area monitoring, you don't want your radio to be jammed because it makes it useless. So if you think about building your equipment, you'd rather build jamming-resistant radios, right?

> I argue that you can in fact have a civilian drone manufacturing which can be repurposed quickly

It could potentially be repurposed relatively quickly if it was well designed. But what tells you it will be? Most software is not very well designed, and in the western drone industry it's particularly right, in my experience. If you subsidise a company to make military drones and they write bad software, you will still end up with a military drone ("the software is bad but it lasts 25min most of the time"). If you subsidise a company to make survey drones in the hope that their design will be good enough to be quickly ported to military needs...

> both sides use civilian service for video streaming and communication which works better than anything "military grade".

I highly doubt that. Civilian radios are easily jammed.

[dani__german]

tell me more about the video streaming and communication aspects of this. Just skype (haha), zoom, and text messaging/phone calls?

[cutemonster]

Thanks for explaining!

I think the article says that the factories are important too, and can be altered to produce these different drones much faster than if starting from zero.

And having one's own already verified and certified backdoor free electronics, rather than buying from what might turn out to be the adversary

[potato3732842]

They're not "based on civilian drones" other than using some basic software and electronics and design principals. Everything else is built around cheap and short lifetime.

[mytailorisrich]

For offensive drones in hot military situations I think one key area of study is swarming. To inflict real damage and evade defenses you need to throw a swarm of drones at the objective. If you that to be robust against jamming you may also want a level of autonomy.

[klooney]

> Feels a bit like saying that subsidising personal cars is good for the tank business.

I mean, in WWII, a lot of car manufacturers made tanks instead. Buick made the Hellcat, Chrysler, Ford and a variety of train manufacturers made the Sherman, and on and on. The skills are much more transferable than a lot of other fields.

In fact, this is explicitly why the US and others subsidize their passenger car industries.

This probably wouldn't work as well today, because most modern automakers just do engine design, assembly, and pick some parts out of a Bosch catalog, but I bet the more ambitious, vertically integrated automakers like BYD or Tesla could do an OK job in a pinch.