You don't need to worry about the drones themselves. The drones are useless without the explosive munitions. Good luck acquiring that without raising any flags. Not to mention somehow acquiring all the training data and ML modals to actually make them useful. Then there is the compute power that would be needed and coms infrastructure.
it puts it firmly out of reach for anyone that isn't state sponsored.
Not saying autonomous weapons won't/couldn't be an issue. just pointing out that drone swarms won't be.
>>explosive munitions. Good luck acquiring that without raising any flags
Not if you were a bit precocious in HC chemistry; I know my friends and I were not even close to alone in knowing all kinds of interesting fireworks recipes, all of which were well tested, and could be configured for anything from a small 'pop' to far more power than a slaughterbot. All from basic chemicals that can be easily be acquired without raising an eyebrow (could alternatively just repurpose a bullet/ shotgun shell, very easily obtained). The ability to put it right on target means that very little is needed.
>>acquiring all the training data and ML modals to actually make them useful.
Sure, if you are looking for fully ai-controlled and integrated with a city-wide CCTV system, that's a big task. But drone control and swarming is already out of the bag. Even a basic open-source Pixhawk does obstacle avoidance, and so can likely be reprogrammed for obstacle-targeting. Multiple off-the-shelf drones have done "follow-me" for years. I'd be astonished if a competent team with five-figure funding couldn't make a working system that could be deployed near a target, ID, it and home in.
We're way past needing state funding. The fact that it is not here suggests that while the scenario is scary, it is not actually that useful? Perhaps that's hoping too much.
I agree with all your points. I've been saying to my friends that I think one of the wealthy billionaires could possibly become a serious power. I think we can all agree Elon Musk could make this happen easily with his existing SpaceX people, and then build a bajillion of them with his Tesla people.
I wonder how far EM could go if he went full evil, full throttle and tried to take over something [the world? a country?] with a drone fleet?
Charging stations seems easily solved, so the drones never need direct human interaction.
I don't. He's the obvious choice for this thought scenario for several reasons:
1) He's the richest man on the planet
2) He has competence (arguably, but that's another debate)
3) He already has all the infrastructure, experience, and people to do close to this exact thing
4) He's already widely considered evil or at least hated
If you sub in Bezos, I'm guessing he's got a lead on EM in the software, but he'll be way behind in other areas.
If you sub in Buffet or Gates, then I have no idea how to even have this flippant conversation, because I have no idea how to rate their ability to get the job done.
I'm confident 3 of the 4 above could solve the problems of "make autopiloting, auto-targeting AI drones, and produce millions of them." I'm sure countless people could solve those problems and have the means to get it done. Have your choice of billionaires and motives, or nation-states and motives, or aliens and motives.
Maybe I should have chosen Santa Claus. I do really have a hard-on for him.
They do, but I don't think this is an example of that — but Musk's position w/Tesla and SpaceX makes him uniquely positioned to develop & deploy such tech.
That said, with all the advantages Musk would have if he "went full Evil", if he challenged something that got the US military involved, I'd bet on the US Military. In a straight-up fight, they're the most lethal force that ever existed. You might get away with 'sucker-punching' them a few times, but not for long.
> Good luck acquiring that without raising any flags.
How is the explosive used in the slaughterbots scenario different from what you can buy by the box in cartridges of bullets? Those aren't exactly highly protected even in the USA.
I imagine e.g. just tying a bullet to a drone, without the structure of a gun around, it isn't exactly going to be effective. But how far off would it be?
It’s a decently large difference. A couple oz of C4 is amazingly effective, no casing required - but an additional couple oz of ball bearings or a copper cone will make something truly scary. [https://youtu.be/AwyniA5ryhY?si=a562j-T8pyQlwaH4]
1/4 lb total for a devastating precision or area weapon.
The casing required for any nitrocellulose based device to do anything dangerous at all will exceed the weight of both of the above and be barely a firecracker. Fragments will be limited in velocity too.
Because if you look around at sophisticated military hardware (which these drones, when they will exist, will be) it's all like that.
I feel that the point our stances diverge is you feel these kind of weapons systems would become something a skilled hacker could make in a shed out of AliExpress parts for a 10$ bom, whereas I view it as being a complex weapon system that would be sophisticated and relies on battlefield infrastructure and complex supply chains to be feasible to properly deploy.
I think if you look at cutting edge weapons, my belief is no hacker could build them in their garage, no matter how sophisticated. Hell, take a look at the pathetic state of "3d printed" weapons to see how I imagine DIY "slaughterbots" would look like.
And acquiring the real weapons would not be easy at all. Heck, even now, if you want to acquire for your projects certain camera or heat sensors you have tons of hoops to jump through to get your hands on them.
Besides, even IF it would be something a DIY-er could do, someone wanting to kill civilians would probably use the most effective tools, least resource intensive, least effort tool, least conspicuous to gather resources and this would not be it. Today any maniac can use a car to plow down civilians or get a gun and go on a rampage. That is low effort high impact. This would be a lot higher effort. Probably kind of the same impact.
> Because if you look around at sophisticated military hardware (which these drones, when they will exist, will be) it's all like that.
I agree that it will be much like sophisticated military hardware is today - but I say that as someone within 20’ of his own third-generation night vision, body armor, and weapons that are approximately on par with that of a contemporary US soldier.
> Heck, even now, if you want to acquire for your projects certain camera or heat sensors you have tons of hoops to jump through to get your hands on them
Nonsense. You can order very usable thermal sensors from China, shipped to your door. You can order even more capable sensors that are covered by ITAR and prohibited for export without so much as signing something. A notice on the website saying that they’re restricted to the US is the extent of it.
> Today any maniac can use a car to plow down civilians or get a gun and go on a rampage. That is low effort high impact. This would be a lot higher effort. Probably kind of the same impact.
I totally agree with you here. Firearms aren’t the most effective way to cause harm to a large number of people even today; drones certainly won’t be for the foreseeable future.
FPV killer drones used in Ukraine now are really slapdash designs: often just motors on a flat fixture with a charge rubber-banded to it. They are used to devastating effect.
The only part of this technology that is remotely possible to tackle on is proliferation of explosives.
Fully agree. I mentioned in my original post I believe them to be very capable on a battlefield. But from there to slaughterbots like targeter drones there is a looong way. I believe my points still remain valid.
These drones won't be sophisticated. They'll be cheaper versions of off the shelf drones. Quantity > Quality.
Imagine a drone that flies towards a cellphone signal and when it gets within range does some basic shape recognition and adjusts it's course to fly towards a the nearest neck until it detonates a crude fragementation grenade.
How much would a drone that can do that cost today? They don't need to be durable as they are meant for single use, and they don't need to be particularly fast, just faster than a person can run.
Consider that despite grenades being dropped from drones onto Russian soldiers in Ukraine on a daily basis, it happens to American civilians never despite America's extremely high homicide rate. This is due to a lack of grenades, not a lack of drones.
The main thing gating slaughterbots from reality is the proliferation of the shaped explosive, not of the delivery mechanism.
Americans kill each other with guns because guns are cheap and plentiful, not because grenades are scarce. If guns were as scarce as frag grenades then it would be a competition between which one is more effective at killing and the availability wouldn't factor into the weapon choice at all.
What you're telling me is that slaughterbots will be armed with guns that shoot bullets, not fragementation grenades.
Well, it's a bit more than that. People kill each other with whatever they have close at hand when they get into a fight; handguns are easily carried into a bar or on the street, which is why more people are killed with handguns than long rifles in America despite more of the latter existing. People aren't going to start carrying drones with propellers around in their pockets.
It could be used by some people who want to do targeted killings, e.g. in gang violence, where it could be better than other methods maybe. But on the other hand, RC helicopters able to carry and shoot a gun have existed since at least 2006: https://www.youtube.com/watch?v=bZCH1492CzA
The number of people who actually want to kill lots of people without being physically present is extremely small; so small that in a country the size of the US, one only crops up every couple decades, e.g. Timothy McVeigh.
I would make a solid bet that proliferation of cheap drones does not impact the homicide or terrorism rate in the US in either direction.
Grenades are low precision weapons that are relatively safe to use. Guns are high precision but very difficult to use in a firefight.
If you watch Ukraine footage, soldiers barely even look where they’re shooting. They’re just trying to establish suppression over an area so that they can toss more grenades to kill people.
That’s because both sides have guns. If you aren’t fighting an armed combatant to the death. The calculus changes a lot.
(Video shows airsoft players using a standard FPV drone with an actuator to press a trigger, and the guy flies around the field getting "kills" and it looks incredibly effective (partly due to no recoil, but .22LR exists))
The ordinance is controlled, and without that the system loses a lot of potency. Not to mention that without decent onboard AI, jamming is a pretty good countermeasure since target selection and much of the control input isn't actually done onboard the vehicle.
Those seem like pretty large hurdles imo.
Edit: and, as others mentioned, goods even tangentially related to defense are routinely controlled. In the extreme case, countries like the US would onshore manufacturing of certain components, acquisition would require licensing, and competing products from overseas would simply be banned.
What is so difficult about creating a small explosive that could take a chunk out of someones neck so they bleed out?
> In the extreme case, countries like the US would onshore manufacturing of certain components, acquisition would require licensing, and competing products from overseas would simply be banned.
Which parts exactly? the PCBs? The motors? the microcontrollers? the batteries? How would that work exactly? How do you regulate any of this without per component DRM? What would that look like exactly? Would I need a license to buy a child's toy that had a microcontroller in it? Would they nerf microcontrollers for devices like appliances and toys so taht they weren't fast enough to handle the needs of a drone?
Materials for making decent explosives are usually controlled. If you're talking homemade explosives, you probably aren't getting that on a quadrotor, or if you are, you're seriously compromising its performance. Bullets are controlled through licensing in most first world countries.
If you wanted to control components going into homebuilt drones, you would probably target the guidance, since even basic waypoint navigation requires something to tell you where you are in relation to the target. To an extent this is already done with GPS (the true precision was classified last I checked, civilian uses are downgraded), but controlling manufacture and distribution of the stack would shut any kind of home robotics down while leaving a lot of channels untouched. If people wanted PCBs, fine, they just wouldn't be able to get certain sensors on them without a license.
You don't need global DRM to control exports. It's routine for some industries.
Of course, much of this is besides the point, because if someone really wants to hurt other people, they'll find a way. But the components to build a cruise missile in your backyard not only exist, they're mostly available COTS, but the critical parts like the payload are controlled. In some hypothetical future where AI drones became a threat, you can bet that the critical paths would be identified fairly quickly and locked down.
How exactly?
Are you expecting some sort of DRM in all commercial EDA and CAD software? Will opensource CAD and EDA software become illegal?
Will customs search all packages from Chinese PCB manufacturers for illicit drone designs?
Of course a nuclear weapon can kill more people than a drone swarm, but that's besides the point.
What exactly does non-proliferation look like for drone swarms? What happens when the cost of a drone swarm drops to $10?