Quadcopters can now visually track targets more effectively | Hacker News Reader

Quadcopters can now visually track targets more effectively

mosfet.net

78 pointsPaulHoule105 comments

I recently read statements from an Azov brigade commander in Ukraine that Russian Orlan-10 observation UAVs are a major problem. Russia is producing a lot of these UAVs and has been using them more heavily and effectively. Orlan-10s reveal targets for Lancet UAVs and other weapons. This leads to Ukrainian crews and weapons being destroyed and forces SPGs and other artillery away from effective positions.

Orlan-10 is a vulnerable system. It's slow, low altitude, easily detected and fragile; basically just a militarized model airplane. A fast UAV designed to detect and kill Orlan-10 would be a huge asset. At the moment Ukraine is using precious and costly AA missiles for this.

I'm thinking this tracking improvement would be a major help for anti-armor UAVs as well. One of the problems UAV operators have when attacking armor is losing FPV signal as the UAV gets close to the ground. If the terminal phase could be handed off to an onboard tracker it would vastly improve these weapons.

> At the moment Ukraine is using precious and costly AA missiles for this.

Apparently gepards seem to have quite a success against shaheds: https://www.youtube.com/watch?v=_PWJ97KcnmA

In the video one of the soldiers mentions that usually 3 rounds are enough, according to this: https://www.defensenews.com/global/europe/2023/09/05/germany... the price per one round is 181000000 / 300000 = 603 USD. I am pretty sure that contract will involve other things to make the restart of production viable so the total cost per round will probably be lower in following contracts.. But 1800 to down the drone does seem super cheap compared to rockets...

https://www.youtube.com/watch?v=pb5_F4_Eod8 this is the next version of the same concept. But it seems that the laser mounted next to it will be even cheaper in the long run https://www.airforce-technology.com/contractors/air-defence/...

Yes, the Gepard platform has been a real success. Gepard has been most effective defending high value fixed assets from simple minded long range UAVs and cruise missiles.

When dealing with Orlan-10 however, the problem is different. Here you must defend mobile forces, close the the front from observation and attack UAVS that are remotely piloted. It's a whole different problem and there will never be enough Gepards and Gepard crews to sweep the skies of Orlan-10, and if there were they would become targets themselves.

Your problem statement (protect mobile forces close to the front) is exactly the original mission set of Gepard and similar platforms (ZSU 23 for example). There's a reason they are tracked AFVs.

Ukraines current situation is that Gepards are more valuable in static defence roles, and they don't have enough platforms to employ them everywhere they want.

And you don't need enough AA sweep the skies clean. You just need enough to add sufficient friction to typical operations, and enough to concentrate together when needed to ensure l that you can locally sweep the skies when needed.

I think the premise that it's viable to "concentrate together when needed" is false. A key tactic on Ukraine's front line, due to the nature of the enemy and the nature of Ukraine's weapons, is dispersion. Concentrating force for any reason before Russia's overwhelming artillery has always proven costly. When that premise fails the utility of point defense like Gepard fails. The only way to salvage it is to imagine Gepards in quantities that a.) don't exist and b.) can't be crewed for lack of trained manpower.

I seek a larger force multiplier: a system that can effectively cover large areas by a small number of men at low cost. One Gepard could pay for many light fox hunting UAVs. With sufficient autonomy a large number of these could be operated from secure areas by small crews. The system should be capable of reaching behind enemy lines to preempt UAVs before they reach combat areas; a capability the can't be matched by any point defense.

What you’re asking for isn’t possible since the development of non-gps & ai based navigation. Any system capable of launching physical strikes will be taken out by counterbattery fire almost immediately, and with the hardening against EW the drones have received you essentially can’t really do much without the equivalent of an EMP, which would likely cause you more damage than the enemy.

The problem is that the orlan only needs to survive long enough to give approximately coordinates for a lancet or artillery strike. Revealing the position of a Gepard in exchange for taking down a drone is a… poor trade, at best.

This puzzle is really starting to fit together nicely but SO FRUSTRATING that I still can’t tell whether it’s an image of Skynet or Big Brother.

Perhaps if I add one more piece…

It's that YouTube movie where micro drones are essentially like magic heat-seeking bullets that scares me the most. Targeted assassinations that are unstoppable with modern defenses - useful for both the terrorists and bloodthirsty nation-states which make us all live in a dystopia with fear of death from the sky

https://undergroundhousing.com/

Maybe you should rather worry how to help the people who are fighting for their lives.

We should worry about both. What we shouldn't do is let allies get slaughtered because Russia can make a lot of model airplanes, or bankrupt ourselves killing them with million dollar AA missiles. There must be a better way.

The better way was in not starting this war, or not undermining three attempts at peace negotiations before it ever got to this point.

The US has been involved in one proxy war after another with Russia for decades. It’s just standard US foreign policy, and it simply can’t be denied - US ISR is providing actual strike targeting information.

Regardless, it’s just more of the same. See: Vietnam, Ethiopia/Somalia, Afghanistan, and half a dozen others.

Proxy wars are just the way that wars are going to be fought between peer/near-peer powers, because a full blown confrontation risks global catastrophe.

Ukraine was just foolish enough to sign up for the role of “US’s cannon fodder” in the never ending US-Russia war.

When - not if - when, China ends up getting involved, we will likely see a shift from a unipolar world to a multipolar world and the end of US hegemony and European colonialism.

I’m sure I’ll get flagged for going against the Russia bad US good narrative though.

> Ukraine was just foolish enough to sign up for the role of “US’s cannon fodder” in the never ending US-Russia war.

I think you will find its Russia who invaded, and Russia alone is the one who decided this.

No one else did, everyone else wanted peace, Russia alone wants war.

So the constant shelling of crimea by Ukraine prior to the war never happened?

The best way to counter the smaller UAVs that operated with roughly a 6k-10k range is via EW. Jam their control links and they either land right where they are or RTB. Either way they're out of the picture. Unfortunately, that also tends to jam your own UAVs as well (at least the commercial ones converted for spotting and FPV attacks.)

Radar doesn't work that well since the RCS of these craft are so small and the radar systems have gate filters that exclude slow moving aircraft. If you tweak this lower, then you end up with a far shorter range with a lot more noise. And if you broadcast too much, you end up eating a KH-58. It's not a trivial problem to solve. Eventually things like better acoustic and vision sensor processing will help.

Despite the US wanting to ignore the impact of these drones, they really are a game changer. Cheap, distributed ISR via drones is invaluable. And imagine in a few years when the control systems use frequency hopping etc. I'm also envisioning a drone controlled via fiber optic cable. It won't be able to go far, but it can go high and act as a spotter, limited only by its cameras. It won't be jammable, nor limited by bandwidth. Kind of like how balloons were used in the Civil War and WW1 to adjust artillery fire. And if the invasion of Ukraine has shown anything at all, it's that if you get spotted, you die.

"The best way to counter the smaller UAVs that operated with roughly a 6k-10k range is via EW."

EW has been brought up a few times now in this thread. I know both side have been able to defeat UAVs with EW on occasion during this conflict. However, I also know that the Russians have lost a lot of costly EW gear to Ukraine: EW is inherently vulnerable because EW lights up the world with RF signatures that get targeted. Also, despite the large amount of EW gear that Russia has employed, they lose lots of tanks, IFVs and men to hobby grade drones every day. Russian EW hasn't been able to stop that.

So I question EW and whether it can really deliver against systems that rely on RF. Then, when you get to fully autonomous systems that don't rely on RF at all, I write EW off as next to pointless.

It's important to remember that all of the videos on YT only show successful attacks. Ukraine is reportedly burning through 10k drones a month.

Russia was terrible at almost all phases of the invasion during the first year, but has really buttoned down in two areas; AD and EW. Russia has been a leader in EW for a long time and is slowly responding.

EW can't be everywhere though, and the smaller drones can avoid it in some situations since it's line of sight.

Also, don't mistake EW as just jamming. Sometimes it's taking over the control channel for a drone, or something even more esoteric like microwaves that fry the systems.

The better way was not releasing Victor Bout, who secured Shahed drones with his Iranian connections

The US has developed both low-cost interceptors like you suggest and Directed Energy Weapons for a counter-UAS role.

https://www.army.mil/article/268403/army_joint_force_work_to...

DEWs are likely to win out in the long term (at least for defensive purposes) because the cost per interception is extremely low when compared even to the price for the 30mm autocannon ammo used by Gepard and similar SPAAGs, never mind when compared to the price of a guided drone or missile. DEWs are also capable of intercepting a broader range of targets, including mortar and artillery rounds. With increased power levels, DEWs should be able to defeat ballistic and hypersonic missiles as well. The US DoD has already developed DEWs more than capable of the 50kW power levels required to destroy a small-UAS like an Orlan and is in the process of integrating them with Strykers and other armored platforms so they can be deployed en masse, but these systems are still in development and thus have not yet been made available to the Ukrainians. e.g., for a system that is already production ready: https://www.rtx.com/raytheon/what-we-do/integrated-air-and-m...

But that's why you don't hear much about re-establishing SPAAG production lines or scaling up interceptor drone production. Everyone knows something cheaper and more capable is coming very soon. Interceptor drones may play a role in clearing out enemy-controlled airspace prior to an offensive action when there is a need to destroy drones that are out of line of sight (although I expect DEWs mounted on attack helicopters or large drones to fulfill that role longer term), but defensively nothing will be able to compete with DEWs on cost and capability in the very near future.

Speaking of Orlan-10, its manufacturer is ZALA Aero Group.

https://www.sensusq.com/blog/sensusq-analysis-on-the-zala-42... Son of the de facto owner/ceo of the company currently works at UN Institute for Disarmament Research (UNIDIR) in Geneva. UN is so useless its not even funny.

The best way to kill drone is with better drones. We are at the dawn of a new arms race.

That remains to be seen. Cannons, lower-cost missiles, or directed energy weapons might end up being more effective at killing drones. This is a rapidly evolving field and I don't think anyone can reliably predict the outcome.

The big problem is target detection. Small quadcopters may be able to track nearby targets, but they lack the sensors needed to detect targets at useful ranges. At a minimum defensive systems will still need large, expensive sensors mounted on ground platforms, aerostats, or larger fixed-wing drones.

I believe currently most 'kills' of small drones are because of EW ('electronic warfare' i.e. jamming and spoofing, both for the control link and for GPS navigation), which causes them to crash or land in enemy-controlled location.

> The big problem is target detection.

Generally yes, detection is hard. The Orlan-10 case, however, is easier. They can often be seen and/or heard from the ground and they are transmitting a live video feed, so their RF profile is a dead giveaway. If a fast, fox hunter counter UAV could be directed to the vicinity and then pick up the target efficiently a small crew could cover wide areas.

I imagine something like a fast, traditional airplane form with directional antennae in the wings for long range tracking, and visual tracking (as per the linked story) at short range. The actual weapon would be a small firearm: the targets are fragile.

No, the best way to kill a drone is with a cheaper better drone that you can make more of.

If it's a kamikaze thing.. The alternative is a world of e.g. dogfighting drones with a small weapon mounted on the hunter. Think of it as a lightweight airborne AA gun. There you might justify a higher price with e.g. the hunter having a >80% likelihood of surviving an encounter.

Maybe no need for a gun even: dangle a detachable rope or web underneath your drone, now you just need to be fast & maneuverable enough to catch the other drone, and tangle the rope into its propellers.

Just because it's along the lines of what I've been thinking- such drones could have longish lengths of 'fishing wire' trailing them, and perhaps some sort of expansive spiders web, so that their engagement with such drones in a counter drone environment might aid bringing them down by snagging the aggressor's propellor

Russian Lancet new generation already uses target locking. I suspect if it is possible to clone a piece of equipment, the only answer is an economy of scale because neither side will have technological advantage.

How about soldiers on the ground with shotguns? Or are they too high for that?

"Low altitude" in this case is 1-2km. Small arms can't deal with that.

Shotguns have been used to counter Lancet in terminal phase. But that's a desperation tactic. The goal would be that Lancet et al. is useless because the Russians are blind.

Wouldn't old school AA guns be perfect for that sort of thing?

With radar directed[0] gunnery and proximity fuzed rounds, absolutely. For SHORAD, or short range air defense. You'd need a lot of guns parceled out over a lot of frontage to protect your positions.

[0]Which, of course, means it has emissions that can be tracked, pinpointed, and targeted.

No. The front is a huge; tens of kilometers deep and hundreds of kilometers long. You would need a stupid number of AA guns and crews to cover it all. The idea is fast hunter UAV that could sweep large areas and respond quickly to enemy UAVs.

My understanding is that AA guns work best against targets the size of a fighter or bomber. As the target shrinks, so do your odds of a hit.

Here's an Orlan-10 on its launch mount. Perspective means it appears larger than it actually is.

https://en.wikipedia.org/wiki/Orlan-10#/media/File:SlavicBro...

So, no, not perfect. Probably not even within an order of magnitude of the same effectiveness. Possibly not even two orders.

AUTONOMOUS AA guns would be perfect for that sort of thing

The US military has had Phalanx CIWS for decades to handle this type of threat

what both Russia and Ukraine could use are cheaper lower caliber versions of this

And then the attackers will locate these expensive systems and attack with a cheap drone swarm of 10-15 drones at once. Doesn't matter if you lose 9, if one gets past you're likely in trouble.

Most of the AA systems coming on line can easily handle a swarm of 15 drones. Watch a video of Skyranger with proximity fuses wiping the gnats out of the sky. The problem is that each system can only handle a small area of the battle front.

The problem is that each 30mm projectile costs ~5k. Not a good trade off if you swarm it with $300 Chinese drones. And as soon as it starts firing, artillery will start raining down on it. Plus, the Russians have already demonstrated the ability to attack AA systems from multiple sides to get through the blind spots. Any attempt to cluster multiple AA guns or radar systems will make for an incredibly attractive target. Modern production lines can make what, a few hundred of them a year at best? Russia has purchased several thousand drones from Iran alone last year, and they have supposedly produced several hundred thousand lancets.

The US desperately needs to stop feeding the MIC and Wall Street and to reindustrialize the US economy, because we are simply not able to produce anywhere near as much missiles, shells, tanks, etc to fight the war the US wants to fight. The current combined yearly shell production capacity of Germany and the US is less than half of Russia’s alleged production capability, and they just signed a deal with NK that will likely increase their production capacity by a good 50% at least.

Where are you getting $5K for a 30mm round?

Russia still hasn't destroyed any Gepards yet, so I don't see why they'd be able to successfully attack Skynex. And Russia hasn't shown any ability to "swarm" anything. At most, they've used an Orlan to spot for a Lancet. And if they can't fly Orlans around, they won't spot the Skynex system for artillery fire.

Also, the Shaheed drones from Iran aren't cheap. They're easily several thousand euros. And there's no indication at all that Russia has produced 200K + Lancets. Russia simply doesn't have the industrial capacity for that, or artillery rounds.

And Germany doesn't need to be the only partner with the US in providing artillery rounds. All of NATO is doing so, and if you think that North Korea and Russia can out produce the US and the EU (not to mention Japan and South Korea), you've been reading too much Russian propaganda.

Exactly, so what is needed is a cheaper version of these weapons. A cheap drone swarm versus a cheap turret swarm built off the browning 50.cal rather than some complex 20 mm chain gun.

Too expensive in terms of cost, materials, and time. Also too vulnerable to swarm attacks. A gepard costs what, 400k? Lancets are 20-40k. I can almost guarantee that it won’t take anywhere near 10-20 lancets to kill one, maybe 3-5. And a non-SPAAG AA gun is a non-starter with modern ISR.

The only reasonable counter-tactic against cheap drones have been EW measures. They're by no means perfect but it's the best any of the two sides has at this moment (yes, it's better than the AA weaponry they have at their disposal).

Here's a video [1] of a Ukrainian drone hitting Russian EW equipment on a tower and here's a Russian video [2] presenting their EW works against incoming Ukrainian drones.

[1] https://old.reddit.com/r/UkraineRussiaReport/comments/17fbxo...

[2] https://old.reddit.com/r/UkraineRussiaReport/comments/17gnna...

Easy to defeat with automatic target recognition and tracking. You only need to get in the general vicinity and let the drone figure it out, which is what I suspect new Lancet is doing.

Yeah, even cheaper and lighter batteries will allow for even more processing to be carried out "locally", on the drone itself, with reasonable good automatic target recognition becoming viable.

Not for the larger surveillance drones, no. Those are a km up, you are unlikely to even know it’s there.

For small drones a shotgun is great… if it’s semi or full auto, if the soldier knows exactly where a drone coming right at him is, and has time to prepare while it’s still around 50 yards out.

Outside of those conditions, still not great.

Shotguns are roughly “effective” up to 100yards with slugs, and half of that with buckshot. Drones dropping grenades are generally well above 200 yards up, generally more like 500. So it’s really only the suicide ones that can be reasonable targets. The problem is a purpose build FPV drone typically used in suicide runs can haul ass, and cover 50yards in a second, are hard to see and predict the path of, even when you know one is there. Consumer drones can generally cover 50 yards in roughly 2-4 seconds.

Trying to shoot one moving across your view dive bombing on something else is going to have you more likely accidentally shooting the drones target or something unintended rather than the drone itself.

I do anticipate vehicle mounted scattergun systems meant for close in defense of drones though, as well as quickly/auto launched anti-drone drones for further off defense. More powerful drones would be needed for dealing with stuff 1km up and beyond though, the smaller stuff will struggle getting close to that.

What's going on with Anduril ? Isn't this literally their product market fit?

Anyone know of any good sources on the topic to stay up on the recent developments?

Watch the four million videos on YT that come from Telegram, and follow the OSINT crowd on X.

How many years until cops use drones and very fast robotic dogs to chase and apprehend people?

This is not meant as some kind of criticism, I believe that this is a fact which will become true regardless of us approving or disapproving it.

I mean from a technical standpoint, where the on-board AI and tracking abilities as well as the physical ones will be so capable that they will be able to replace humans chasing others, with the ability of evading thrown stones or other objects, possibly using tasers or fine but strong nets to prevent the chased person from continuing to flee.

A bit less than 10 years? Then how much longer until this is no longer a novelty?

Apprehend, not in the foreseeable future.

Chase, however, I'm surprised we haven't been seeing for at least a few years. Here in Chicago, following a high-profile bystander death resulting from a police pursuit, the department adopted a new pursuit policy that effectively bans vehicular pursuits in a majority of situations.

While many contend that this policy goes too far, it's indisputable that vehicular pursuits are dangerous. The potential for tracking, for example, a robbery suspect from the air (with a considerably lower bar than scrambling one of the city's two police helicopters) via drone seems like an obvious win for safety, both in reducing dangerous pursuits and apprehending dangerous offenders.

This would be great for car chases. Police cars in high speed chases too often lead to collateral damage and lots of people get away when police disengage as a matter of policy. A drone follower would be a great improvement.

Black Mirror: Metalhead.

It’s one of the best episodes.

https://en.m.wikipedia.org/wiki/Metalhead_(Black_Mirror)

Would not work very well in an urban environment where you could duck into the front of a business and go out the back way. Maybe drones + ability to remotely lock the doors of any building in the city. Or a drone swarm with built in maps of the layouts of all the buildings in the city so if someone goes into a building, they automatically split up and start watching all the exits.

I've seen videos of Ukraine where operators hit specific windows in high rises where they know Russian meetings are happening

Yeah but I’m talking about apprehending criminals around the city. Can’t just send missiles through a Starbucks.

Controversial Opinion: This is a good idea. There is zero need to make these Black Mirror lethal but tracking and apprehending is a scary job and Cops are human (there are also bad cops). Automating that process seems like Step 1 to getting out of our current strategy of

"Well cops are bad, so we can't enforce any laws".

Well-intentioned ACLU action and Public backlash have made these very hard to roll out, but long term, this may be the only the solution (along with wholesale long-term welfare overhaul of course)

> zero need

There's zero need for local rural cops to have MRAPs and SWAT teams, and yet...

You’re making my point. Humans are prone to this sort of largesse and they use human safety as an excuse to do so (“why don’t you want our cops safe?”).

Let’s remove that excuse.

Meta, but the way you felt you had to word your opinion really rubs me the wrong way.

Are we really at a point where we need to have a preface, two mid-sentence disclaimers, and carefully worded language just so one does not get torn apart by one's own in-group?

If that's where we're at, I'd sooner support the other camp, whatever it is.

> There is zero need to make these Black Mirror lethal

Nope, loot ak Ukraine. And military surplus can easily end up with the police.

I'm assuming of course we have some political control here. Just a walking robot was shamed out of existence [1]. So the public has some control over what is and isn't acceptable.

We don't need to accept lethal robots, but wholesale saying "no robotic equipment at all" will simply leave us in the logjam we have now between bad cops and zero enforcement.

[1]:https://www.nytimes.com/2021/04/28/nyregion/nypd-robot-dog-b...

Based on the behavior of many police departments in the US, I wouldn't put as much faith in your assumption of political control. From the "Blue Flu" to how many departments behaved both during the BLM protests and during COVID, police departments wield an amazing amount of control over their own TTP and who they seek to control.

Right but that behavior is driven (either earnestly or not) by safety concerns for the cops.

Lets remove that concern, no excuses left.

Safety concerns? Ha!

Cops would have you believe that LEO is the most dangerous job in America, but it's not even in the top 10.

Some cops may be afraid of getting shot, but consider that for every cop killed, cops kill roughly 21 citizens.

Copaganda has made everyone think that crime is terrible, cops are underpaid, overworked, and going into combat every time they report for work.

I literally said “earnestly or not”. It doesn’t really matter what the underlying truth is, it’s an excuse, let’s remove it.

No more SWAT teams needed if it’s just robots doing tracking and catching.

We've already had a swat team use a robot to kill someone using a small explosive charge. Maybe let's not head right to Cyberdyne before we address institutional issues in law enforcement before we get even more droids asking us for our papers etc.

> "Well cops are bad, so we can't enforce any laws"

When did that become our current strategy? I haven't seen any sign of it in my neck of the woods.

I live in the Bay Area. And this is basically the mentality spreading through most large cities.

Maybe that's true, but I honestly haven't seen evidence of it (at least in the big cities I'm familiar with where this claim is sometimes made). I've just seen people who claim that it's the case. From my observations, it appears to be a misunderstanding/miscommuncation that leads to this belief.

Well enforcement is down through the floor. You can measure that [1].

Cops claim it’s because of too much regulation (regulations put in place because of bad cops). It doesn’t really matter if they’re lying or not, there’s now no enforcement.

And all the political capital spent putting regulations in place now means there is none left to solve the underlying root causes of crime.

This is the dynamic playing out in most cities, esp in cities with strong police unions.

[1]: https://sfstandard.com/2023/07/01/ask-the-standard-san-franc...

But none of that supports the assertion that the attitude is "Well cops are bad, so we can't enforce any laws". It may support the assertion that "the way we're trying to address abusive policing is misguided", which is an entirely different thing.

“ the way we're trying to address abusive policing is misguided”

and as a result we have no enforcement now

That’s all I’m saying.

>How many years until cops use drones and very fast robotic dogs to chase and apprehend people?

I'd go farther than that. I'd wager that within 10 years, someone will have been shot by a mostly autonomous drone operated by a police unit somewhere. An operator will have had a approve the shot. We are probably 20-30 years from fully autonomous takedown of terrorists.

How many years until a domestic terrorism attack operated by a drone pilot followed by us never finding the person

Better chance of that happening by the end of 2024 than of Twitter being a dominant financial platform by the same time, to cross the streams of current HN threads.

That's a win for 'fusion of unrelated subjects' I think.

Do drones make it significantly easier to attack without being caught than, say, timed explosives or even firearms?

You don’t need to know where the target is going to be exactly in advance, and you don't have to personally be in firearm range or line of sight.

As robotic dogs that apprehend suspects play a major role in Fahrenheit 451, this is one of those things that should really make us pause and reflect.

Say no more! There's a YC funded startup on it already

https://paladindrones.io

Ethics issues...

Drones? Already, today.

Robotic dogs to apprehend people? I dunno.

Not only robotic dogs apprehending people, robotic dogs with posh English accents.

https://www.youtube.com/watch?v=djzOBZUFzTw

https://youtu.be/Bmy-Cr2bkZ0

they almost do that at the end.

For very slow motion stuff right now, yea. Not chasing people down though. At least not yet...

Reminds me of the Slaughterbots video [1] that the Autonomous Weapons project [2] released five years ago.

[1] https://youtu.be/HipTO_7mUOw?si=L1z_NYk77BHAbFel

[2] https://autonomousweapons.org

Is our humanity really so far gone? At this rate, researchers should come together and refuse to publish further research in this area.

I would hope UN can come together and at a minimum make an arms treaty to limit/prevent/undo development of AI tools that track/hurt individual humans this way. At a pathetic bare-bones minimum, these drones should never be allowed to kill or maim humans autonomously, and must be hardcoded to "accept a surrender" if someone stops holding their hands up.

If we limited chemical weapons, we can limit these terrifying abominations too.

The only reason chemical and biological weapons aren't used in warfare is because they just aren't very useful. An equivalent amount of high explosives is more lethal, easier to deliver, easier to transport and store, and less dangerous to your own troops. Which sucks, because it implies that arms control measures for landmines, cluster munitions, autonomous drones, and nuclear weapons are probably doomed to failure.

https://acoup.blog/2020/03/20/collections-why-dont-we-use-ch...

The cat is already out of the bag.

There's no meaningful way to limit access to any of this when it's purely software based and the hardware required to run it are commodities.

It's nothing like chemical weapons were the reagents can be controlled as they're hard to come by or synthesise in a amateur lab.

Isnt the same technology used on consumer drones https://www.youtube.com/watch?v=oM7cg4Gi_BU ? This is just a bit of an improvement on all that.

> I would hope UN can come together and at a minimum make an arms treaty to limit/prevent/undo development of AI

Well good luck getting US / China to sign something like that.

Paper seems really thin on details on how the targets are initialized. It mentions a "foundation detector" and claims its better than YOLO results, but then routes that conversation to the Ultralytics library (which has implementations of YOLO and SAM).

Am I missing something here? Is it SAM? That would make sense , SAM is amazing and running SAM (fastSAM?) onboard a quadrotor is a great achievement.

Weird, right? I couldn’t tell if they are trying to rebrand 2019’s yoloact model as a “foundation model” for marketing, something else with ultralyrics, or if there is hide-the-ball happening.

As far as non-military uses of this sort of thing: it’d be nice if I could deploy some drones that would spot and track off leash dogs BEFORE they’re running straight at me and my dog. Send one forward along our path and get warning that “hey, your irresponsible neighbor’s dog is loose around this corner”.

A dystopian future awaits... once Starlink is globally active, a drone can locate and track any human, anywhere on the face of the earth. Starlink communication will eventually be compact enough to be mounted on a quadcopter...

It already is. There a phones that support it and copters that can tow that weight.

Cheap, individual-seeking flying bombs are very close to being available to folks who can soldier and experiment a bit.

I rather suspect a civilian anti-drone defense market is starting to spin up now...

I think anti-facial recognition makeup will become a lot more prevalent.

completely ineffective, the more you research the more bits of information you realize are leaking off every single thing that makes you unique. your gait, smell, heart-beat (detectable at range), posture etc are all just as identifiable by a computer as your face.

It supposedly is true that putting on black and white face makeup as if you are a fan of Insane Clown Posse, however, it would tend to make you stand out in a crowd, if you were walking down the street...

I'll have a Klaus Schwab mask on the back of my head, like the tiger hunters of old.

Depends on if the target is a specific person or "any human trying to run away"

It's already done. In Ukraine. Remember the dispute with SpaceX

what could possibly go right

Reply on news.ycombinator.com