Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
The Military AI

Ukraine Turning To AI To Prioritize 700 Years of Landmine Removal (newscientist.com) 221

MattSparkes shares a report from NewScientist: The Russian invasion of Ukraine has seen so many landmines deployed across the country that clearing them would take 700 years, say researchers. To make the task more manageable, Ukrainian scientists are turning to artificial intelligence to identify which regions are a priority for de-mining, though they expect some may simply have to be left as a permanent "scar" on the country. The model considers vast amounts of data, including tax and property ownership records, agricultural maps, data on soil fertility, logs from the military and emergency services of where bombs and shells have landed, information gleaned from satellite images and interviews with local civilians and the military. Even climate change models and data on population density derived from mobile phone operators could be assessed. The AI then weighs factors such as civilian safety and potential economic benefits to determine the importance of a given piece of land and how urgent it is to make it safe. Ihor Bezkaravainyi, a deputy minister at Ukraine's Ministry of Economy, is leading the team, and he likens the task of de-mining during an ongoing war to designing and building a submarine entirely underwater, except that the water is on fire. "It's a big problem," he says.
This discussion has been archived. No new comments can be posted.

Ukraine Turning To AI To Prioritize 700 Years of Landmine Removal

Comments Filter:
  • by alternative_right ( 4678499 ) on Tuesday June 18, 2024 @09:28PM (#64559703) Homepage Journal

    Europe is still de-mining and clearing unexploded munitions (UXBs) from WW1. Ukraine will have another century at least of finding dangerous explosives buried in the topsoil, no matter the outcome of this war. Russia will have another seventy years of taking care of war wounded and permanently disabled. Even the most conservative capitalist analysis will show us that war is a very expensive business, especially in terms of human cost.

    • by AmiMoJo ( 196126 )

      At least 20,000 Ukrainians have had amputations since the start of the war too. Gaza is probably even worse, but it's hard to get reliable data.

  • Russia, a backwards country with an incompetent and drunk army and no industrial base, somehow managed to bury 700 years' worth of landmines in a year. An amount and time frame not even seen in WWII and all wars since.
    Sure. "Researchers".

    • by piojo ( 995934 ) on Tuesday June 18, 2024 @10:26PM (#64559789)

      Russia, a backwards country with an incompetent and drunk army and no industrial base, somehow managed to bury 700 years' worth of landmines in a year.

      Are you misunderstanding this deliberately? I don't know exactly how they calculated that time span, but you can bet if someone buried a few land mines in your field, that hour's work would take you months or years to undo. No matter how you crunch the numbers, it's faster to plant land mines than to clear them.

      • How many of those mines were planted between 2014 and 2022, you know, during the Ukranian civil war?

        Also, how many of those landmines were supplied to Ukraine by the USA? e.g. https://www.nytimes.com/2023/0... [nytimes.com]
        • by dargaud ( 518470 )
          ll, Ukraine most likely has a map of their own minefields, so they can remove them after the war is over. I don't think that's what's being discussed here.
        • by Cyberax ( 705495 )

          How many of those mines were planted between 2014 and 2022, you know, during the Ukranian civil war?

          Not much. Both sides mostly used mortars, so most of UXO (unexploded ordnance) were mortar shells. They typically were not buried and unlike mines they are not designed to be tamper-proof.

    • So that save Ukraine some cleaning up.

    • by AmiMoJo ( 196126 )

      Russia has mines that can be deployed from aircraft, from missiles, and from mortars. They are released in the air and spread over a wide area. They can very quickly cover large areas with them.

      There is no easy way to remove them either. Even ropes with explosive charges at intervals are not 100% effective, and of course the explosions give away your position. Even in peace time there is no effective way to clean large area safely and reliably.

      Some Ukrainian solders are being given special shoes that absorb

      • Do you look at the ground for mines, or up for snipers?

        The guys in front look for mines, the guys in the middle look for snipers, and the guys in the rear look worried

    • How do you not understand that land mines are a form of extended asymmetric warfare, and take much longer to remove than they do to set? Were you in fact born last night? And why, given your apparently complete lack of information about land mines, did you feel qualified to comment about them?

    • I am not sure what your point is. Russia has inherited all the mines the USSR manufactured for WW3, and also the tools to mine areas quickly [wikipedia.org].

  • by RightwingNutjob ( 1302813 ) on Tuesday June 18, 2024 @09:44PM (#64559729)

    A: An already-exploded mine poses no danger to anyone, so one way of clearing a minefield is to explode them all. Mines explode when stepped on, so one way to ensure every mine explodes is to step on each one.

  • What "AI"? (Score:5, Insightful)

    by cstacy ( 534252 ) on Tuesday June 18, 2024 @09:58PM (#64559757)

    This would not be an application for an LLM, and I doubt they are asking ChatGPT where to demine. It doesn't even sound like any kind of NN problem.

    It sounds like a straightforward traditional "AI" program that is scoring parcels based on real-estate factors; this is the kind of thing you use rule-based (aka "production") declarative programming for.

    The hard part is getting the input data for the rules. Who knows what they are doing there. Some kind of language model (not a chatbot, though) might be useful for reading unstructured data (eg. random text documents). The result is the input facts for the rule-based system.

    A big problem these days is that people think that "AI" means an LLM such as ChatGPT. When actually, those LLM systems are not even what I would call "AI".

    The term "AI" has never been well-defined and is always a moving target. These days, it means "magic computer shit so give me money".

    • Re:What "AI"? (Score:5, Informative)

      by necro81 ( 917438 ) on Wednesday June 19, 2024 @08:21AM (#64560617) Journal

      It doesn't even sound like any kind of NN problem.

      IEEE Spectrum had coverage on this [ieee.org] back in April. They also have a podcast episode [ieee.org] that gets into it. It's not a LLM model - it's more about image recognition - looking for the signature of landmines from drone-gathered image data, which is most definitely a neural net application. Safe Pro AI has investigated different modalities for their 'image" data - visual, lidar, ground-penetrating radar, thermal/infrared, magnetometry. They're focused on visual imagery right now, largely because it is the easiest and cheapest to gather and process.

      • by cstacy ( 534252 )

        ^^^^ Mod Up Info +5 please

        IEEE Spectrum had coverage on this [ieee.org] back in April. They also have a podcast episode [ieee.org] that gets into it. It's not a LLM model - it's more about image recognition - looking for the signature of landmines from drone-gathered image data, which is most definitely a neural net application. Safe Pro AI has investigated different modalities for their 'image" data - visual, lidar, ground-penetrating radar, thermal/infrared, magnetometry. They're focused on visual imagery right now, largely because it is the easiest and cheapest to gather and process.

  • Let's say you have a walking robot that has about the same ground pressure as an adult human footstep, it doesn't need to be anything too advanced - could be just a big hexapod, and it's connected to a RTK GPS system. The robot is armored and is powered from an armored extension cord. Wouldn't it be faster, easier, and ultimately safer to have that robot step on every square inch of a minefield until it's stepped on all of it, thus ensuring by brute-force that there are no mines that can be triggered there?

    • Nice plan except robots will never kill their own. Only humans are stupid enough to frag their own species. The landmine won't activate.

    • it doesn't need to be anything too advanced - could be just a big hexapod, and it's connected to a RTK GPS system.

      Or you could just buy a deminer [wikipedia.org]. A souped-up tractor with a flail attachment sounds more practical than a bunch of semi-autonomous armoured robots.

      • They'd be fully-autonomous armored robots, and I think they might be more practical than a huge manned vehicle powered by liquid fuels...

        • The British used a number of different specially equipped for different tasks. One of them, the Matilda Scorpion [wikipedia.org] was first put into service in North Africa in 1942, and were even used for mine clearance on D Day. Only moderately successful, but good enough to keep around at least until the end of the war.
      • it doesn't need to be anything too advanced - could be just a big hexapod, and it's connected to a RTK GPS system.

        Or you could just buy a deminer [wikipedia.org]. A souped-up tractor with a flail attachment sounds more practical than a bunch of semi-autonomous armoured robots.

        The assumption being all mines are pressure sensitive and a flail will generate enough to detonate them, which may not be the case.

    • Let's say you have a walking robot that has about the same ground pressure as an adult human footstep, it doesn't need to be anything too advanced

      Stop there. If you're talking about a walking robot that's already advanced enough to be expensive. You want to walk all over all of the affected areas, setting off mines in the process, with expensive robots. Nope nope nope, that's never going to be sensible.

      It's a much better plan to use remote sensing of the mines, it's expensive but you can conceivably borrow the equipment since it's not going to be used up.

  • So for the first 500 years of landmine removal, nobody bothered to ask whether landmines had even been invented yet?

  • by Tailhook ( 98486 ) on Wednesday June 19, 2024 @01:11AM (#64559965)

    It's not going to take 700 years. Ukraine has found that thermal imaging can spot mines effectively. At certain times, like early evening when the ground cools, the heat retained in the dense mines can be seen by thermal cameras. The mines shine out as spots on the ground. They've already done this with drones, and it's a short walk to using ML to automate image analysis at a large scale. When the time comes "we" will be able to demine much more efficiently than the conventional methods this "700 year" figure is doubtlessly based on.

    • The cost of demining HAS to be built into calculations, as is VA costs, and loss of capital, and loss of land productivity. Thermal cameras are no good in dense urban areas, or where a building rubble abounds (Like Gaza). Mines should have a use-by date and deactivate - but no - zero agreement on that. In addition there are 'smart mines' designed not to explode unless it hears what it expects - or trigger an AP mine many meters away. Yet here we are in WW1 attrition BS. If you swim, or by boat, landmines c
      • Thermal cameras are no good in dense urban areas

        What? That's nonsense.

        Mines should have a use-by date and deactivate - but no - zero agreement on that.

        The whole point of mines is to shit on your opponent both now and in the future. It's a design feature that they kill children for decades. War isn't happy fun time, it's shit designed to enrich fuckfaces forever and ever amen.

    • by Cyberax ( 705495 )

      It's not going to take 700 years. Ukraine has found that thermal imaging can spot mines effectively

      Unfortunately, this only works with mines that are on the surface or close to the surface. Mines have a tendency to get buried, and then surface years (or decades) later. People still die from mines placed during the WWII.

  • IEEE did a podcast episode with the young developer who is spearheading this effort. Truly an inspiring story. https://spectrum.ieee.org/clea... [ieee.org]
  • Why bother with AI? Windows provides a free tool for it! [wikipedia.org]

We are Microsoft. Unix is irrelevant. Openness is futile. Prepare to be assimilated.

Working...