4 ms·

Waste of money. Within 5 years the most effective battle aviation will be unmanned and partially AI powered (for when EW takes out the signal). It is plainly ob

by ein0p 2y ago

Waste of money. Within 5 years the most effective battle aviation will be unmanned and partially AI powered (for when EW takes out the signal). It is plainly obvious that from here on out our investment should go mostly into unmanned systems

JumpCrisscross 2y ago
> Within 5 years the most effective battle aviation will be unmanned and partially AI powered You’re parroting old tropes, which confuse the roles of CAS and SEAD.
- CamperBob2 2y ago
  Whatever those are, they don't need priceless humans in an expensive airplane. Not anymore. Explain why, don't just downvote. At this point, the burden of proof is on you to explain why the human pilots are still needed... considering that if you didn't have to keep a human pilot alive, you could afford to build 10-100 times as many of whatever aircraft we're talking about.
  - JumpCrisscross 2y ago
    > they don't need priceless humans in an expensive airplane. Not anymore What are you basing this on? > you could afford to build 10-100 times as many of whatever aircraft we're talking about What are you basing this on? Planes are capability limited by human life support. The most expensive bits inside them are still technology. Yields on some of these novel materials processes are below 1%. And that’s with expensive feedstocks like titanium. ChatGPT can’t direct a squadron through enemy radiation and successfully conclude SEAD. Not yet. (And we have been doing that for decades. They’re called anti-radiation missiles because they’re disposable.)
    - CamperBob2 2y ago
      What are you basing this on? The same thing you're basing your assertions on: what I see as common sense. ChatGPT isn't designed to win dogfights, but other models are, and gee, guess what, that's exactly what's happening: https://news.sky.com/story/first-known-test-dogfight-between-ai-and-human-pilot-carried-out-us-military-says-13118545 https://news.sky.com/story/first-known-test-dogfight-between... It is no longer enough to assert blindly that "ML models can't do X." You must explain exactly why ML models can't do X, and admit that you were wrong the last 10 times you asserted that ML models can't do X. (Radiation? Really?)
      - JumpCrisscross 2y ago
        > same thing you're basing your assertions on: what I see as common sense Well, that and my background in aerospace engineering and training to be a general-aviation pilot. Not quite expertise. But more than common sense. (I have some military-pilot [helicopter] and designer friends and family, but for obvious reasons, they don’t inform my public views.) > ChatGPT isn't designed to win dogfights, but other models are, and gee, guess what, that's exactly what's happening Terrific, can it also aim a trebuchet, because they’re equally germane to a modern battlespace. > that you were wrong the last 10 times you asserted that ML models can't do X What? When? > Radiation? Really? I didn’t name this crap [1]. I’ve followed your account for a while. Reasoning on modern military bleeding-edge technology using the fallback of common sense seems beneath you. NGAD is in design and pre-production; you’ll find their trading off what decisions are made by man versus machine interesting [2]. [1] https://en.m.wikipedia.org/wiki/Anti-radiation_missile https://en.m.wikipedia.org/wiki/Anti-radiation_missile [2] https://en.m.wikipedia.org/wiki/Next_Generation_Air_Dominance https://en.m.wikipedia.org/wiki/Next_Generation_Air_Dominanc...
        CamperBob2 2y ago
        Reasoning on modern military bleeding-edge technology using the fallback of common sense seems beneath you. Well, some days we all wake up in a trolling^Wprovocative frame of mind, and F35 meta-dogfights on message boards are always a hoot, but I'm only being half-facetious here. Video games had "AI" that did most of this stuff in the 1980s and the rest of it by the 2000s. So why, given modern sensor packages and modern software engineering, isn't it "just common sense" that machines are better at it than humans now, or will be soon enough? E.g., we had a crude version of SLAM in the 1950s, with cruise missiles that combined real-time terrain correlation mapping with inertial guidance. (You have to google TERCOM to learn more about that, rather than SLAM, because SLAM meant something even more ludicrous-sounding at the time.) Now we use vastly more sophisticated techniques in household vacuum cleaners. What exactly does close-air support require that can't be handled by an AI (if not an RPV)? As for ARM and SEAD, I honestly don't understand where you're going with that. Electronics can take a lot more nuclear radiation than you and I can, and as for EM, well, so? We know how to deal with that. On the offensive side, the Wild Weasel that fires an ARM is just the first stage of the missile, isn't it, at the end of the day? Can you elaborate on (again, exactly) why this role is inherently impossible to automate?
- ein0p 2y ago
  Take a look at what’s going on in the only high intensity near peer battlefield in recent decades. It’s 95% drones. $500 drone with an RPG duct taped to it immobilizes an Abrams MBT and the other two finish it off. Then Mavics drop grenades on the crew. For reach there are larger drones, cruise missiles, ballistics and hypersonics. “CAS” is bullshit in 2024. At best, planes are launching missiles or guided bombs from tens of miles away.
  - JumpCrisscross 2y ago
    > 95% drones. $500 drone with an RPG duct taped to it immobilizes an Abrams MBT and the other two finish it off You’re describing conflicts where nobody is meaningfully contesting air superiority. (And CAS!) > ”CAS” is bullshit in 2024 Stand-off CAS is still CAS. Air support for ground and naval units is far from obsolete.
Latty 2y ago
I'd suggest this video as a good overview of why pure unmanned isn't the route countries are taking, and is unlikely to be the best option: https://www.youtube.com/watch?v=1-0L5Wv86fQ https://www.youtube.com/watch?v=1-0L5Wv86fQ (and this video on manned systems that preceded it: https://www.youtube.com/watch?v=RPrWm6fWuaM https://www.youtube.com/watch?v=RPrWm6fWuaM as well) People expect manned-unmanned teaming to be the more likely thing, rather than just unmanned systems.

by ein0p 2y ago

JumpCrisscross 2y ago

> Within 5 years the most effective battle aviation will be unmanned and partially AI powered You’re parroting old tropes, which confuse the roles of CAS and SEAD.

CamperBob2 2y ago
Whatever those are, they don't need priceless humans in an expensive airplane. Not anymore. Explain why, don't just downvote. At this point, the burden of proof is on you to explain why the human pilots are still needed... considering that if you didn't have to keep a human pilot alive, you could afford to build 10-100 times as many of whatever aircraft we're talking about.
- JumpCrisscross 2y ago
  > they don't need priceless humans in an expensive airplane. Not anymore What are you basing this on? > you could afford to build 10-100 times as many of whatever aircraft we're talking about What are you basing this on? Planes are capability limited by human life support. The most expensive bits inside them are still technology. Yields on some of these novel materials processes are below 1%. And that’s with expensive feedstocks like titanium. ChatGPT can’t direct a squadron through enemy radiation and successfully conclude SEAD. Not yet. (And we have been doing that for decades. They’re called anti-radiation missiles because they’re disposable.)
  - CamperBob2 2y ago
    What are you basing this on? The same thing you're basing your assertions on: what I see as common sense. ChatGPT isn't designed to win dogfights, but other models are, and gee, guess what, that's exactly what's happening: https://news.sky.com/story/first-known-test-dogfight-between-ai-and-human-pilot-carried-out-us-military-says-13118545 https://news.sky.com/story/first-known-test-dogfight-between... It is no longer enough to assert blindly that "ML models can't do X." You must explain exactly why ML models can't do X, and admit that you were wrong the last 10 times you asserted that ML models can't do X. (Radiation? Really?)
    - JumpCrisscross 2y ago
      > same thing you're basing your assertions on: what I see as common sense Well, that and my background in aerospace engineering and training to be a general-aviation pilot. Not quite expertise. But more than common sense. (I have some military-pilot [helicopter] and designer friends and family, but for obvious reasons, they don’t inform my public views.) > ChatGPT isn't designed to win dogfights, but other models are, and gee, guess what, that's exactly what's happening Terrific, can it also aim a trebuchet, because they’re equally germane to a modern battlespace. > that you were wrong the last 10 times you asserted that ML models can't do X What? When? > Radiation? Really? I didn’t name this crap [1]. I’ve followed your account for a while. Reasoning on modern military bleeding-edge technology using the fallback of common sense seems beneath you. NGAD is in design and pre-production; you’ll find their trading off what decisions are made by man versus machine interesting [2]. [1] https://en.m.wikipedia.org/wiki/Anti-radiation_missile https://en.m.wikipedia.org/wiki/Anti-radiation_missile [2] https://en.m.wikipedia.org/wiki/Next_Generation_Air_Dominance https://en.m.wikipedia.org/wiki/Next_Generation_Air_Dominanc...
      - CamperBob2 2y ago
        Reasoning on modern military bleeding-edge technology using the fallback of common sense seems beneath you. Well, some days we all wake up in a trolling^Wprovocative frame of mind, and F35 meta-dogfights on message boards are always a hoot, but I'm only being half-facetious here. Video games had "AI" that did most of this stuff in the 1980s and the rest of it by the 2000s. So why, given modern sensor packages and modern software engineering, isn't it "just common sense" that machines are better at it than humans now, or will be soon enough? E.g., we had a crude version of SLAM in the 1950s, with cruise missiles that combined real-time terrain correlation mapping with inertial guidance. (You have to google TERCOM to learn more about that, rather than SLAM, because SLAM meant something even more ludicrous-sounding at the time.) Now we use vastly more sophisticated techniques in household vacuum cleaners. What exactly does close-air support require that can't be handled by an AI (if not an RPV)? As for ARM and SEAD, I honestly don't understand where you're going with that. Electronics can take a lot more nuclear radiation than you and I can, and as for EM, well, so? We know how to deal with that. On the offensive side, the Wild Weasel that fires an ARM is just the first stage of the missile, isn't it, at the end of the day? Can you elaborate on (again, exactly) why this role is inherently impossible to automate?
ein0p 2y ago
Take a look at what’s going on in the only high intensity near peer battlefield in recent decades. It’s 95% drones. $500 drone with an RPG duct taped to it immobilizes an Abrams MBT and the other two finish it off. Then Mavics drop grenades on the crew. For reach there are larger drones, cruise missiles, ballistics and hypersonics. “CAS” is bullshit in 2024. At best, planes are launching missiles or guided bombs from tens of miles away.
- JumpCrisscross 2y ago
  > 95% drones. $500 drone with an RPG duct taped to it immobilizes an Abrams MBT and the other two finish it off You’re describing conflicts where nobody is meaningfully contesting air superiority. (And CAS!) > ”CAS” is bullshit in 2024 Stand-off CAS is still CAS. Air support for ground and naval units is far from obsolete.

Latty 2y ago

I'd suggest this video as a good overview of why pure unmanned isn't the route countries are taking, and is unlikely to be the best option: https://www.youtube.com/watch?v=1-0L5Wv86fQ https://www.youtube.com/watch?v=1-0L5Wv86fQ (and this video on manned systems that preceded it: https://www.youtube.com/watch?v=RPrWm6fWuaM https://www.youtube.com/watch?v=RPrWm6fWuaM as well) People expect manned-unmanned teaming to be the more likely thing, rather than just unmanned systems.