5 ms·

One of the major weaknesses of fighter aircraft is the pilot. Not their skills or abilities: the requirement for a soft, squishy human that deforms under G forc

One of the major weaknesses of fighter aircraft is the pilot. Not their skills or abilities: the requirement for a soft, squishy human that deforms under G forces to pilot the thing is a disadvantage. The person in the cockpit is the most breakable part of the system. Everything else on the jet is capable of withstanding far more material strain. Just being able to replace a human-driven fighter jet with one that is piloted by an AI - even a comparatively dumb one - would be an advantage for fighters. The AI driven jet would be more maneuverable straight out of the gate.

nostrademons 6y ago
It's also by far the limiting resource on aircraft production. It takes 18 years to grow a human to the point where they can pilot a jet, and then a few years more to train them. Many years more to train them really well. In any sort of war of attrition, you'll run out of trained pilots long before you run out of planes. The Japanese weren't defeated in the Pacific because they ran out of planes, they were defeated because they ran out of pilots. At the time of the Marianas Turkey Shoot, Japan still had a large carrier force (9) and a large number of aircraft (750), but their pilots were all green, and so weren't very effective in combat. The Battle of Santa Cruz Islands is considered a strategic victory for the U.S. even though it was a huge tactical defeat, because it depleted the stock of trained pilots enough that Japan wasn't able to mount effective resistance for the rest of the war.
- cushychicken 6y ago
  The Battle of Santa Cruz Islands is considered a strategic victory for the U.S. even though it was a huge tactical defeat, because it depleted the stock of trained pilots enough that Japan wasn't able to mount effective resistance for the rest of the war. Didn't know this. How sad. The more I learn about WWII and its abject brutality, the more I marvel at how much US culture fetishizes it. Related: this season of Revisionist History has a series about Curtis LeMay and the history of napalm that is equal parts fascinating and sickening.
  - gknoy 6y ago
    > equal parts fascinating and sickening That's how I felt when I watched the documentary about Robert MacNamara [0], when they started talking about how we are pretty sure that what we did to Japan _before_ the atomic bombs counted as war crimes. 0: https://en.wikipedia.org/wiki/The_Fog_of_War https://en.wikipedia.org/wiki/The_Fog_of_War
    - cushychicken 6y ago
      That's most of what the podcast covers. They quote MacNamara in The Fog of War as well.
    - RhysU 6y ago
      Please do recall everything mentioned above pertaining to Japan was after https://en.wikipedia.org/wiki/Attack_on_Pearl_Harbor https://en.wikipedia.org/wiki/Attack_on_Pearl_Harbor
      - saagarjha 6y ago
        When else would it be? I'm unsure what your point is…
        RhysU 6y ago
        All the horrible Pacific WW2 brutality by the US followed a surprise attack on its homeland intended to kneecap its military abilities so that Japan could invade countries around it unimpeded. Doesn't make any of it right. Just putting it in context.
      - cushychicken 6y ago
        The US did some disproportionately heinous shit after Pearl Harbor. For example: 2,500 people died in the attacks on Pearl Harbor. The firebombing of Tokyo killed somewhere between 90,000 and 120,000 people - most of them civilians, and most either incinerated or asphyxiated by incendiary bombs over the course of six hours. It's the single most lethal day in human history - even greater than the atomic bombs, where over half the deaths were lingering affects of radiation over weeks, months, or years. Curtis LeMay, the Allied Air commander in the Pacific, freely admitted that he would have been tried as a war criminal for the firebombing of Tokyo had the Japanese won. Don't rush to justify everything the US did just because we weren't the first to strike.
- kcb 6y ago
  There was another there major difference between the Battle of Santa Cruz Islands and the Marianas Turkey shoot. The US went from flying the vastly inferior Wildcat to the mostly superior Hellcat. Japan wasn't able to produce a more modern fighter than the Zero in significant numbers for the rest of the war and by 1944 it was obsolete.
- mattmanser 6y ago
  Funny, I thought airforces generally just recruited pre-grown humans called adults.
- jonfromsf 6y ago
  The Japanese had excellent pilots, but they did not invest in the infrastructure of making more pilots. Every excellent pilot flew until he was killed. The Americans pulled top pilots back to train the next generation of pilots, creating a massive talent gap over time. The lesson is one that applies to startups as well as to aviators: you have to invest in your people, not just in maximizing value delivered today.
fatbird 6y ago
Many years ago there was an AI game for PC called Creatures, using neural nets for the Norns, the characters you would raise in the game by training them. The devs who wrote it were originally writing AI software for fighter pilots, and continually won simulated engagements because the AIs had no body-based limitations, which opened up a range of maneuvers that would literally kill a human. The best example of such is a hard dive: if you're flying level, and dive too quickly, the G forces drive all the blood into your brain, potentially causing a hemhorrage. The maneuver human pilots use for a hard dive includes a half barrel roll just to avoid this, and it's still less effective than suddenly pointing down. The AIs used hard dives effectively to shed human pursuers. This led to the insight that bodies (in some sense) were a missing component of any human-like AI, so they made Creatures to test the theory. The neural nets had virtual bodies with hunger, fatique, and pleasure/pain receptors; the player interacted with a god-like hand cursor that could pet them or spank them. For the 90s, it worked pretty well on Windows 98.
- cushychicken 6y ago
  What a clever dual market strategy. Consumer PC games as failed fighter pilot software. Capitalism is incredible.
  - fatbird 6y ago
    The story I told comes straight from articles about the game as it was released, and the marketing definitely worked on me. Raising cuddly neural nets seems much more interesting when you believe that the AI was already proven in another area.
SkyBelow 6y ago
Humans also have a limit on what they will do. A military with more AI can engage in actions that a group of soldiers would refuse to do. As more of the military becomes AI driven, it effectively concentrates more power behind those who operate the AI.
- 082349872349872 6y ago
  https://www.youtube.com/watch?v=n7wluszxDrY https://www.youtube.com/watch?v=n7wluszxDrY On the bright side, horses are now spared much of the direct experience of modern warfare, so maybe humans will eventually follow. On the flip side, if the oligarchy-with-social-policies was a knock-on effect of mass conscription during the napoleonic wars, a military with more AI suggests an oligarchy without social policies. CN covid responders' celebration: https://www.youtube.com/watch?v=daVCbgNsVEM https://www.youtube.com/watch?v=daVCbgNsVEM UK covid responders' celebration: https://www.youtube.com/watch?v=4bQG9aB8dWY https://www.youtube.com/watch?v=4bQG9aB8dWY (to be fair, I think the latter was celebrating 72 years of NHS, so the choice of a folkloric airframe was likely deliberate.)
- cushychicken 6y ago
  Excuse me while I run to incorporate SkyNet Systems, LLC
- manfredo 6y ago
  Alternatively, humans will also do things that the military doesn't want them to do: like loot, rape, and kill non-combats. Not every war is WWII where military leaders actively directs troops to commit atrocities. More often than not, crimes are committed by troops against the wishes of their superiors.
- LatteLazy 6y ago
  That's a solved problem already really, basic training gives you soldiers who will do what they're told...
kleiba 6y ago
Wouldn't that be an argument for remote controlled fighter aircrafts?
- cushychicken 6y ago
  Latency between the jet and the pilot is an insurmountable problem. You either have a very laggy connection to a human pilot who is far away, or a wireless link to a more local pilot who is vulnerable due to being in a combat zone. Plus, it creates an EM signal that could be disrupted as a tactical weakness. The latency, though, would be the primary killer. It's why the Air Force still needs to send drone pilots to Afghanistan - you can have a person piloting the drone during the mission out of a container in Nevada, but they don't have enough reaction time to safely and reliably land the thing in Afghanistan. Control has to be handed off to a local pilot who has a lower-latency control link to the aircraft.
  - goatherders 6y ago
    THank you for sharing that. I hadn't thought of that in that way before.
mNovak 6y ago
Interesting to note however that in the actual competition [1], the AIs didn't seem to find benefit from vastly exceeding the performance limits of humans, with regard to G-forces. Maybe these limits come through in the design of the jet as a whole, and would be a different story in a built-to-purpose UAS. The winning AIs here mostly benefited from super-human precision in aiming. [1] https://www.youtube.com/watch?v=NzdhIA2S35w https://www.youtube.com/watch?v=NzdhIA2S35w
- cushychicken 6y ago
  Maybe these limits come through in the design of the jet as a whole, and would be a different story in a built-to-purpose UAS That's what I'd suspect. A jet that's purpose built to be flown by a robot would likely be much more acrobatic, as you're designing for the engineering limits of the materials, rather than a human. Regardless of the real answer, it's not making me feel like climbing into a jet cockpit and picking a fight with an AI.

nostrademons 6y ago

It's also by far the limiting resource on aircraft production. It takes 18 years to grow a human to the point where they can pilot a jet, and then a few years more to train them. Many years more to train them really well. In any sort of war of attrition, you'll run out of trained pilots long before you run out of planes. The Japanese weren't defeated in the Pacific because they ran out of planes, they were defeated because they ran out of pilots. At the time of the Marianas Turkey Shoot, Japan still had a large carrier force (9) and a large number of aircraft (750), but their pilots were all green, and so weren't very effective in combat. The Battle of Santa Cruz Islands is considered a strategic victory for the U.S. even though it was a huge tactical defeat, because it depleted the stock of trained pilots enough that Japan wasn't able to mount effective resistance for the rest of the war.

cushychicken 6y ago
The Battle of Santa Cruz Islands is considered a strategic victory for the U.S. even though it was a huge tactical defeat, because it depleted the stock of trained pilots enough that Japan wasn't able to mount effective resistance for the rest of the war. Didn't know this. How sad. The more I learn about WWII and its abject brutality, the more I marvel at how much US culture fetishizes it. Related: this season of Revisionist History has a series about Curtis LeMay and the history of napalm that is equal parts fascinating and sickening.
- gknoy 6y ago
  > equal parts fascinating and sickening That's how I felt when I watched the documentary about Robert MacNamara [0], when they started talking about how we are pretty sure that what we did to Japan _before_ the atomic bombs counted as war crimes. 0: https://en.wikipedia.org/wiki/The_Fog_of_War https://en.wikipedia.org/wiki/The_Fog_of_War
  - cushychicken 6y ago
    That's most of what the podcast covers. They quote MacNamara in The Fog of War as well.
  - RhysU 6y ago
    Please do recall everything mentioned above pertaining to Japan was after https://en.wikipedia.org/wiki/Attack_on_Pearl_Harbor https://en.wikipedia.org/wiki/Attack_on_Pearl_Harbor
    - saagarjha 6y ago
      When else would it be? I'm unsure what your point is…
      - RhysU 6y ago
        All the horrible Pacific WW2 brutality by the US followed a surprise attack on its homeland intended to kneecap its military abilities so that Japan could invade countries around it unimpeded. Doesn't make any of it right. Just putting it in context.
    - cushychicken 6y ago
      The US did some disproportionately heinous shit after Pearl Harbor. For example: 2,500 people died in the attacks on Pearl Harbor. The firebombing of Tokyo killed somewhere between 90,000 and 120,000 people - most of them civilians, and most either incinerated or asphyxiated by incendiary bombs over the course of six hours. It's the single most lethal day in human history - even greater than the atomic bombs, where over half the deaths were lingering affects of radiation over weeks, months, or years. Curtis LeMay, the Allied Air commander in the Pacific, freely admitted that he would have been tried as a war criminal for the firebombing of Tokyo had the Japanese won. Don't rush to justify everything the US did just because we weren't the first to strike.
kcb 6y ago
There was another there major difference between the Battle of Santa Cruz Islands and the Marianas Turkey shoot. The US went from flying the vastly inferior Wildcat to the mostly superior Hellcat. Japan wasn't able to produce a more modern fighter than the Zero in significant numbers for the rest of the war and by 1944 it was obsolete.
mattmanser 6y ago
Funny, I thought airforces generally just recruited pre-grown humans called adults.
jonfromsf 6y ago
The Japanese had excellent pilots, but they did not invest in the infrastructure of making more pilots. Every excellent pilot flew until he was killed. The Americans pulled top pilots back to train the next generation of pilots, creating a massive talent gap over time. The lesson is one that applies to startups as well as to aviators: you have to invest in your people, not just in maximizing value delivered today.

fatbird 6y ago

Many years ago there was an AI game for PC called Creatures, using neural nets for the Norns, the characters you would raise in the game by training them. The devs who wrote it were originally writing AI software for fighter pilots, and continually won simulated engagements because the AIs had no body-based limitations, which opened up a range of maneuvers that would literally kill a human. The best example of such is a hard dive: if you're flying level, and dive too quickly, the G forces drive all the blood into your brain, potentially causing a hemhorrage. The maneuver human pilots use for a hard dive includes a half barrel roll just to avoid this, and it's still less effective than suddenly pointing down. The AIs used hard dives effectively to shed human pursuers. This led to the insight that bodies (in some sense) were a missing component of any human-like AI, so they made Creatures to test the theory. The neural nets had virtual bodies with hunger, fatique, and pleasure/pain receptors; the player interacted with a god-like hand cursor that could pet them or spank them. For the 90s, it worked pretty well on Windows 98.

cushychicken 6y ago
What a clever dual market strategy. Consumer PC games as failed fighter pilot software. Capitalism is incredible.
- fatbird 6y ago
  The story I told comes straight from articles about the game as it was released, and the marketing definitely worked on me. Raising cuddly neural nets seems much more interesting when you believe that the AI was already proven in another area.

SkyBelow 6y ago

Humans also have a limit on what they will do. A military with more AI can engage in actions that a group of soldiers would refuse to do. As more of the military becomes AI driven, it effectively concentrates more power behind those who operate the AI.

082349872349872 6y ago
https://www.youtube.com/watch?v=n7wluszxDrY https://www.youtube.com/watch?v=n7wluszxDrY On the bright side, horses are now spared much of the direct experience of modern warfare, so maybe humans will eventually follow. On the flip side, if the oligarchy-with-social-policies was a knock-on effect of mass conscription during the napoleonic wars, a military with more AI suggests an oligarchy without social policies. CN covid responders' celebration: https://www.youtube.com/watch?v=daVCbgNsVEM https://www.youtube.com/watch?v=daVCbgNsVEM UK covid responders' celebration: https://www.youtube.com/watch?v=4bQG9aB8dWY https://www.youtube.com/watch?v=4bQG9aB8dWY (to be fair, I think the latter was celebrating 72 years of NHS, so the choice of a folkloric airframe was likely deliberate.)
cushychicken 6y ago
Excuse me while I run to incorporate SkyNet Systems, LLC
manfredo 6y ago
Alternatively, humans will also do things that the military doesn't want them to do: like loot, rape, and kill non-combats. Not every war is WWII where military leaders actively directs troops to commit atrocities. More often than not, crimes are committed by troops against the wishes of their superiors.
LatteLazy 6y ago
That's a solved problem already really, basic training gives you soldiers who will do what they're told...

kleiba 6y ago

Wouldn't that be an argument for remote controlled fighter aircrafts?

cushychicken 6y ago
Latency between the jet and the pilot is an insurmountable problem. You either have a very laggy connection to a human pilot who is far away, or a wireless link to a more local pilot who is vulnerable due to being in a combat zone. Plus, it creates an EM signal that could be disrupted as a tactical weakness. The latency, though, would be the primary killer. It's why the Air Force still needs to send drone pilots to Afghanistan - you can have a person piloting the drone during the mission out of a container in Nevada, but they don't have enough reaction time to safely and reliably land the thing in Afghanistan. Control has to be handed off to a local pilot who has a lower-latency control link to the aircraft.
- goatherders 6y ago
  THank you for sharing that. I hadn't thought of that in that way before.

mNovak 6y ago

Interesting to note however that in the actual competition [1], the AIs didn't seem to find benefit from vastly exceeding the performance limits of humans, with regard to G-forces. Maybe these limits come through in the design of the jet as a whole, and would be a different story in a built-to-purpose UAS. The winning AIs here mostly benefited from super-human precision in aiming. [1] https://www.youtube.com/watch?v=NzdhIA2S35w https://www.youtube.com/watch?v=NzdhIA2S35w

cushychicken 6y ago
Maybe these limits come through in the design of the jet as a whole, and would be a different story in a built-to-purpose UAS That's what I'd suspect. A jet that's purpose built to be flown by a robot would likely be much more acrobatic, as you're designing for the engineering limits of the materials, rather than a human. Regardless of the real answer, it's not making me feel like climbing into a jet cockpit and picking a fight with an AI.