[mrtksn]

Elon Musk recently made some remarks about F-35 on Twitter: https://x.com/elonmusk/status/1860574377013838033

It's exactly what I was thinking about: Having a pilot onboard is overhead and its a limiting factor for the flight envelope as you have to keep it alive. Besides, it's not invisible despite calling it that. It's visible in the visible spectrum and hearable in the human hearing range, which means that you can build detection and tracking systems in those wavelengths instead of pretending that its invisible because its hard to detect at radar wavelengths.

Why just not drop any manned vehicles and go for the remote control + AI? What is the logic? Sunken cost fallacy? Military industrial complex needs it?

The only thing I can think of is the political implications of downing plane with a soldier on board.

[oefrha]

> it's not invisible despite calling it that. It's visible in the visible spectrum and hearable in the human hearing range, which means that you can build detection and tracking systems in those wavelengths instead of pretending that its invisible because its hard to detect at radar wavelengths.

That’s not a gotcha you seem to believe it is, it’s like the first thing you learn about stealth technology and has been apparent to anyone following military news to any extent in the past forty plus years. That you somehow think it’s something that went unnoticed and the thousands of people working on this every day has been “pretending” after reading some Twitter posts is absurd and funny.

[mrtksn]

Wow really? Someone else already noticed that invisible to radar isn’t invisible to eyesight and event teaching it? lots of smart people out there.

I’m happy that you’re feeling smart and superior but that’s not the point, the point is signal collection and processing came a long way, and maybe low radar cross-section isn’t that big of a deal anymore. At least enough to justify spending huge sums on a plane that keeps having issue and falling short.

[compass_copium]

>maybe low radar cross-section isn’t that big of a deal anymore

That's simply not true. There is not a peer or near-peer country that is not developing a variety of low-RCS aircraft.

[fmajid]

It’s not as if stealth is foolproof even to radar:

https://en.m.wikipedia.org/wiki/1999_F-117A_shootdown

[Reubachi]

This incident happened precisely because the human controller activated radar twice to get a read on ground AA targets.

This is an operational/user fatigue error rather than a fault of the plane. It was completely blind to any and all AA until it turned on it's radar.

Which is to say, f35 Stealth characteristics are worth their weight until exactly the moment a human turns off "lane keep assist" because "they can do it better."

[oefrha]

Your target is long gone by the time you can detect short wavelength visible spectrum. Jets don’t need to get physical with you to do damage, it’s not WW2 anymore. That’s not to say countermeasures aren’t a focal point of research.

There are lots of “smart people” out there shitting on very serious work after reading some tweets from captain obvious. There are lots of even smarter people who don’t think everyone else’s an idiot and therefore think long and hard before shitting on work they don’t understand.

[compass_copium]

>Besides, it's not invisible despite calling it that. It's visible in the visible spectrum and hearable in the human hearing range, which means that you can build detection and tracking systems in those wavelengths instead of pretending that its invisible because its hard to detect at radar wavelengths.

There have absolutely been programs to reduce the visual signatures of aircrafts. Yehudi lights, COMPASS GHOST, BoP, etc. Don't get your information on anything from Musk.

[mrtksn]

Good point but all these things can be fitted on a drone too.

[compass_copium]

It's not like UCAVs aren't being developed, though. Piloted aircraft have lower latency and are better for some tasks. If UAVs were ready to replace piloted aircraft in all combat roles, they would have. They have almost completely replaced piloted aircraft in ISR, for example, where they are a great fit (no Gary Powers, no latency issues). The Pentagon doesn't want to lose trained pilots.

[mrtksn]

Does it justify the expense and the shortcomings though? Horses also have some advantages over tanks, like refueling from the grass on the roadside but they become mostly ceremonial anyway.

[compass_copium]

Yes, because AI is not good enough to fly a plane, detect an enemy craft, and react before a human pilot could. Like I said, they are developing UCAVs. They are not replacing all human-piloted strike planes, let alone all human-piloted air-superiority fighters, yet. Being able to pull a few more gs in a turn doesn't give you that much of an edge, yet.

[JohnBooty]

It depends.

In a lot of cases unmanned craft are obviously superior.

But by the time you load an unmanned craft up with a powerful radar and multiple bombs and missiles - you're also not that far away from the size and expense of a manned craft.

The radar is something you might not appreciate. Modern fighters have very large and powerful onboard radars and a lot of onboard processing power. They are not just bomb and missile trolleys, they are powerful sensor platforms.

Think about it from another perspective. Scroll down and look at the cutaway drawing of an F-16:

https://www.flightglobal.com/defence/why-lockheed-martins-f-...

...clearly, a signficant portion of the plane is dedicated to the pilot. But honestly, not that much. Strip away the cockpit and how much smaller is the plane? 20-30% smaller and cheaper?

[reisse]

> Why just not drop any manned vehicles and go for the remote control + AI? What is the logic?

When the F-35 program was started in late 1980s-early 1990s, neither reliable remote control nor AI existed (I'm not even sure supersonic-reliable remote control exists now). Now, if there exists research programs utilizing unmanned fighter jets, they're likely classified and we won't know about them for quite some time.

[euler_angles]

DARPA's Air Combat Evolution program (ACE) began with AIs fighting each other in a simulated environment in a tournament. Then the winning AI fought against a human (USAF Fighter Weapons School graduate) in that simulated environment, and won. The company that developed the winning AI, Shield AI, has gone on to deploy an AI in an actual F-16 that has flown against a human in trials.

https://www.darpa.mil/program/air-combat-evolution

[zcw100]

Many people underestimate the challenges of working in this environment. The DARPA challenge is like saying that boxing is the same as hand to hand combat. There are many similarities but the first has rules and parameters and the second has none. I find that people in IT tend to arrogantly proclaim things along the lines of, "I've set up a Kubernetes cluster. This can't be any more difficult." but it's more like setting up a Kubernetes cluster where you're paying someone else just like you to do everything they can to destroy it including giving them a knife and stabbing you. Then letting them loose in the datacenter.

[euler_angles]

I've worked with quite a few people who were part of the AI effort, and my current boss was the architect for ACE a few years ago. All of those people were painfully aware of the gap you describe here and were actively working to bridge it.

[zcw100]

I'm sure you and your team were very aware. That was a very cool project and I'd be interested in hearing about your experience. My comment was more directed towards the armchair tech generals out there.

[euler_angles]

I didn't work on this effort, I am just in the same circles with a lot of people who did and still do.

[542354234235]

First, visible/invisible isn’t a dichotomic distinction, it is a spectrum. Stealth isn’t to make an aircraft “invisible” to radar, it is to reduce its radar signature, which means that radar can’t see an aircraft at all beyond a certain distance, and can make it more difficult to identify and track at closer locations. What that means is that a stealth aircraft can fly “in between” radar stations without being detected, where other aircraft couldn’t, and get closer to an adversaries location before being detected, and be more difficult/take more time to identify it as hostile and engage. That adds up to being able to strike many more of your adversary’s targets and escape without them being able to meaningfully engage you and gives you a significant advantage.

It is just like regular camouflage. You cant just throw on a camo jacket and walk around Walmart and be “invisible”. But in the woods, with the right camo, and the right technique, you can move very “stealthily” and if it allows you to get in a position to kill your enemy before they can see and kill you, that is the point.

Second, is an issue with AI I didn’t see anyone bring up. Related to my first point, stealth is about minimizing signatures and that that includes sending/receiving signals. Situations where an aircraft is operating in a highly contested environment means it may not be able to communicate reliably or safely and maintain stealth (they do have MADL but I am sure it can’t work 100% and maintain stealth perfectly). You would have to delegate complete shooting authority to the onboard AI. That is very difficult for many reasons, just one being accountability (who is responsible if an AI commits a war crime?).

[lm28469]

> Besides, it's not invisible despite calling it that. It's visible in the visible spectrum and hearable in the human hearing range

We didn't wait Musk to know planes weren't literally invisible or silent lmao, maybe don't take your military analysis from a man child with 0 experience in the domain.

They either fly high enough that you neither hear nor see them, or low/fast enough that you're dead long before you're even aware something is coming.

Also we already have unmanned aircrafts, a lot of them. Internet army experts will tell you f35 are useless because they're not invisible (duh) meanwhile in eastern Europe people are getting killed by 70+ years old tanks and other ww2 era surplus

[mrtksn]

Musk can be many bad things, but he is also right in some things.

Flying high might make it invisible for human observers but the idea is that it’s not invisible in that wavelength, therefore it must be possible to create devices that can detect it.

Also, this is a brand new machine that is still not ready. Just write it off, liquidate any useful work that might have been done on it and go all in drones. What’s the point of insisting on a job not done when already looks obsolete?

[lm28469]

If a bunch of microphones and binoculars would defeat stealth fighters (or any kind of jets) don't you think someone in the US, Chinese or Russian army would have thought about it ? Just as a reminder the thing is coming at mach 1.5-2 and as soon as it can it'll send a little present coming your way at mach 2-4

> therefore it must be possible to create devices that can detect it.

What's the probability some over worked dude who tweet 20 times an hour came up with something the US military–industrial complex hasn't thought about in the last 50 years ?

Remember the early Ukraine invasion when a couple of bayraktars almost single handedly saved the country during the initial wave ? It was neither stealthy nor fast

https://defence-blog.com/bayraktar-tb2-drones-saved-the-coun...

btw his brand new idea is at least a hundred years old: https://4.bp.blogspot.com/-eyPsCUn0O68/V8jmQwIYR5I/AAAAAAAAK...

[red-iron-pine]

> Remember the early Ukraine invasion when a couple of bayraktars almost single handedly saved the country during the initial wave ? It was neither stealthy nor fast

The Bayraktars are the budget-drone option. A big part of their success was less that they're good, and more that the Russians kept all of their EW and AA turned off to achieve tactical surprise. Which they sort of did, but not enough to anchor the fight, and the budget drones were effective at killing a lot of AA early on, increasing their window of lethality.

Once they got their EW & AA game together the Bayraktars stopped being dangerous very quickly.

[orwin]

I highly doubt f35 will use supersonic speed for anything else than repositioning/travel. In combat theaters i doubt it speed will go higher than Mach 0.8.

And anyway, if you want CAS, to stop an army, A-10 will probably be a dozen time better suited than any multirole, and especially the f35 with its ridiculously low availability rate, or even better in that particular case, an AC-130 (that is probably able to direct a drone fleet in its latest revision, but that was speculation last time i checked)

(the A-10 is the best modern plane in my opinion, i really like the F15, f18 and Rafale (those curves!), because i really like the idea of aircraft carrier but that plane is the best.)

[Albatross9237]

Supersonic speeds would certainly be used for beyond visual range (BVR) engagements. Higher speeds mean you and your missiles have more energy and maneuverability. Since the F-35 needs to use it's afterburners to reach supersonic speeds, it likely wouldn't use the extra fuel just for transport.

[euler_angles]

The default for an F-35 is to be at something like M=0.9 and altitudes of 20-25kft.

[maxglute]

>would have thought about it

IIRC in 2000s there was talk by USAF that stealth is a 10-20 year advantage. Not something undefeatable. The actual degree to which stealth can be / is defeated, we wouldn't know until shooting starts. It's ancient technology at this point.

[_djo_]

Stealth doesn’t stand still. The latest low-observable US aircraft have lower radar and visual signatures than the previous generation.

It’s not an indefinite ‘lead’, of course, but having decades of experience with LO shaping and materials means you can keep making effective incremental improvements.

[maxglute]

IMO keyword is incremental, but we don't know what relative gap is being maintained against detection... if any. As in detection, queuing, targetting might well have caught up, but the virtue of stealth may still be it forces adversaries to have denser sensor networks, even if it means stealth is dead if caught within net. My understanding focus is more on LO materials vs LO shaping in terms of design (to not compromise kinematics), but report suggest coating is not reliable after decades. If LO fighter that doesn't have optimzied LO shape can't fly with LO coating is not a good combo.

[mrtksn]

Ukrainians already built a microphone network to detect incoming missile and planes.

[hkpack]

Microphone network is mostly used for low speed and high noise Iranian Shahed drones which are also made from composite materials and have very low radar visibility. (For example, Moldovan military recently mentioned that they cannot detect them with their old USSR-made radars).

When one is flying towards you, you hear it from few kilometres away for minutes as it has very loud petrol-powered engine.

In contrast, when you hear low passing cruise missile, you will have just few seconds until it passes over you.

[JohnBooty]

Musk: "Ah, yes, let me use microphones to listen for objects approaching at greater than the speed of sound"

I mean, yeah, if you control the ground in the battlespace and have microphone arrays all over the place, sure, maybe acoustic data could be... some part of... some kind of fused sensor network.

But even with perfectly smart software and shitloads of microphones over a huge radius that is going to be a very laggy and imprecise way of measuring (checks notes) objects that may be hundreds of miles away, moving nearly or above the speed of sound, using a spectrum (acoustic waves in air) that only propagates signals at the speed of sound.

[invalidname]

He is wrong about that with the current technology. All you have to do to see that is look at Israel vs. Iran. In a decade drone technology might be enough to do something similar to what the F35 does, but right now the F35 is still the peek of technology. There's a reason Israel ordered more of them.

AI might be able to do a dogfight which is great in terms of flight envelope, but completely unnecessary in modern stealth warfare. Despite everything you heard, stealth does work. It isn't perfect but it destroyed Russia's top of the line anti-aircraft missiles in Iran without a problem. The planes are ghosts, by the time you see them it's already too late.

Drones have the advantage of reduced risk to the pilot but since a human sitting at the base will have to deal with signal delay, transmission jamming and low resolution... The difference in having a pilot physically present is huge. AI is unpredictable and unreliable e.g. Iranians were able to fool a US army drone by sending it signals that made it land. Then they took it apart and reverse engineered it.

[rstuart4133]

Yet in Ukraine we see the reverse effect. Russia has some very advanced planes that are barely used.

It seems to come down to this: for the same money you can buy 1 F35 or 10,000 long range drones. If you are an army with a few SAM's, what you scare you more: a single F35 coming over the hill, or a swarm of drones so large you had no hope of taking them all out?

[JohnBooty]

    Russia has some very advanced planes that are barely used.

I wouldn't extrapolate too much from the fact that the Su-57 are rarely if ever seen in this conflict.

For a long time they apparently only had two operational Su-57s. Apparently now they have 5-7? This is not a "production" aircraft as we would understand it in the west. Its actual stealth ability is also highly suspect. The official photos as well as photos from airshows have shown some sloppy physical construction that would compromise any stealth ability.

     If you are an army with a few SAM's, what you scare you 
     more: a single F35 coming over the hill, or a swarm of 
     drones so large you had no hope of taking them all out?

For some kinds of targets, the drones.

For other kinds of targets, the F35.

While your comparison makes sense from a budget perspective, it's not necessarily realistic though. Nobody has the ability to launch that many drones at once and nobody is flying in a single F35.

Also, respectfully, a lot of the anti-manned-fighter arguments boil down to "drones are really good and useful!"

Which is true, but also not something that anybody disputes. Even the most diehard defender of manned aircraft is going to tell you that drones are a huge part of the future of war. And that manned fighters are niche.

[invalidname]

Russia has nothing like the F35 and Ukraine has planes too (although not as great). Once Ukraine got F16s they had an impact (even though it's a very old model).

The idea of a drone swarm is science fiction at this time. First, it's not 10,000 drones. Maybe the low quality stuff Iran builds is that cheap. A good western drone will be expensive but also of far higher quality.

If you try to send a drone cloud then they are easily detected and you can just shoot them down. If you send them one by one then they get detected one at a time. A few get through as we see with Israel who dealt with well over 30,000 drones/rockets over the past year... But it took them a year to launch 30,000 rockets/drones. They did very little damage.

You need logistics to send them out big logistics are a big target for an F35. If you do it from far away (like sending drones from Iran) then radars have a lot of time to pick them up and shoot them down. If you do it from close by (like Lebanon) then some might get through but the F35 in the sky will destroy you very fast.

Finally, they all need to fly autonomously which is flawed. You can take them down like ducks in a row. Any soldier with a smart scope can just bring down a drone. Not to mention their deep vulnerability to electronic jamming.

I used to think like you as an engineer. But having spoken to people who actually know this stuff I understood the difference. Yes, there is a price disparity which is why the Israeli army has both drones and F35. Different tools for different jobs. A drone can't carry the damage and logistics an F35 can. But sending an F35 to shoot down drones is a remarkable waste of resources. That's why Israel is working on energy/laser based defense systems which will make a swarm of 10,000 drones completely useless but won't even scratch an F35.

[mangamadaiyan]

> Musk can be many bad things, but he is also right in some things

Indeed. A broken clock also tells the time correctly twice a day.

[542354234235]

>It's visible in the visible spectrum and hearable in the human hearing range, which means that you can build detection and tracking systems in those wavelengths instead of pretending that its invisible because its hard to detect at radar wavelengths.

Most anti aircraft radar works by sending out radar pulses and measuring the return. Visual spectrum light is scattered significantly by the atmosphere. It is a physics problem and visual range is just not as good of a medium as HF, UHF, VHF etc radar. Which there are multiple over-the-horizon radars with ranges in the hundreds of miles. And that is before we add in clouds, fog, smoke, snow, or you know, darkness, where radio waves easily penetrate. Visual detection needs far more contrast than radio wave detection. Optical detection is just not well suited.

As for using sound to detect and track an object coming at you at faster than the speed of sound, I think it encapsulates this entire comment’s level of thought and insight.

[wepple]

> remote control + AI?

We might be getting close to advanced AI for a lot of domains, but are we ready to have one making independent decisions with bombs?

I’m not a military expert but I’d much prefer having a human making decisions rather than AI for at least the next decade. I’m not sure that remote connectivity is reliable and high bandwidth enough everywhere for a drone fighter jet

[mrtksn]

We already have drones that are making independent decisions with bombs, but that’s not the point. You can still have people in the loop, people that are not on board.

[wepple]

For 100% of tasks that an F35 can perform?

I know we currently have this capability, but aren’t up to speed if you can rely on it everywhere and every situation.

[lloeki]

Arguably even before drones, missiles have some form of "AI" to autonomously make calls, e.g Tomahawk TERCOM and DSMAC (arguably navigation, but hey, for a missile navigation ends in controlled descent into terrain)

[potato3732842]

This is already the air force's tentative plan. They made several announcements (10+yr ago) when it was new about the F35 being the last manned fighter and then kinda walked that back because PR but kept pursuing it.

[AnimalMuppet]

> It's visible in the visible spectrum and hearable in the human hearing range

Sure. It's just that radar lets you see tens or hundreds of miles away, and visual doesn't. And sound lets you hear where something is only at the speed of sound, which is less useful for something that can fly faster than the speed of sound.

So developing a weapons-quality track from visual and sound data is problematic. That means that, while not invisible, it's "invisible enough".

[orwin]

Once an AI is able to land aircrafts on any airports, highway or carrier, is able to decide on an emergency landing, or on an emergency discharge an won't crash because of a faulty sensor (MCAS) or weird UFOs that trigger those sensors, i won't be opposed to AI-driven aircrafts.

In the meantime, Aircraft+"Ai-driven" drone is a great idea (look up to "nEUROn" if you want an idea of a combat drone capability)

[JohnBooty]

    Having a pilot onboard is overhead and its a limiting 
    factor for the flight envelope as you have to keep it alive.

The advantages of unmanned craft, and disadvantages of manned craft, are numerous.

There are still plenty of cases where you want human pilots present. There's a lot of realtime decision making by pilots when it comes to identifying, selecting, and firing upon targets.

Remotely-piloted drones rely on seamless drone-to-base communications so they can be, you know, remotely piloted. These communications can be denied by an enemy. As far as autonomous drones that can act on their own without a datalink go, let's just say I think the current SOTA in automonous anything shows it's going to be a long time until we're there.

Also a few of the "obvious" advantages of unmanned craft aren't as relevant as one might think....

Drones can obviously be smaller and more manuverable than manned fighters because they don't have to carry that extra weight (pilot, ejector seat, life support, etc) and because they don't need to worry about g-force restrictions as much. However, the sort of close range high-G dogfighting maneuvering seen in movies is vanishingly rare. It's all about BVR (beyond visual range) missile launches.

Additionally, attack aircraft need to carry missiles and bombs. The missiles and bombs need to be a certain size because they need to carry X kilograms of explosive, Y kilograms of fuel, and Z kilograms of guidance electronics. If you want to put 2, 4, 6, 8 of these on a drone, and give the drone itself some kind of large-enough usable flight range, guess what -- it starts approaching the size and cost of a manned fighter pretty quickly.

As far as optical detection of stealth fighters goes...

(deep breath)

Sure, in some cases.

Probably not in ways that are as useful as you hope. First, there are these things called "clouds" and "nighttime" that are going to put a damper on the visual thing, no matter how good the camera and how smart the AI.

Also I want to point out the scale of modern aerial combat. Air to air missiles and surface to air missiles have ranges up to hundreds of miles.

There is probably a role for some kind of sufficiently smart visual spectrum... something... as part of future sensor networks, augmenting radar. Especially in parts of the world (deserts) where you typically have clear skies.

And as far as sound goes? Since Musk mentioned that too?

I'm just going to point to some basic laws of physics here. Gonna be hard to hear things coming in useful amounts of time if they're going near the speed of sound, and impossible to hear them coming if they're going faster than the speed of sound. There's also some significant lag involved that you don't have with EM spectrum stuff. So even with smart enough analysis the best you're going to be able to do is sort of guess that some stealthy fighters are in an approximate area, assuming you control the ground and have a sufficient number of acoustic sensors scattered all about the place and smart enough sensors. Again, this could be part of some kind of wide-spectrum sensor network fusing lots of different data, maybe, but it's not some kind of "gotcha" that invalidates current stealth hardware.