How Hot is Ukraine Gonna Get?

[chrisl]

2 hours ago, Butschi said:

Aren't you shifting the goal posts now? You claimed that "there really isn't any reasoning behind the answers" of LLMs. Providing sources, etc. is not to really a part of reasoning, especially for common knowledge. If I ask you to tell me how fast a spacecraft goes after a given time with speed v and acceleration a, I don't expect you to tell me what textbook you had at school.

Sources is an overloaded word.  I didn't mean references, I mean the actual data it used from the inputs in the reasoning and the pattern it actually followed (not a pattern that it produces because there are a bunch of examples in the training data that give both the logic and the explanation. It's popping out a pattern that "explains the reasoning", but dollars to donuts it didn't actually follow that reasoning in solving the problem - it's just the standard pattern humans use for that problem, so it spat it back.

Quote

None of the papers you provided (interesting ones, really), actually concluded that LLMs are incapable of reasoning, just that their performance for abstract problems or ones not in included in the training data are often not very good. But frankly the same can be argued for humans. Back at school the transferring knowledge to a new problem was what we needed to get an A not to pass the test ...

That's because their "reasoning" isn't reasoning, it's spitting back an interpolation of a best fit based on the input set given some overlap of the novel inputs with the training set.  It dies at the edges for the same reason string theory can't predict anything and I can fit an elephant with a function that has 10^6 free parameters but that function can't describe anything outside the elephant.  High order multiparameter fits get really ugly outside the range of the data they're fit to.

ChatGPT answered that correctly because that problem is solved countless times in its training set with very similar wording.  That's exactly an example of a physics 101/102/201/202 problem that appears repeatedly in problem sets and with complete solutions all over the internet.

Here's a simple one that doesn't appear all over the internet.  I asked it three different ways and it got it wrong all three times.  The correct answer is arccos(10/17).  IIRC, if the spheres slide then it's arccos(2/3).  The radii of the spheres don't matter.

Quote

A sphere of radius a is balanced on top of a sphere of radius b and then given an infinitesimal push so that it rolls off. Find the angle between the vertical and a line through the centers at the moment the moving sphere separates from the fixed one.

 

ChatGPT

To find the angle between the vertical and a line through the centers of the spheres at the moment the moving sphere separates from the fixed one, we need to consider the geometry and dynamics of the rolling motion.

Given:

• Radius of the moving sphere (top sphere): a
• Radius of the fixed sphere (bottom sphere): b

Let's denote:

• θ: Angle between the vertical (direction of gravity) and a line through the centers of the spheres at the moment of separation.

At the moment of separation, the point of contact between the spheres has moved to the highest point on the moving sphere's surface that is in contact with the fixed sphere.

To find θ, we use the condition that the contact point must satisfy the rolling without sliding condition:

1. Rolling Without Sliding Condition: The rolling sphere has a linear velocity and an angular velocity such that the point of contact has zero relative velocity with respect to the fixed sphere.

2. Geometry of Contact Point: At the moment of separation, the contact point is at the topmost point on the surface of the moving sphere that is in contact with the fixed sphere.

The vertical distance from the center of the fixed sphere to the contact point on the moving sphere is b+acos⁡θ, where θ is the angle we seek.

The horizontal distance from the center of the fixed sphere to the contact point on the moving sphere is asin⁡θ.

For rolling without sliding, the relative velocity condition at the contact point must be zero, which gives us the relation:

tan⁡θ=asin⁡θ/(b+acos⁡θ)

To find θ, we solve this equation. This involves a transcendental equation which can be tricky to solve analytically. Typically, this would be solved numerically or using iterative methods. However, the exact angle θ can be found using the inverse tangent function after substitution.

Therefore, the angle θ between the vertical and a line through the centers of the spheres at the moment the moving sphere separates from the fixed one is given by:

θ=tan⁡−1(a/b)

This angle represents the critical angle at which the moving sphere just separates from the fixed sphere due to the infinitesimal push given to it.

there's no reasoning in there.  It's telling a story like a Phys 101 student but doesn't have even a simple physical model. It's following a standard pattern and filling in the blanks, but it's nice sentences so it looks like reasoning.  Fancy Mad Libs.

Quote

Now, we humans, although our brains are also just parametric fits, are capable of forming models of how our world works. Often just heuristics but even learning algebraic formulae.

I've stabbed a few neurons in dishes with pointy sticks, and even tried to train them to do absurdly simple things, and I've never been entirely convinced that the CS model of neural computing is as close to actual neural computing as people seem to act that it is.  People have something like 10^12 neurons and ~10^14 interconnects (synapses).  That's an enormous number of free parameters, and is used to fit a far, far smaller input data set than any of the chatbots have, but the chatbots still kind of suck, but sound good doing it. Given the relative number of free parameters in humans, you'd expect them to go off the rails far more often than they alraedy do (but it maybe explains conspiracy theorists...)

Humans develop models that are much more complex than simply heuristics plus learning some algebraic formulae.  And with far less training data than the Chatbots.  I've read way more of the internet than I care to think about, a ton of books, and done a ton of problems. But I've ingested only a tiny fraction of the input data that your average Chatbot has, but I get annoyed by the second paragraph when they start to BS me like a 1st year undergrad who doesn't know the answer to something really that they probably read the day before.  There's no model of the universe in there.  So you can expand the context arbitrarily, and all you're going to get is more complicated wrong answers if there's not a physical model.

 

 

 

 

Edited July 8 by chrisl

[ASL Veteran]

3 hours ago, The_Capt said:

Kind of my thinking. I mean are we really trying to hide 30-60 ton vehicles that run on small explosions? If we can do that then a small bird sized UAS is basically invisible. So this really is less about individual drones v a platoon of vehicles, and more about battle space management.  That platoon is going to get picked up well back, and if it somehow manages to stay hidden its logistics tail will definitely give it away.  Once detected, there is a suite of systems to go hunt and kill them from good old artillery of all shapes and sizes, to NLOS ATGMs, to drones, to UGVs, to other vehicles, to mines.

So it is great if you have some sort of wireless and silent comms system in your platoon…but the autonomous drones will be sent right to your grid square along with a bunch of other stuff.  After thinking on this, I am not sure passive or active will really matter.  Once detected they can and will be engaged. Then even if they managed to hold off the first swarm wave, all those guns blasting off like a Serbian wedding is damn sure to give up their position.  And in come the PGM and cluster munitions.

A foundational element of land warfare is an ability to seek cover from observation and fires.  Once that is gone, land warfare theory is pretty much dead.  At that point we need a new theory, probably closer to maritime warfare.  And before someone pipes up that there is PD in maritime warfare, well 1) we have no evidence that our current theories of war on the sea are still effective and 2) because land is still…well land…the dynamics are fundamentally different.  We can disaggregate on land far more than at sea - so I think there are adjustments to be made.

So here we are…still.

 

[The_Capt]

36 minutes ago, chrisl said:

Sources is an overloaded word.  I didn't mean references, I mean the actual data it used from the inputs in the reasoning and the pattern it actually followed (not a pattern that it produces because there are a bunch of examples in the training data that give both the logic and the explanation. It's popping out a pattern that "explains the reasoning", but dollars to donuts it didn't actually follow that reasoning in solving the problem - it's just the standard pattern humans use for that problem, so it spat it back.

That's because their "reasoning" isn't reasoning, it's spitting back an interpolation of a best fit based on the input set given some overlap of the novel inputs with the training set.  It dies at the edges for the same reason string theory can't predict anything and I can fit an elephant with a function that has 10^6 free parameters but that function can't describe anything outside the elephant.  High order multiparameter fits get really ugly outside the range of the data they're fit to.

ChatGPT answered that correctly because that problem is solved countless times in its training set with very similar wording.  That's exactly an example of a physics 101/102/201/202 problem that appears repeatedly in problem sets and with complete solutions all over the internet.

Here's a simple one that doesn't appear all over the internet.  I asked it three different ways and it got it wrong all three times.  The correct answer is arccos(10/17).  IIRC, if the spheres slide then it's arccos(2/3).  The radii of the spheres don't matter.

there's no reasoning in there.  It's telling a story like a Phys 101 student but doesn't have even a simple physical model. It's following a standard pattern and filling in the blanks, but it's nice sentences so it looks like reasoning.  Fancy Mad Libs.

I've stabbed a few neurons in dishes with pointy sticks, and even tried to train them to do absurdly simple things, and I've never been entirely convinced that the CS model of neural computing is as close to actual neural computing as people seem to act that it is.  People have something like 10^12 neurons and ~10^14 interconnects (synapses).  That's an enormous number of free parameters, and is used to fit a far, far smaller input data set than any of the chatbots have, but the chatbots still kind of suck, but sound good doing it. Given the relative number of free parameters in humans, you'd expect them to go off the rails far more often than they alraedy do (but it maybe explains conspiracy theorists...)

Humans develop models that are much more complex than simply heuristics plus learning some algebraic formulae.  And with far less training data than the Chatbots.  I've read way more of the internet than I care to think about, a ton of books, and done a ton of problems. But I've ingested only a tiny fraction of the input data that your average Chatbot has, but I get annoyed by the second paragraph when they start to BS me like a 1st year undergrad who doesn't know the answer to something really that they probably read the day before.  There's no model of the universe in there.  So you can expand the context arbitrarily, and all you're going to get is more complicated wrong answers if there's not a physical model.

 

 

 

 

We too have been through a bit of an AI scourge.  LLM can do stuff but their real value added is a work in progress.  What “AI” is good at is taking ridiculous amounts of data and pulling “stuff we ask for” from it.  The AI does not understand context or background, it does not have history or narratives. But it can push all that data around like no one’s business.  

As it relates to warfare, AI/ML/whatever, is at a strange crossroads.  Clausewitz was right, war is actually pretty simple.  Find and kill the other guy until he stops working. What is hard is actually doing that simple.  Now AI can do simple, to a fault.  And it sidesteps a lot of the hard - doesn’t need sleep, never gets scared, wife doesn’t mess around on him while he is in country, doesn’t care about cold, wet or hungry. So it is a good fit, at least on the surface. It can see, and now it can remember.  It can recognize and chase.  It can even predict with some help.  Where the trouble starts is in the pairing with humans.  We still do not really know what the human-machine team will look like but it is becoming clearer.

What AI can do is create synthetic mass. So concentrations of empowered capability built from a core other than squishy human flesh bags. That is a game changer but the quality of the AI is key. It needs to be able to learn, and teach.  It needs to be able to know when to act and when not to.  It needs to understand consequences.  Some of this is emerging.  Other parts are a long way off.

[chrisl]

6 minutes ago, The_Capt said:

We too have been through a bit of an AI scourge.  LLM can do stuff but their real value added is a work in progress.  What “AI” is good at is taking ridiculous amounts of data and pulling “stuff we ask for” from it.  The AI does not understand context or background, it does not have history or narratives. But it can push all that data around like no one’s business.  

As it relates to warfare, AI/ML/whatever, is at a strange crossroads.  Clausewitz was right, war is actually pretty simple.  Find and kill the other guy until he stops working. What is hard is actually doing that simple.  Now AI can do simple, to a fault.  And it sidesteps a lot of the hard - doesn’t need sleep, never gets scared, wife doesn’t mess around on him while he is in country, doesn’t care about cold, wet or hungry. So it is a good fit, at least on the surface. It can see, and now it can remember.  It can recognize and chase.  It can even predict with some help.  Where the trouble starts is in the pairing with humans.  We still do not really know what the human-machine team will look like but it is becoming clearer.

What AI can do is create synthetic mass. So concentrations of empowered capability built from a core other than squishy human flesh bags. That is a game changer but the quality of the AI is key. It needs to be able to learn, and teach.  It needs to be able to know when to act and when not to.  It needs to understand consequences.  Some of this is emerging.  Other parts are a long way off.

I don't care if it can "reason".  I'd be happy if it can show me on an N-dimensional plot how it parametrized and clustered the input data so I can do the equivalent of slapping it on the nose with a rolled up newspaper and say "No! Don't do that to that cluster!"

[dan/california]

40 minutes ago, The_Capt said:

We too have been through a bit of an AI scourge.  LLM can do stuff but their real value added is a work in progress.  What “AI” is good at is taking ridiculous amounts of data and pulling “stuff we ask for” from it.  The AI does not understand context or background, it does not have history or narratives. But it can push all that data around like no one’s business.  

As it relates to warfare, AI/ML/whatever, is at a strange crossroads.  Clausewitz was right, war is actually pretty simple.  Find and kill the other guy until he stops working. What is hard is actually doing that simple.  Now AI can do simple, to a fault.  And it sidesteps a lot of the hard - doesn’t need sleep, never gets scared, wife doesn’t mess around on him while he is in country, doesn’t care about cold, wet or hungry. So it is a good fit, at least on the surface. It can see, and now it can remember.  It can recognize and chase.  It can even predict with some help.  Where the trouble starts is in the pairing with humans.  We still do not really know what the human-machine team will look like but it is becoming clearer.

What AI can do is create synthetic mass. So concentrations of empowered capability built from a core other than squishy human flesh bags. That is a game changer but the quality of the AI is key. It needs to be able to learn, and teach.  It needs to be able to know when to act and when not to.  It needs to understand consequences.  Some of this is emerging.  Other parts are a long way off.

And one or two breakthroughs in narrow areas can be extremely consequential. 

[Battlefront.com]

7 hours ago, chrisl said:

His example was reasonable - it didn't assume predictable path so much as that newton's laws are still valid at the drone scale.  If a drone has a particular momentum at point (x,y,z,t) and maximum acceleration of a_max, there's a well defined limited volume available for it to occupy at some later point (x1,y1,z1,t1).  You just have to fill that volume of spacetime with stuff at a density high enough that at least one of piece of your shrapnel is guaranteed to intesect the volume occupied by the drone.

Yes, there is the "spray and pray" type defense.  And yes, in theory you can plaster a space large enough to account for a drone suddenly pulling a 12g maneuver at 15m/s.  But really... how practical is such a solution?  By that I mean if you have a single point defense system that is being attacked from all sides by a swarm, is it really practical to pump that much shrapnel into the airspace?  I don't think it is. 

Smaller airburst munitions, that are aimed close to where the drone is likely to be, is more practical in theory, however this runs into the problem of requiring a fairly predictable path a fair ways away from the point of interception.  This is why I'm so thumbs down on APS.

You waive away friendly casualties, but the West doesn't think that way.  So maybe the Russians could field something like you describe, but the West won't.

Steve

[ArmouredTopHat]

 

I found an excellent channel from a T-64 crew fighting in Ukraine. Sorta shows what they get up to with both direct and indirect fire. We should bear in mind that things like FPVs are not turning every single vehicle to scrap just yet. The first video in particular highlights how useful a direct fire capability can be and how ATGMs can be mitigated by short, effective fire missions. 
 

This particular video seems to feature them doing indirect fire in particular, quite accurate too at that. 
 

Really do like the extended footage that shows the reasonable amount of thought the tankers are putting with regards to moving around, firing for short periods of time then moving. They even show a load of videos with them maintaining the vehicle between runs. The tank crew has been in combat since 2022, largely with T-64 it seems like. Not bad for such an old tank!

Edited July 8 by ArmouredTopHat

[ArmouredTopHat]

2 minutes ago, Battlefront.com said:

You waive away friendly casualties, but the West doesn't think that way.  So maybe the Russians could field something like you describe, but the West won't.

Trophy features a similar potential problem for friendly fire yet a lot of countries are looking into adopting the system or others similar to it. It seems the risk of potentially hitting friendlies is outweighed by the advantages such a system bestows. I wonder if the Israelis have any data on friendly fire caused by the system in action?

[Battlefront.com]

7 hours ago, poesel said:

You repeat kimbosbreads arguments here. And I wrote him a PM to not drag this one out but no luck.

The fun part is that we all agree and say the same thing using different words.

TL;DR: you can go totally silent with some (probably small) loss of coordination. And even if you communicate, it can be made very hard to detect and nearly impossible to track.

Yes, exactly that.  So all this fuss over EW is really just a temporary annoyance for drones.  All it does is force the opponent to invest in autonomous capabilities.  This will soon be available to anybody on the battlefield, including resource poor non-state actors.

7 hours ago, poesel said:

Why would the enemy know where my vehicles are from the information my vehicles share with each other? See above for drone to drone communication. Same problem, same fix.

Already partially covered.  But to add to what was just said about the realities of ISR, each vehicle in the network needs to broadcast its position in order for the network to know where it is.  Intercepting that information, at least sufficient to identify the location of the source, is something that can be done with today's technology.  Making sense of it using AI makes acting upon that information far more practical than it was in the old days.

EW still has a future in terms of disrupting and defending such signals.

7 hours ago, poesel said:

Yes, there are places where PD does not work. Try to avoid them.

A friendly base that cannot defend itself versus drones is not a base.

Trucks in tight formation is a dumb idea for other reasons, too.

There's only so much avoidance possible.  Not going through civilian areas is pretty much impossible, so that right there is a significant problem.  The problems that logistics have to deal with is very long already, so this makes things even more complicated for them.  Bases will need to be able to defend themselves, but this again means the bases will have to be situated away from civilian infrastructure.  That's not necessarily easily done.

My point is that there's a host of complications for point defense systems aside from them actually tracking and hitting targets.  These complications are not trivial to solve and ignoring them would be a very bad idea.

7 hours ago, poesel said:

I assume our common assumption is that the defender has been detected, and the attack will be done by a bunch of drones.

If the defender is supposedly undetected, then we talk about a free roaming drone swarms and that is another can of worms.
If you don't attack with drones and use e.g. artillery, then the PD we talk about here has no use at all.

We should assume going forward that detection is inevitable.  Any proposed defense that doesn't have that as a primary design criteria is almost certainly going to wind up being useless. 

Free roaming loitering munitions are already "a thing".  We don't need to talk about them, we can just watch videos of what the results are.

ISR UAS feeding coordinates to artillery is already "a thing".  Which is a major reason why I don't think PD is the solution we need to aim for.

7 hours ago, poesel said:

chrisl already explained the upper part nicely.

I don't know how far danger close is for you, but from my ballpark calculations, I guess we are talking about 200m out.

I responded to chrisl, but I will expand here...

I predict "danger close" to be measured in 10s of meters, not 100s.  If I know you have a PD capability, I will fly my drones at ground level using terrain features for cover.  I've covered this ground several times now so I won't bother with it again.

Another possibility is to fly UAS into ambush positions when you know there are no PDs in the area.  This is really good against attacking forces, but it is also good against logistics for a defending force.  A well positioned ambush might not even give the defending PD 100m warning.  Might not even given it 20m warning.

Oh, and then there's UGVs which PD isn't even designed to handle.  Or a hybrid approach where UAS delivers UGVs or UGVs deliver UAS. 

These are the things designers of counter systems need to take into account.  Not just the things we know are available today, but the things that we should reasonably expect will be on the battlefield soon.  Or at a minimum, soon enough to make it clear that some counter systems are a really bad investment.

Steve

[Battlefront.com]

22 minutes ago, ArmouredTopHat said:

Trophy features a similar potential problem for friendly fire yet a lot of countries are looking into adopting the system or others similar to it. It seems the risk of potentially hitting friendlies is outweighed by the advantages such a system bestows. I wonder if the Israelis have any data on friendly fire caused by the system in action?

It really comes down to the risk of friendly casualties.  The point I've been making is that the sorts of PD systems we've been discussing here against the sorts of targets that they will encounter in their deployed lifetime has far greater potential for causing friendly casualties.

As was argued back in the not-to-be-revived TheTankIsDead discussion, APS has never been deployed in large numbers in wide ranging combat situations.  So, there's that.

I expect that whatever friendly casualty statistics the IDF has are classified.  I know I would keep them classified if I were them, unless the number were 0. Which I bet it isn't.

Steve

[ArmouredTopHat]

19 minutes ago, Battlefront.com said:

It really comes down to the risk of friendly casualties.  The point I've been making is that the sorts of PD systems we've been discussing here against the sorts of targets that they will encounter in their deployed lifetime has far greater potential for causing friendly casualties.

As was argued back in the not-to-be-revived TheTankIsDead discussion, APS has never been deployed in large numbers in wide ranging combat situations.  So, there's that.

I expect that whatever friendly casualty statistics the IDF has are classified.  I know I would keep them classified if I were them, unless the number were 0. Which I bet it isn't.

Steve

My only counter point in the absence of data is that the Israelis seem to consider it a worthwhile risk, which should be at least noted given how conserving of their manpower they tend to be. If the system was killing dozens of soldiers then they would never use it. 

[Battlefront.com]

If anybody here doesn't understand why I'm so hostile to the combo of industry, military, and civilians in weapons funding and procurement, I present to you a timely (short) article on Exhibit A - the F-35:

https://thehill.com/policy/defense/4754671-congress-f-35-program-problems/

https://www.defensenews.com/air/2024/04/15/f-35s-to-cost-2-trillion-as-pentagon-plans-longer-use-says-watchdog/

The total projected cost is now $2 TRILLION over its lifetime.  A cost that is for sure going to go up.  But even without that factored in, each plane costs around $100m to acquire and over $6m per year to keep it in service.  And it doesn't even work very well, as the above articles point out.  It was delivered many years late, 10 years on it still isn't working as advertised, but it does cost 2x more than was originally priced at.

Despite all of the problems with the F-35, including the fact they still don't work reliably, a significant number of nations have signed up to buy them.  This proves that customer demand ≠ evidence all is well with the system.

I do think they will eventually get the F-35 working mostly as expected.  However, I also think all that money could have been better spent on something else to maintain air capabilities while the R&D was given time to mature into something practical.

Steve

[Battlefront.com]

1 minute ago, ArmouredTopHat said:

My only counter point in the absence of data is that the Israelis seem to consider it a worthwhile risk, which should be at least noted given how conserving of their manpower they tend to be. If the system was killing dozens of soldiers then they would never use it. 

They also thought their pre-October 2023 defenses were adequate.  They also thought launching a massive ground war into Gaza was a good idea.  Not to mention what their concept of a good idea was for dealing with Hamas more generally.

Which is to say, I don't share your faith in the IDF's decision making as much as you do.  The F-35 post I made while you posted the above is further evidence that taking things at face value really isn't a good idea.

Steve

[ArmouredTopHat]

1 hour ago, Battlefront.com said:

If anybody here doesn't understand why I'm so hostile to the combo of industry, military, and civilians in weapons funding and procurement, I present to you a timely (short) article on Exhibit A - the F-35:

https://thehill.com/policy/defense/4754671-congress-f-35-program-problems/

https://www.defensenews.com/air/2024/04/15/f-35s-to-cost-2-trillion-as-pentagon-plans-longer-use-says-watchdog/

The total projected cost is now $2 TRILLION over its lifetime.  A cost that is for sure going to go up.  But even without that factored in, each plane costs around $100m to acquire and over $6m per year to keep it in service.  And it doesn't even work very well, as the above articles point out.  It was delivered many years late, 10 years on it still isn't working as advertised, but it does cost 2x more than was originally priced at.

Despite all of the problems with the F-35, including the fact they still don't work reliably, a significant number of nations have signed up to buy them.  This proves that customer demand ≠ evidence all is well with the system.

I do think they will eventually get the F-35 working mostly as expected.  However, I also think all that money could have been better spent on something else to maintain air capabilities while the R&D was given time to mature into something practical.

Steve

A little bit of misinformation about the F-35 here, which I dont blame you because the program has received a lot of unfair flak.  

F-35 is actually -cheaper- currently than a number of 4th gen platforms to buy these days, and that cost ratio only becomes more favourable when you factor in the older age of a lot of those 4th gen platforms, which get more expensive every year to run. Its also by far the cheapest stealth platform by any metric. 

Jets in general are pretty expensive to operate but F-35 mitigates those costs due to the scale of its production. We talked about how the west needs to improve on its production scales? The F-35 shows that super high end platforms can be made much more economical when produced at a proper scale. The project itself was a remarkable achievement of multinational development.

F-35 gets a lot of controversy but it has actually enjoyed less accidents per flight hour than most US jets when you actually look at the long term data. Its had far less accidents than F-16s, F-15s or other legacy platforms per 100k flight hours, just compare the accident rates from F-35 to F-16, which in itself is an excellent jet. F-35 has had less mishaps per flight hour than pretty much any US Jet in service, and a lot less deaths. 

https://www.safety.af.mil/Portals/71/documents/Aviation/Aircraft Statistics/FY21_F-35.pdf

https://www.safety.af.mil/Portals/71/documents/Aviation/Aircraft Statistics/F-16.pdf

Its a same case for readiness rates, which people often cite. Readiness rates are in fact in line with other jets in service, which always tend to be low in peacetime. Look up the readiness rates of Israeli F-35s (that have been running ops for most of this year with zero issue, including scuffles with Russian AD based in Syria) and you will find they have had no real sustainment problems

The high development cost is worth it for the capabilities as well. Why do you think everyone is trying to buy them? 5th gen planes utterly slaughter 4th gens in pretty much every combat exercise (Were talking 20-1 kill ratios), and F-35s can act as major force multipliers to less capable planes with its sensor integration. If you want to do anything outside of extreme BVR and bomb tossing then you simply need that stealth capability and battlefield sensor mastery against a near peer/ peer opponent. They truly are terrifying pieces of kit that are the right direction when it comes to mastery over BVR air combat. Making over a -thousand- 5th gens is no mean feat either. In most metrics the program has been a monumental success. 

There is a reason people are buying them, because they are quite simply the best and for a paying customer, quite affordable. Its really not a good example if you want to point out bad procurement policy. 

https://lexingtoninstitute.org/the-f-35-is-the-safest-and-most-capable-fighter-the-u-s-military-has/

 

Edited July 8 by ArmouredTopHat

[dan/california]

5 minutes ago, ArmouredTopHat said:

A little bit of misinformation about the F-35 here, which I dont blame you because the program has received a lot of unfair flak.  

F-35 is actually -cheaper- currently than a number of 4th gen platforms to buy these days, and that cost ratio only becomes more favourable when you factor in the older age of a lot of those 4th gen platforms, which get more expensive every year to run. Jets in general are pretty expensive to operate but F-35 mitigates those costs due to the scale of its production. We talked about how the west needs to improve on its production scales? The F-35 shows that super high end platforms can be made more economical when produced at a proper scale. The project itself was a remarkable achievement of multinational development.

F-35 gets a lot of controversy but it has actually enjoyed less accidents per flight hour than most US jets when you actually look at the long term data. Its had far less accidents than F-16s, F-15s or other legacy platforms per 100k flight hours, just compare the accident rates from F-35 to F-16, which in itself is an excellent jet. F-35 has had less mishaps per flight hour than pretty much any US Jet in service, and a lot less deaths. 

https://www.safety.af.mil/Portals/71/documents/Aviation/Aircraft Statistics/FY21_F-35.pdf

https://www.safety.af.mil/Portals/71/documents/Aviation/Aircraft Statistics/F-16.pdf

Its a same case for readiness rates, which people often cite. Readiness rates are in fact in line with other jets in service, which always tend to be low in peacetime. Look up the readiness rates of Israeli F-35s (that have been running ops for most of this year with zero issue, including scuffles with Russian AD based in Syria) and you will find they have had no real issues. 

The high development cost is worth it for the capabilities as well. Why do you think everyone is trying to buy them? 5th gen planes utterly slaughter 4th gens in pretty much every combat exercise, and F-35s can act as major force multipliers to less capable planes with its sensor integration. If you want to do anything outside of extreme BVR and bomb tossing then you simply need that stealth capability. They truly are terrifying pieces of kit that are the right direction when it comes to mastery over BVR air combat. Making over a -thousand- 5th gens is no mean feat either. In most metrics the program has been a monumental success. 

There is a reason people are buying them, because they are quite simply the best and for a paying customer, quite affordable. Its really not a good example if you want to point out bad procurement policy. 

https://lexingtoninstitute.org/the-f-35-is-the-safest-and-most-capable-fighter-the-u-s-military-has/

 

I agree the F-35 is coming mostly good in the end, albeit a decade late and over budget by the GDP a medium sized European country. Now explain the littoral combat ship fiasco.

Edited July 8 by dan/california

[ArmouredTopHat]

6 minutes ago, dan/california said:

I agree the F-35 is coming mostly good in the end, albeit a decade late and over budget by the GDP a medium sized European country. Now explain the littoral combat ship fiasco.

The Littorals were an utter disaster and serve as a far better example of a bad procurement and how role / capability creep can be crippling. They also are a cautionary tale of how modular designs are not always the best option. I still dont know why the hell they were built in the first place.

You will note how no one outside of the US was lining up to buy them. 

Edited July 8 by ArmouredTopHat

[ArmouredTopHat]

26 minutes ago, dan/california said:

GDP a medium sized European country.

A further note on this as well:

https://www.forbes.com/sites/beltway/2011/06/27/massive-cost-estimate-for-fighter-program-is-misleading/?sh=48f7cfab6625

The whole 2 trillion thing is frequently misleading. (That cost is literally for half a CENTURY of service) Using the same metrics, its going to cost America over four times that in order for them to maintain their far less capable 4th gen inventories. If you want to go after wastes of money, go after A-10s or something that currently serve no purpose at all in a near peer / peer environment with any modicum of air defence. 

Edited July 8 by ArmouredTopHat

[chrisl]

2 hours ago, Battlefront.com said:

Yes, there is the "spray and pray" type defense.  And yes, in theory you can plaster a space large enough to account for a drone suddenly pulling a 12g maneuver at 15m/s.  But really... how practical is such a solution?  By that I mean if you have a single point defense system that is being attacked from all sides by a swarm, is it really practical to pump that much shrapnel into the airspace?  I don't think it is. 

Smaller airburst munitions, that are aimed close to where the drone is likely to be, is more practical in theory, however this runs into the problem of requiring a fairly predictable path a fair ways away from the point of interception.  This is why I'm so thumbs down on APS.

You waive away friendly casualties, but the West doesn't think that way.  So maybe the Russians could field something like you describe, but the West won't.

Steve

I don't disagree with any of that.

In the case of a single drone, or maybe 2, that get through the outer layers it may be reasonable (as in physically possible) to have some kind of airburster that will fill their predicted location sphere with little sharp bits of fast moving stuff, depending on the actual parameters of the drone and the gun.  

But I wouldn't spend a lot of money on it.  It's likely to be a relatively brittle defense, both because it's PD and because it depends a lot on the external environment and the attacker.  I'd spend my money on layers of more distributed defense that loses performance slowly as you break pieces of it.

And it's only @poesel that waves away friendly casualties. I think it's an unpleasant environment that I would avoid having live humans anywhere in the area.  I'm sending robots.

[chrisl]

1 hour ago, Battlefront.com said:

If anybody here doesn't understand why I'm so hostile to the combo of industry, military, and civilians in weapons funding and procurement, I present to you a timely (short) article on Exhibit A - the F-35:

https://thehill.com/policy/defense/4754671-congress-f-35-program-problems/

https://www.defensenews.com/air/2024/04/15/f-35s-to-cost-2-trillion-as-pentagon-plans-longer-use-says-watchdog/

The total projected cost is now $2 TRILLION over its lifetime.  A cost that is for sure going to go up.  But even without that factored in, each plane costs around $100m to acquire and over $6m per year to keep it in service.  And it doesn't even work very well, as the above articles point out.  It was delivered many years late, 10 years on it still isn't working as advertised, but it does cost 2x more than was originally priced at.

Despite all of the problems with the F-35, including the fact they still don't work reliably, a significant number of nations have signed up to buy them.  This proves that customer demand ≠ evidence all is well with the system.

I do think they will eventually get the F-35 working mostly as expected.  However, I also think all that money could have been better spent on something else to maintain air capabilities while the R&D was given time to mature into something practical.

Steve

Augustine's Laws are holding up at least as well as, if not better than, Moore's law.

https://en.wikipedia.org/wiki/Augustine's_laws

Scroll down to the bottom for the plot for Law 16:

Quote

In the year 2054, the entire defense budget will purchase just one tactical aircraft. This aircraft will have to be shared by the Air Force and Navy 3½ days each per week except for leap year, when it will be made available to the Marines for the extra day.

 

[sburke]

1 hour ago, Battlefront.com said:

It really comes down to the risk of friendly casualties.  The point I've been making is that the sorts of PD systems we've been discussing here against the sorts of targets that they will encounter in their deployed lifetime has far greater potential for causing friendly casualties.

As was argued back in the not-to-be-revived TheTankIsDead discussion, APS has never been deployed in large numbers in wide ranging combat situations.  So, there's that.

I expect that whatever friendly casualty statistics the IDF has are classified.  I know I would keep them classified if I were them, unless the number were 0. Which I bet it isn't.

Steve

the fun item that hasn't been discussed - to beat a dead horse somewhat - is again the logistical system.  It was suggested earlier that the PD would be ubiquitous i.e. it would be mounted on pretty much everything.  So I'm cruising along in my wonderful merry way in a supply column for that armor combat team 10 km up the road.. and a drone attack happens... and all my PD starts opening up... in the middle of my column of fuel trucks.  No problem there right?

[holoween]

1 hour ago, sburke said:

the fun item that hasn't been discussed - to beat a dead horse somewhat - is again the logistical system.  It was suggested earlier that the PD would be ubiquitous i.e. it would be mounted on pretty much everything.  So I'm cruising along in my wonderful merry way in a supply column for that armor combat team 10 km up the road.. and a drone attack happens... and all my PD starts opening up... in the middle of my column of fuel trucks.  No problem there right?

What is worse? The drones hitting or the point defense?

Also if we have sensors that are good and cheap enough to recognize a truck we can also put it in the point defense and have it not target where our own trucks are.

[Maciej Zwolinski]

On 7/7/2024 at 1:12 AM, Battlefront.com said:

We are seeing nothing of the sort.  We are seeing carefully controlled, super expensive, very limited weapons systems that are knocking down unrealistically simplistic targets. They also have zero, and I mean ZERO, chance of being scalable to the extent they are needed to protect anything other than fixed facilities.  And even then, I don't think they hold out much promise.

Which gives rise to a very interesting question, why this kind of situation created defence dominance. On the face of it, the situation where anything that is attacked is killed, should result in problems for the defender, who has to take at least some fixed positions which can be reconnoitred and targeted. Whereas the attacker can be somewhere out there on his approach march or scooting around. On the face of it, a similar paradigm existed at the time when US army expected tac nukes to be used as a matter of course from day 1 (about 1955-1965). And then, the idea of appropriate behaviour was to be always on the move, always spread out and always attacking. Aboiut the opposite of what we are seeing now.

[Maciej Zwolinski]

5 hours ago, Battlefront.com said:

You waive away friendly casualties, but the West doesn't think that way.  So maybe the Russians could field something like you describe, but the West won't.

Steve

However, that approach is not absolute. IIRC for a time, ERA was thought to be unusable in the West because it could kill friendly infantry and the projected cost of resultant lawsuits made peacetime officers quake in their boots. 

[chrisl]

26 minutes ago, Maciej Zwolinski said:

Which gives rise to a very interesting question, why this kind of situation created defence dominance. On the face of it, the situation where anything that is attacked is killed, should result in problems for the defender, who has to take at least some fixed positions which can be reconnoitred and targeted. Whereas the attacker can be somewhere out there on his approach march or scooting around. On the face of it, a similar paradigm existed at the time when US army expected tac nukes to be used as a matter of course from day 1 (about 1955-1965). And then, the idea of appropriate behaviour was to be always on the move, always spread out and always attacking. Aboiut the opposite of what we are seeing now.

The defender at least can make some attempt at concealment and cover, and can stockpile supplies.  The attacker has to move and has to maintain a longer and more exposed logistics tail.  

[The_Capt]

2 hours ago, holoween said:

What is worse? The drones hitting or the point defense?

Also if we have sensors that are good and cheap enough to recognize a truck we can also put it in the point defense and have it not target where our own trucks are.

Can you start to see how impractical that is going to be?  We are talking logistics lines of potentially hundreds of kms.  So every convoy now has to behave like a fighting unit - covering arcs and networked.  Then there is the armor problem because no PD is going to be 100%, likely not even close. So logistics lines are harried and hit from all directions at long ranges from multiple systems. You honestly think we can keep that up for any length of time? How many fuel trucks do we have in a brigade? And we have not even talked about mines/UGVs that can be dropped along that route. EW won’t work because these systems will be fully autonomous.  And ISR will know exactly where those convoys are.

Once again, the forward looking crowd on this thread are not the impractical ones.  We have got a mountain of real world evidence of the environment, and it is plausible to extrapolate from that.  The military conservatives have got some pretty tenuous theories, trends in military procurement that pre-date this war, and corporate shilling videos.

Here is a crazy idea…way out there…let’s build lighter forces with much smaller logistics tails…and have no convoys to hit.  I know…insane.