Too funny, I went through flight school with Colin. Great guy and glad to see he picked up command but disagree with him on the benefits of BFM and the true value of manned fighter jets over the coming decades. The US military is caught between two wars. One, against opponents who can’t really hit back where fighter jets are overkill. Better off with AC-130s and the like. The other against opponents who can hit back, where the battle will be won or lost in space, cyber, economically, and possibly through nuclear warfare. Even with regards to fighters, the amount of money spent on BFM is really a boondoggle. It’s like spending billions training the infantry to knife fight. It’s not Vietnam and the last gunfighter with regards to missiles and targeting. Is not the AIM-9X targeting AI itself?
A great book that rings more true than ever talks about the obsolescence of weapons systems. I believe it was Freedman[1] A key thesis is that as a predominant weapon system becomes more vulnerable, it becomes more expensive, as it’s burdened with more defensive measures to survive from cheaper more expendable systems. In World War 2 the canonical example was the battle ship and the aircraft. Now it’s the fighter jet and the missile (or drone) etc. Just look at the ridiculous state of the JSF.
Manned missile trucks are valuable for countering electronic attack, but not dogfighting. Still, practice dogfighting is one of the most fun things you can do, too bad it hasn’t really been applicable since Vietnam. This post from the other day reminded me how fun it is and seems like a great fit for VR.[2] Onward last Samurais! Just try not to spend all the taxpayers money on the way out.
I've been clocking in 10+ hours/wk dogfighting in VR with a decent HOTAS setup since quarantine. You're absolutely right about how fun it is. No other VR experience has been such a consistent draw for me. It's just an excellent place between a fast paced shooter and a strategy game. I love thinking about relative energy and the capabilities of the enemy plane and how they match up with mine in order to come up with an engagement strategy. I also love the tension as you identify a target and you're both trying to come into the engagement from a favorable position. It reminds me of match sprints in track cycling. Very much recommend it.
Do you do anything to deal with the motion sickness, or did you just never experience it?
I can play games where real-world motion matches in-game motion (Beat Saber, PokerStars VR, Superhot VR, Zombie Training Simulator, Space Pirate Trainer, etc.), but any sort of driving or flying makes me sick. Even games that make me use the controller to walk long distances make me dizzy and nauseous.
I started watching some DCS dogfights on YouTube because of the last MSFS post. It looks incredible. If I could afford it, I'd definitely buy a stick/throttle and try to get into it.
Particularly Growling Sidewinder has a lot of great videos and I can't stop watching them.
The military understands this. The f35 is designed for long range kills.. not dog fighting.
But the ability to control the plane is the minimum capability of an AI. And in combat, controlling a plane includes the capability to deploy countermeasures and take evasive action to avoid getting shot down, and defend itself to stop an attack.
I doubt the purpose of this was to create a great dog fighter... but rather it was to prove it's minimum capabilities work so they can continue developing it to operate in more dangerous/common situations... like taking out ground assets in the initial wave; or adding it to existing drones so they can take some defensive measures automatically; etc.
I mean an interesting question and to your point is: how have wars been won in the past decades - or not been won.
To me it seems that almost all wars of the past decades transformed into: “irrespective of your weapon superiority - it will all cumulate into asymmetric partisan wars that drag on for years and cannot be won”.
In the end it almost feels like “whoever drops more ordinances on the civil population to irrevocably break their morale may come out as a winner a bit more”.
Having read some some history, the obsolescence of weapons systems is indeed a cultural problem that people become unprepared for in the euphoria of advancements. Classic example being Syracuse vs Rome, where Rome was held back by the weapons of Archimedes only to be sacked after the Syracusans decided to have a party and left a gate open.
Are manned missile trucks even valuable for countering electronic attack? I'm thinking "no". A missile truck can definitely be AI'ed or controlled remotely. Perhaps only CAS can't be made unmanned.
Cost is a huge problem. We now have unmanned carrier launched refueling trucks that I bet are incredibly cheap. Make them a tad bigger, put missiles in them, and you have a super cheap replacement for the F-35. The F-35 can stick around in small numbers for when you really need a human, if you really need a human in the cockpit. Cost has to drive us to this.
You can conceptualize all you want about dog fights, the truth is nobody wants to be the pilot having fired all the missiles and having nothing left to engage the remaining enemy planes.
It's a bit like saying a soldier doesn't need a bayonet / combat knife. 99.99999% of the time yes, but would you bet YOUR life on it?
No shortage of criticisms -- literally "JSF problems" into google/bing/duckduckgo will find you plenty.
Also keep in mind that these are billion-dollar programs, so expect to get a lot of FUD and propaganda from the government contractors building them. I remember riding the DC Metro and rolling my eyes when I saw ads for the F-22 at Metro Center and Dupont Circle.
Saw an interesting video on the military channel a couple of years back that pointed out something important about these super powerful, super fast, super expensive aircraft.
You'd be better off from a cost perspective running a swarm of hundreds of drone strapped with explosives, only one of them needs to get lucky to disable a multi-million dollar plane and the drones themselves can be pretty cheap.
> You'd be better off from a cost perspective running a swarm of hundreds of drone strapped with explosives, only one of them needs to get lucky to disable a multi-million dollar plane and the drones themselves can be pretty cheap.
Why do you think fighters carry a bunch of “drones strapped with explosives” (air-to-air missiles) for air-to-air combat?
The thing is, when you optimize those, they tend to be comparatively short range and have no loiter time, so it's nice to have something to carry them into an engagement area and, if you need to patrol or escort, it's nice for it to also be something with some loiter time.
So proponents of hundreds of cheap drones theory tend to forget that kinetic energy is 1/2mv^2. So Energy required to catch up with a fighter in increases with square of velocity and modern fighters can all supercruise. That energy has to be stored on the drone to be expended which increases mass which increases energy require8d (linearly) and so on. If you need to add loiter time and manoeuvrability, again you need more energy and so mass.
So just like the famous rocket equation, you'll find that the design space narrows down to drones that are no longer cheap or casually expendable - essentially supersonic capable cruise missiles with loiter times
that tend to cost millions of dollars a piece. Certainly cheaper than manned aircraft but not something that can be used in the hundreds casually per engagement.
The other option is of course what's done today - strapping a relatively cheaper air to air missile to an even more expensive but fast and survivable aircraft which brings it to the war zone and gives it kinetic energy at launch.
> Certainly cheaper than manned aircraft but not something that can be used in the hundreds casually per engagement.
That was perfectly visible in the recent Syrian Civil War. Once the Russians decided to actively step in the conflict sure that they sent a rocket or two directly from their bases in Southern Russia/Caucasus, but I suspect that was more for propaganda reasons.
Soon enough they set up an air-base for good in Northern Syria (only a few tens of kms away from rebel lines) and they also brought in a battleship (or whatever is called nowadays the ship from which you can send rockets) off the coast of Syria. I'm pretty sure that proved a lot more cost-effective for them.
Your tactics change too. Instead of building a drone that can keep up with a fighter, focus on area denial. Throw up tens of thousands of tiny drones into the atmosphere that each travel at say 50 mph (like a model aircraft), and train the AI to put them in the path of anything moving within that airspace. The goal isn't to hit the fighter, it's to make the fighter hit you. Think of a flock of Canada geese over an airport: they're tiny, they move slowly, but they're super unsafe to fly through. Now imagine that instead of trying to get out of the way, the geese try to get in the way - and if you get within 10 miles of one, you may have a Sidewinder on your tail.
Then when not engaged in air defense, these drones engage in CAS. Taking down fighters is a secondary mission - their real role is to dominate the battlefield, and part of that is by making it so unsafe to enter the airspace above the battlefield that it's futile suicide to try.
> You'd be better off from a cost perspective running a swarm of hundreds of drone strapped with explosives, only one of them needs to get lucky to disable a multi-million dollar plane and the drones themselves can be pretty cheap.
It appears to be hard to predict the effect of the introduction of even a single weapon/platform into a complex conflict which consists of many, qualitatively different weapons/platforms. One of my career-defining "Aha" moments is described on page 19 of https://www.iitsec.org/-/media/sites/iitsec/link-attachments... in the section entitled "Forward Area Air Defense (FAADS)."
While I do not believe I am at liberty to provide details (even lo so many years later), I was witness to the first use of (arguably) VR to prototype and introduce a new weapons platform into a combined arms battlefield simulation. The short version is that on Monday morning, despite all the deep thinking done by smart people about how this would all work out, none of us came close to predicting what turned out to be the net effects of the new system as seen by Friday. I was there in my capacity as a nerd simply to keep the blinking lights blinking (vs the capacity of a combat domain expert), but watching the whole thing unfold was a mind-blowing demonstration of the "Law of Unanticipated Consequences."
None of us are as smart as we think we are. We are no smarter than our adversaries. The world is more complex than either of us can know.
Any drone that can operate with reasonable endurance at the same altitude as those super powerful, super fast, super expensive aircraft will not be cheap. An MQ-9 is somewhat cheaper than a manned fighter, but not by orders of magnitude.
If you're going to use the drone in an A2A role then it needs a powerful radar for accurate targeting. Targeting from other platforms using data links is possible in theory, but in practice only works in a narrow set of ideal conditions. Modern radars and other sensors are themselves super expensive.
And an MQ-9 is incredibly slow, can't maneuver worth a damn, and requires clear communication to control it. I don't understand why this question keeps coming up every time drones are in a news article. Sure something like Loyal Wingman might pan out, but the idea that swarms of cheap drones are going to be able to target something that's stealthy, operating at 50K, and around Mach 2 is just a fantasy.
That's a good point. But you also have to take into account how many dogfights have been avoided because the enemy knew there was no point in starting one.
The article talked about that; fighter pilots are taught dogfighting skills mostly so that they can become comfortable doing extreme maneuvers while conserving energy. Basically, it makes you a better pilot in general even if you aren't actually likely to get into a dogfight in a real battle.
The A part of the F/A 18 means Attack. Usually Attack aircraft are capable of close air support, which is a huge and integral part of our military tactics.
Without close air support, troops on the ground can't overwhelm enemies with firepower. Without air superiority, we don't provide close air support. Our strategy dictates air superiority to enable our tactics of close air support.
The reason there's only been one dogfight in the past 20 years is not because these fighters aren't effective, or because they're a relic of the past. It's that they're too effective. We own the sky and nobody is capable of challenging us.
The funny thing is that for a long time the opposite applied to fighter planes post WW2 - a more advanced one could take down entire airwings of older planes as the tech dominated so well and spam wasn't the answer.
Of course drones lack the biological limits on G-Force and the economic scales and factors like pilot deaths mean the situations are not identical.
The real question to be answered is probably one of "expendible drones vs premium super drones" long term as the age old battle of quality vs quantity.
You're designing a mini jet at that point, one that has to go faster than it's target yet be maneuverable enough to hit it. Which is pretty much just a smart missile at that point.
We learned in WWII that flak works just fine by exploding close to aircraft, which is much easier than directly hitting them. The same applies to drones.
You can have a wall of evenly spaced drones with explosives. When an aircraft passes all adjacent drones explode, the shockwave damaging the aircraft.
Are drones anywhere near as fast as a missile? In certain situations, an F/A-18 under afterburner can outrun a SAM missile - outrun meaning go fast enough that the missile can't close the distance before it runs out of fuel. How much chance would a drone have of catching it even while not operating under afterburner?
The unclassified ceiling is 50,000 feet, with a 250 m/s climb rate. Could a drone strapped with explosives come anywhere near that level of performance?
Programming the drones would be a pain. The wikileak ROE from 2008 was 27 pages long whereas Alexa struggles to converse on the topic of two plus two. Maybe some century we can replace the entire policy generating middle management of the Army with an AI; not any time soon it seems.
That discussion about swarms always devolves to "just push human beings near the action before launching the pitbull outta control killer drones" which for the last 30 years has been repeatedly successfully been implemented as modern jet fighters carrying short range sidewinder IR guided missiles, which annoys the people who thought they invented the new idea of explosive drones, little did they know we've been doing that for generations and its unfortunately probably even more expensive than swarms of WWII era P-51s with human pilots.
The really cool thing, especially for a ship, would be a dome of drones. They could come back to the ship to recharge and fly out again, and even, depending on weather, travel with the ship. Each one could sense incoming ordinance and move to pre-explode it well before it hit the ship. Of course, this wouldn't work against some things, like heavy kinetic stuff, or really fast stuff. But ordinary missiles? Other drones? Sure.
Why use drones in this missle defense system your describing? Drones are incredibly slow and missiles are INCREDIBLY fast. If you have the trajectory of the missile enough to put some sort of drone or hard object on it, why not just shoot it. That's what the phalanx missile defense system does.
Interesting thought experiment, a single Arleigh Burke destroyer has 96 vertical launch cells and can trivially be arranged to fire to a time and incoming bearing on a target.
So if an enemy ship, perhaps an aircraft carrier, really pissed off a single US destroyer, it can decide to send all hundred supersonic anti-ship missiles in from the same direction at the same instant of arrival, or from every point on the compass, or maybe three directions at once or who knows what kind of crazy four dimensional incoming attack pattern, like waves of ten missiles at a time every second for ten seconds from ten different directions. So how do you distribute your drones so as to soak up a hundred incoming from one direction while not getting sneak attacked by say five from every 30 degrees on the compass at the same time?
Its like say you have a Greek Phalanx, that's pretty well defended over a couple hundred feet, but try to use the same one phalanx spread out to defend all one thousand acres of city-state farmland simultaneously and its not going to work out well.
But you need to make sure that control radio signals are not jammed or that the onboard AI is intelligent enought that it will not target friendly or civilian targets.
I dont think thats a problem that can be solved today to the exclusion of manned aircrafts. If you are a terrorist it might not matter but for the kind of nation states that can field fighter jets it certainly matter.
This tactic only really works during peace time. The main strategy relies on the fact that the US (or whoever) isn't willing to shoot on sight. The swarm might well be able to get a good shot or two in but that would be the end of it.
Once a war is going on the you can move to a much more aggressive strategy. The US can just declare that any boats within a certain region are subject to summary attack and all shipping should steer clear. Attack aircraft can start sinking these small boats as soon as they leave harbour (or most likely cruise missiles blow them up at anchor).
I think this also is subject to the fact that today the US navy has not entered any area in force on a wartime footing. You can't just go around with your Phalanx armed if there might be anything non-hostile around you today (be it just an oil tanker of your foe - the PR by causing a massive oil spill is horrendous). A phalanx (or probably the 25mm cannons on the newest Arleigh-Burke destroyers) will quickly tear your swarm to pieces.
Fry: "I heard one time you single-handedly defeated a horde of rampaging somethings in the something something system"
Brannigan: "Killbots? A trifle. It was simply a matter of outsmarting them."
Fry: "Wow, I never would've thought of that."
Brannigan: "You see, killbots have a preset kill limit. Knowing their weakness, I sent wave after wave of my own men at them until they reached their limit and shut down. Kif, show them the medal I won."
>>the argument about the death of the dogfight, or that there is no need for within visual range engagements anymore is a tired one. There was a pretty popular movie in the ‘80s about that very argument, so I am not going to rehash it here.
Regardless of the actual merits of whether dogfights are a thing any more, I feel like "because Top Gun" is not exactly a sound argument.
CSBA has a report that goes through the data. It's rather one sided. Unless something like the vietnam rules of engagement concerning visual id happens again, dogfights are done. Add stealth to that and the arguments get monumentally stronger.
What's interesting to me - I thought I was a pretty big fan of Top Gun, but now I'm not so sure.
I did not pick up there is an argument about whether or not dog fighting still was needed.
The only "argument" I recall was the recap of how pilot deaths increased after training stopped, and decreased after training started. But it was a one off history lesson as part of the Top Gun school intro.
But the rest or the movie seemed very clearly - yes, the US needs dog fighting to defend its interests across the globe. There seemed to be absolutely no challenging of that idea within Top Gun.
Top Gun the Movie is actually pretty accurate when you look at the last 30 years of "no fly zone" enforcement, which was the mission in the movie.
Many dozens of times over multiple decades in multiple theaters against multiple opponents, the vast majority of stereotypical NFZ engagements have ended in "two american planes primary and wingman have a brief dogfight against enemy aircraft to obtain ideal firing positions, involving visual identification, then end the fight with short range IR guided sidewinder missiles". There are, of course, occasional exceptions but the VAST majority of NFZ aerial combat stories look like this.
Its interesting that for 30 years NFZ ROE documents have been classified and are generally still classified and only rarely can you find an ROE illegally posted on wikileaks. HOWEVER the secret squirrel types seem to have ignored the internet and OSINT in general. So you can't legally find a piece of paper titled RoE describing how visual ID was necessary and BVR generally prohibited for all practical purposes, however, its trivial to read on wikipedia the stories of about a hundred NFZ engagements and then reverse engineer the classified ROE to match the large number of mostly identical individual engagements.
Or to summarize the above, the RoE for NFZ operations for the last 30 years remain classified, although the results of seemingly all air to air engagements have been publicly released, and traditional OSINT techniques applied to hundreds of "isolated" public anecdotes can easily reverse engineer the contents of those classified RoE documents, and the reverse engineered RoE documents pretty much mirror the fight scene at the end of the Top Gun movie; get close, dogfight into missile launch position while performing visual ID, end fight by launching short range IR missiles.
My main issue with it is that the movie is over 30 years old and the argument in the movie is based on data from a conflict that began 55 years ago.
Maybe I'm completely wrong (I'm not a pilot and have no military knowledge), but I feel like things have moved on since then in terms of both the technology and the types of wars fought.
Maybe someone more knowledgable than me can answer... when was the last time a US combat aircraft engaged in a real dogfight?
Drones can be, and have been jammed, or even captured [1]. You need now, and will continue to need for the foreseeable future pilots in planes. But teams of piloted fighters and drones? That will be the winning combination. And how can a single human command a flock of 20 aircraft? By giving short orders, and letting AI taking care of the rest. The ethics of killing will never be left to AI, but the technicalities of acquiring and prosecuting the target? You bet.
Ha. I know of a device with a very simple AI program that can tell if an enemy is standing on it. If so, the enemy is killed.
It’s called a mine.
They have been used in every land (and naval) war since they were invented.
>The ethics of killing will never be left to AI, but the technicalities of acquiring and prosecuting the target?
This is unfortunately completely wrong. Fully autonomous weapons have already deployed for years, and are currently in use on the North/South Korean border. Fully autonomous weapons systems, including drones that use AI to autonomously choose and destroy their targets have been under development for many years.
There's a reason experts are sounding the alarm - because the ethics of killing are being shifted to AI right now.
>Kallenborn, an expert in unmanned systems and WMD, describes one type of swarm that he calls an Armed, Fully-Autonomous Drone Swarm, or AFADS. Once unleashed an AFADS will locate, identify, and attack targets without human intervention
> The ethics of killing will never be left to AI, but the technicalities of acquiring and prosecuting the target?
Other have already answered this, but I want to chip in with my own experience working for a submarine and military boat manufacturer + talking to navy officer.
They already have system to acquire and try to distinguish friend of foe. The only reason why the system is not pushing the launch button by itself is because there is a lot of tactic involved in starting an engagement. Especially for submarines who might prefer to remain hidden than launch a torpedo which will uncover its position. Also, those system are not perfect, one navy officer who was on radar duty told me that he activated the alarm on his ship after seeing something on the radar that ended up being a flock of seagull.
If the system was able to manage tactic and be able to distinguish friend of foe reliably, they would give it the power to kill.
I hope you're right, but what makes you so sure? In a wartime scenario, if a country is short on skilled personnel why wouldn't they release AI combatants if they had the tech? Especially if they're facing a perceived aggressor or defending their territory.
Keep in mind that electronically jamming a modern fighter is effectively downing it, irrespective of the human inside. I have read that are beyond the ability of a human to fly directly, and absolutely require fly-by-wire to stay aloft.
As for taking control of an AI drone, you can eliminate that possibility by only allowing abort orders remotely.
Classical drones still require a remote human. This is a different league.
"electronically jamming a modern fighter" isn't going to do anything to its flight controls.
You're correct that flight computers are generally required to keep dynamically unstable aircraft flying -- but that's done at a much lower level than anything that's connected to a radio.
The US military already has drones that are flown remotely. It seems to be a small step to add a dogfighting AI to a plane and just have the remote pilot sic the AI at another plane. The remote pilot just has to reengage after the fighting is over. That’s assuming that they can add the sensors and AI to track the other plane.
As the author mentioned, the dogfighting AI seems to counter the real pilot’s moves effectively while maximizing the amount of energy maintained. It seems to have caused the real pilot to overshoot or undershoot while maintaining or gaining energy. The countering also included adjusting aim to hit the real pilot as the real pilot tried to maneuver away. It wasn’t just perfect information or efficient flying.
1. The AI had "total situational awareness" (perfect position, direction, velocity of opponent). Even given 360 cameras, radar, etc there is still uncertainty, errors, etc.
2. The human was an expert pilot, not a sim expert (see videos, he was going way too fast several times, not something he'd do in a real plane). A DCS master would likely have done better.
3. The AI wasn't trying to avoid head-ons.
With this and the Starcraft2 blink micro "win" I'm just going to these from now on.
Why should it? Presumably the AI pilot is backed up / remote / not precious - while the human pilot has an inherent value - isn't reckless disregard one benefit of automated systems?
You don't risk an expensive plane with head-ons - that's what missiles are for.
This was testing the AI. If real-world AI was going to be deployed in warfare shooting down manned planes, you'd make custom hardware that is smaller, faster and 10-100x cheaper than a f-35.
It’s not minded, it’s not any different in than alphago or alphastar or the OpenAI dota bots, it doesn’t understand anything outside of the context of the game it’s playing.
It doesn’t have a mind. On the “intelligence” spectrum it’s much closer to the guidance system in a heat seeking missile to than say even a dog, yet alone anything that we would define as having cognition.
The only thing that will address that is a broad international agreement with sanctions and political consequences for aggressor AIs that violate it, just like agreements that mostly keep horrifying biological weapons off the table and mostly kept nuclear weapons programs in check.
"mostly" still leaves a lot of room for bad actors, but these sorts of compromises have worked surprisingly well enough that the failures were notable. Political problems are only going to have political solutions, and almost by definition those are imperfect.
Phew. Thankfully since we decided not to invest in defensive AI our power rivals also decided not to invest heavily in offensive AI. I was afraid the world wasn’t a fair and moral place for a minute there!
Home security companies working in AI also tends not to end well in the movies, video games.
It amuses me to think that Aperture Science had a protocol for rogue AIs with the codename "Birthday Cake". GLaDOS did not understand why all of the humans kept talking about cake as they died. They really seem to like cake...
it's as inevitable as nuclear proliferation following WW2. you can slow it down, try and create a few key players and build around that, but it's always coming.
Despite all the Smedley Butler memes the US actually spends a pretty normal amount of money for a country with a large economy. Most people don't grasp how much the US economy dwarfs that of other nations.
Germany and Canada both spend 1.3% of their GDP’s on the military. United Kingdom and China are at 1.9%. However, the US is all the way up at 3.4%. A 30% cut would still leave the US well above the international average and due to it’s sheer economic might result in a clear military advantage. https://en.wikipedia.org/wiki/List_of_countries_by_military_...
Consider the US Air-force is unquestionably the #1 most powerful Air-force in the world with the US Navy comfortably holding the #2. That’s frankly expensive.
US spends so much on military because other countries essentially outsource their own military to the US. For example, the EU spends very little on military in large part because of a large US military presence in Europe.
You could just as well double the EU's relative share by halving the US military budget. The US would still spend more than China, Russia, and a bunch of other countries combined. Surely that's more than enough.
It's just that US politicians always frame it as if the total spending had.to remain constant.
Hopefully we can remove many or most of those bases in the future. No reason to be the world police or to let other counties outsource their military to American lives.
A lot of this has always confused me. In times of peace you should invest more in civilian research. In times of war or near war, more military. They complement one another too. Civilian research will always be used by military, from basic understanding of how people act, to lasers, to economic theory. One helps the other. The reverse direction is slower though because it is often held close to the vest for a bit of time.
> Civilian research will always be used by military
It goes both ways, and military applications create market pressures and timescales that wouldn't normally exist. Take for example, sustained fission, computing, or the semiconductor industry as a whole (which was subsidized by the government in its early years for aerospace applications, some civilian and some not).
But that said it's more about industrial capacity than research. We build planes and tanks not because we need them today, but because we might need the ability to build them tomorrow. Continued research on military applications prevents institutional knowledge from dying and tools from rotting.
Peace time is the best time to invest in military research because you'll be more prepared to secure an advantageous position when a conflict emerges. If you only invest during an in-progress conflict then what you're really doing is playing catch up.
Government science investment frequently comes from fear. ARPA was the US government's big "give money to research" program. It was started in response to Sputnik because we feared we were falling behind the Soviets.
Then in 1969-73, Senate majority leader Mansfield decided that this research might be being misused to fund science in ways that wasn't directly useful to fighting Soviets, so ARPA became DARPA (the D is for Defense!) and now researchers needed to explain why their science might help build missiles.
If war breaks out, it's too late. Especially nowadays, where technology is crazy complex and it takes absurd amounts of time, effort and coordination to get something like a new fighter plane all the way from the drawing board into production. And then trying to iron out all the bugs while fighting a shooting war? That might have worked out back in WW1 or 2, but modern wars would be over in a flash (comparatively). Either you have the technology and you win, or you don't and you lose. Our modern capabilities of blowing each other into oblivion are too effective to only start research when war breaks out (or is "near" to breaking out).
That said, DARPA has always been one of the biggest funders of research in the US, whether it's basic research or applied. Just think of the DARPA Grand Challenge (self-driving cars in an off-road setting). While, yes, ultimately their goal is military application, this cultivation of research into self-driving cars is something that will benefit everybody, not just the military.
The reason you're seeing this story is not because the Pentagon felt like you would enjoy some candor on a whiz-bang AI thing. No, they're way out in front of this issue. This is part of a prolonged PR campaign to prepare the public for reducing or eliminating human-in-the-loop limitations. A wet dream for systems guys since they were children reading sci-fi.
The Pentagon has been using robots to kill robots and other humans since before 9/11. This story has a long arc. We meat bags are the major limitation for machines of war and that limitation will be engineered out.
Readit News
A great book that rings more true than ever talks about the obsolescence of weapons systems. I believe it was Freedman[1] A key thesis is that as a predominant weapon system becomes more vulnerable, it becomes more expensive, as it’s burdened with more defensive measures to survive from cheaper more expendable systems. In World War 2 the canonical example was the battle ship and the aircraft. Now it’s the fighter jet and the missile (or drone) etc. Just look at the ridiculous state of the JSF.
Manned missile trucks are valuable for countering electronic attack, but not dogfighting. Still, practice dogfighting is one of the most fun things you can do, too bad it hasn’t really been applicable since Vietnam. This post from the other day reminded me how fun it is and seems like a great fit for VR.[2] Onward last Samurais! Just try not to spend all the taxpayers money on the way out.
[1] https://www.amazon.com/Future-War-History-Lawrence-Freedman-... [2] https://news.ycombinator.com/item?id=24255191
I can play games where real-world motion matches in-game motion (Beat Saber, PokerStars VR, Superhot VR, Zombie Training Simulator, Space Pirate Trainer, etc.), but any sort of driving or flying makes me sick. Even games that make me use the controller to walk long distances make me dizzy and nauseous.
Particularly Growling Sidewinder has a lot of great videos and I can't stop watching them.
But the ability to control the plane is the minimum capability of an AI. And in combat, controlling a plane includes the capability to deploy countermeasures and take evasive action to avoid getting shot down, and defend itself to stop an attack.
I doubt the purpose of this was to create a great dog fighter... but rather it was to prove it's minimum capabilities work so they can continue developing it to operate in more dangerous/common situations... like taking out ground assets in the initial wave; or adding it to existing drones so they can take some defensive measures automatically; etc.
And the analysis of this incident: https://www.youtube.com/watch?v=yy16Bv14I-c
To me it seems that almost all wars of the past decades transformed into: “irrespective of your weapon superiority - it will all cumulate into asymmetric partisan wars that drag on for years and cannot be won”.
In the end it almost feels like “whoever drops more ordinances on the civil population to irrevocably break their morale may come out as a winner a bit more”.
How can wars be realistically be won today?
Having read some some history, the obsolescence of weapons systems is indeed a cultural problem that people become unprepared for in the euphoria of advancements. Classic example being Syracuse vs Rome, where Rome was held back by the weapons of Archimedes only to be sacked after the Syracusans decided to have a party and left a gate open.
Cost is a huge problem. We now have unmanned carrier launched refueling trucks that I bet are incredibly cheap. Make them a tad bigger, put missiles in them, and you have a super cheap replacement for the F-35. The F-35 can stick around in small numbers for when you really need a human, if you really need a human in the cockpit. Cost has to drive us to this.
It's a bit like saying a soldier doesn't need a bayonet / combat knife. 99.99999% of the time yes, but would you bet YOUR life on it?
A bit tangential but I'm curious. Can you enlighten me, or point me to where can I read more about it?
Also keep in mind that these are billion-dollar programs, so expect to get a lot of FUD and propaganda from the government contractors building them. I remember riding the DC Metro and rolling my eyes when I saw ads for the F-22 at Metro Center and Dupont Circle.
You'd be better off from a cost perspective running a swarm of hundreds of drone strapped with explosives, only one of them needs to get lucky to disable a multi-million dollar plane and the drones themselves can be pretty cheap.
Why do you think fighters carry a bunch of “drones strapped with explosives” (air-to-air missiles) for air-to-air combat?
The thing is, when you optimize those, they tend to be comparatively short range and have no loiter time, so it's nice to have something to carry them into an engagement area and, if you need to patrol or escort, it's nice for it to also be something with some loiter time.
They don't typically carry _hundreds_, though. Carrying capacity for air-to-air missiles is pretty low. You get only a few shots.
So just like the famous rocket equation, you'll find that the design space narrows down to drones that are no longer cheap or casually expendable - essentially supersonic capable cruise missiles with loiter times that tend to cost millions of dollars a piece. Certainly cheaper than manned aircraft but not something that can be used in the hundreds casually per engagement.
The other option is of course what's done today - strapping a relatively cheaper air to air missile to an even more expensive but fast and survivable aircraft which brings it to the war zone and gives it kinetic energy at launch.
That was perfectly visible in the recent Syrian Civil War. Once the Russians decided to actively step in the conflict sure that they sent a rocket or two directly from their bases in Southern Russia/Caucasus, but I suspect that was more for propaganda reasons.
Soon enough they set up an air-base for good in Northern Syria (only a few tens of kms away from rebel lines) and they also brought in a battleship (or whatever is called nowadays the ship from which you can send rockets) off the coast of Syria. I'm pretty sure that proved a lot more cost-effective for them.
Then when not engaged in air defense, these drones engage in CAS. Taking down fighters is a secondary mission - their real role is to dominate the battlefield, and part of that is by making it so unsafe to enter the airspace above the battlefield that it's futile suicide to try.
Have you not just described an anti-air missile?
Yes, but a loitering anti-air missile.
Also, drones are far slower than a missile, and even far slower than the airplane they're trying to target.
While I do not believe I am at liberty to provide details (even lo so many years later), I was witness to the first use of (arguably) VR to prototype and introduce a new weapons platform into a combined arms battlefield simulation. The short version is that on Monday morning, despite all the deep thinking done by smart people about how this would all work out, none of us came close to predicting what turned out to be the net effects of the new system as seen by Friday. I was there in my capacity as a nerd simply to keep the blinking lights blinking (vs the capacity of a combat domain expert), but watching the whole thing unfold was a mind-blowing demonstration of the "Law of Unanticipated Consequences."
None of us are as smart as we think we are. We are no smarter than our adversaries. The world is more complex than either of us can know.
If you're going to use the drone in an A2A role then it needs a powerful radar for accurate targeting. Targeting from other platforms using data links is possible in theory, but in practice only works in a narrow set of ideal conditions. Modern radars and other sensors are themselves super expensive.
"There has been only one dogfight involving a U.S. aircraft in the last 20 years: in 2017, a U.S. Navy pilot shot down a Syrian fighter." [1]
So you have to wonder a bit how much time is worth investing in that aspect of these planes.
[1] https://www.wsj.com/articles/duels-among-the-clouds-11565878...
https://en.wikipedia.org/wiki/Close_air_support
Without close air support, troops on the ground can't overwhelm enemies with firepower. Without air superiority, we don't provide close air support. Our strategy dictates air superiority to enable our tactics of close air support.
The reason there's only been one dogfight in the past 20 years is not because these fighters aren't effective, or because they're a relic of the past. It's that they're too effective. We own the sky and nobody is capable of challenging us.
My takeaway from that data point is that the Syrian military might be very interested in such an investment.
Of course drones lack the biological limits on G-Force and the economic scales and factors like pilot deaths mean the situations are not identical.
The real question to be answered is probably one of "expendible drones vs premium super drones" long term as the age old battle of quality vs quantity.
You can have a wall of evenly spaced drones with explosives. When an aircraft passes all adjacent drones explode, the shockwave damaging the aircraft.
https://en.wikipedia.org/wiki/McDonnell_Douglas_F/A-18_Horne...
The unclassified ceiling is 50,000 feet, with a 250 m/s climb rate. Could a drone strapped with explosives come anywhere near that level of performance?
That discussion about swarms always devolves to "just push human beings near the action before launching the pitbull outta control killer drones" which for the last 30 years has been repeatedly successfully been implemented as modern jet fighters carrying short range sidewinder IR guided missiles, which annoys the people who thought they invented the new idea of explosive drones, little did they know we've been doing that for generations and its unfortunately probably even more expensive than swarms of WWII era P-51s with human pilots.
So if an enemy ship, perhaps an aircraft carrier, really pissed off a single US destroyer, it can decide to send all hundred supersonic anti-ship missiles in from the same direction at the same instant of arrival, or from every point on the compass, or maybe three directions at once or who knows what kind of crazy four dimensional incoming attack pattern, like waves of ten missiles at a time every second for ten seconds from ten different directions. So how do you distribute your drones so as to soak up a hundred incoming from one direction while not getting sneak attacked by say five from every 30 degrees on the compass at the same time?
Its like say you have a Greek Phalanx, that's pretty well defended over a couple hundred feet, but try to use the same one phalanx spread out to defend all one thousand acres of city-state farmland simultaneously and its not going to work out well.
I dont think thats a problem that can be solved today to the exclusion of manned aircrafts. If you are a terrorist it might not matter but for the kind of nation states that can field fighter jets it certainly matter.
https://gcaptain.com/can-u-s-navy-defeat-swarm-attacks/
Once a war is going on the you can move to a much more aggressive strategy. The US can just declare that any boats within a certain region are subject to summary attack and all shipping should steer clear. Attack aircraft can start sinking these small boats as soon as they leave harbour (or most likely cruise missiles blow them up at anchor).
Dead Comment
Deleted Comment
Brannigan: "Killbots? A trifle. It was simply a matter of outsmarting them."
Fry: "Wow, I never would've thought of that."
Brannigan: "You see, killbots have a preset kill limit. Knowing their weakness, I sent wave after wave of my own men at them until they reached their limit and shut down. Kif, show them the medal I won."
Just fly in at a higher altitude and if there's some there go around them.
Drones carrying air to air missiles might be more challenging.
Regardless of the actual merits of whether dogfights are a thing any more, I feel like "because Top Gun" is not exactly a sound argument.
https://csbaonline.org/uploads/documents/Air-to-Air-Report-....
I did not pick up there is an argument about whether or not dog fighting still was needed.
The only "argument" I recall was the recap of how pilot deaths increased after training stopped, and decreased after training started. But it was a one off history lesson as part of the Top Gun school intro.
But the rest or the movie seemed very clearly - yes, the US needs dog fighting to defend its interests across the globe. There seemed to be absolutely no challenging of that idea within Top Gun.
Many dozens of times over multiple decades in multiple theaters against multiple opponents, the vast majority of stereotypical NFZ engagements have ended in "two american planes primary and wingman have a brief dogfight against enemy aircraft to obtain ideal firing positions, involving visual identification, then end the fight with short range IR guided sidewinder missiles". There are, of course, occasional exceptions but the VAST majority of NFZ aerial combat stories look like this.
Its interesting that for 30 years NFZ ROE documents have been classified and are generally still classified and only rarely can you find an ROE illegally posted on wikileaks. HOWEVER the secret squirrel types seem to have ignored the internet and OSINT in general. So you can't legally find a piece of paper titled RoE describing how visual ID was necessary and BVR generally prohibited for all practical purposes, however, its trivial to read on wikipedia the stories of about a hundred NFZ engagements and then reverse engineer the classified ROE to match the large number of mostly identical individual engagements.
Or to summarize the above, the RoE for NFZ operations for the last 30 years remain classified, although the results of seemingly all air to air engagements have been publicly released, and traditional OSINT techniques applied to hundreds of "isolated" public anecdotes can easily reverse engineer the contents of those classified RoE documents, and the reverse engineered RoE documents pretty much mirror the fight scene at the end of the Top Gun movie; get close, dogfight into missile launch position while performing visual ID, end fight by launching short range IR missiles.
Maybe I'm completely wrong (I'm not a pilot and have no military knowledge), but I feel like things have moved on since then in terms of both the technology and the types of wars fought.
Maybe someone more knowledgable than me can answer... when was the last time a US combat aircraft engaged in a real dogfight?
[1] https://en.wikipedia.org/wiki/Iran%E2%80%93U.S._RQ-170_incid...
Ha. I know of a device with a very simple AI program that can tell if an enemy is standing on it. If so, the enemy is killed. It’s called a mine. They have been used in every land (and naval) war since they were invented.
https://www.thedrive.com/the-war-zone/25235/the-u-s-is-getti...
https://en.m.wikipedia.org/wiki/Ottawa_Treaty
So really what the OP means is the ethics of what to kill and what to let live should not be left to an AI.
Deleted Comment
This is unfortunately completely wrong. Fully autonomous weapons have already deployed for years, and are currently in use on the North/South Korean border. Fully autonomous weapons systems, including drones that use AI to autonomously choose and destroy their targets have been under development for many years.
https://www.nato.int/docu/review/articles/2017/07/28/autonom...
There's a reason experts are sounding the alarm - because the ethics of killing are being shifted to AI right now.
>Kallenborn, an expert in unmanned systems and WMD, describes one type of swarm that he calls an Armed, Fully-Autonomous Drone Swarm, or AFADS. Once unleashed an AFADS will locate, identify, and attack targets without human intervention
https://mwi.usma.edu/swarms-mass-destruction-case-declaring-...
In this case the AI is onboard, right? So much lower surface area for being hacked.
Other have already answered this, but I want to chip in with my own experience working for a submarine and military boat manufacturer + talking to navy officer.
They already have system to acquire and try to distinguish friend of foe. The only reason why the system is not pushing the launch button by itself is because there is a lot of tactic involved in starting an engagement. Especially for submarines who might prefer to remain hidden than launch a torpedo which will uncover its position. Also, those system are not perfect, one navy officer who was on radar duty told me that he activated the alarm on his ship after seeing something on the radar that ended up being a flock of seagull.
If the system was able to manage tactic and be able to distinguish friend of foe reliably, they would give it the power to kill.
I hope you're right, but what makes you so sure? In a wartime scenario, if a country is short on skilled personnel why wouldn't they release AI combatants if they had the tech? Especially if they're facing a perceived aggressor or defending their territory.
As for taking control of an AI drone, you can eliminate that possibility by only allowing abort orders remotely.
Classical drones still require a remote human. This is a different league.
You're correct that flight computers are generally required to keep dynamically unstable aircraft flying -- but that's done at a much lower level than anything that's connected to a radio.
As the author mentioned, the dogfighting AI seems to counter the real pilot’s moves effectively while maximizing the amount of energy maintained. It seems to have caused the real pilot to overshoot or undershoot while maintaining or gaining energy. The countering also included adjusting aim to hit the real pilot as the real pilot tried to maneuver away. It wasn’t just perfect information or efficient flying.
https://www.popularmechanics.com/military/aviation/a29847417...
https://www.boeing.com/defense/support/qf-16/index.page
1. The AI had "total situational awareness" (perfect position, direction, velocity of opponent). Even given 360 cameras, radar, etc there is still uncertainty, errors, etc.
2. The human was an expert pilot, not a sim expert (see videos, he was going way too fast several times, not something he'd do in a real plane). A DCS master would likely have done better.
3. The AI wasn't trying to avoid head-ons.
With this and the Starcraft2 blink micro "win" I'm just going to these from now on.
Why should it? Presumably the AI pilot is backed up / remote / not precious - while the human pilot has an inherent value - isn't reckless disregard one benefit of automated systems?
This was testing the AI. If real-world AI was going to be deployed in warfare shooting down manned planes, you'd make custom hardware that is smaller, faster and 10-100x cheaper than a f-35.
For some reason, we can’t justify more on civilian research but the military easily can, if it embraces it as part of its future.
"mostly" still leaves a lot of room for bad actors, but these sorts of compromises have worked surprisingly well enough that the failures were notable. Political problems are only going to have political solutions, and almost by definition those are imperfect.
- If you don't, then the enemy will
- If you do, then you're inviting the enemy to do the same
So, without trust and mutual agreement on an international level, this is a self-propelling premise that cannot be stopped.
Game theory 101.
It amuses me to think that Aperture Science had a protocol for rogue AIs with the codename "Birthday Cake". GLaDOS did not understand why all of the humans kept talking about cake as they died. They really seem to like cake...
Consider the US Air-force is unquestionably the #1 most powerful Air-force in the world with the US Navy comfortably holding the #2. That’s frankly expensive.
You could just as well double the EU's relative share by halving the US military budget. The US would still spend more than China, Russia, and a bunch of other countries combined. Surely that's more than enough.
It's just that US politicians always frame it as if the total spending had.to remain constant.
Navy said, great, let's turn it into a tanker!
Northrop said nope but Boeing said they'd be happy to. So billions as long as those robots stay in their lane.
Deleted Comment
It goes both ways, and military applications create market pressures and timescales that wouldn't normally exist. Take for example, sustained fission, computing, or the semiconductor industry as a whole (which was subsidized by the government in its early years for aerospace applications, some civilian and some not).
But that said it's more about industrial capacity than research. We build planes and tanks not because we need them today, but because we might need the ability to build them tomorrow. Continued research on military applications prevents institutional knowledge from dying and tools from rotting.
Then in 1969-73, Senate majority leader Mansfield decided that this research might be being misused to fund science in ways that wasn't directly useful to fighting Soviets, so ARPA became DARPA (the D is for Defense!) and now researchers needed to explain why their science might help build missiles.
That said, DARPA has always been one of the biggest funders of research in the US, whether it's basic research or applied. Just think of the DARPA Grand Challenge (self-driving cars in an off-road setting). While, yes, ultimately their goal is military application, this cultivation of research into self-driving cars is something that will benefit everybody, not just the military.
The Pentagon has been using robots to kill robots and other humans since before 9/11. This story has a long arc. We meat bags are the major limitation for machines of war and that limitation will be engineered out.