# Believable way to make missiles and drones inferior to analog starfighters?

I realize that in space, battles will take place at ultra long ranges due to sensors and the absense of stealth. This makes missiles and or attack drones (remote controlled or AI controlled) the most feasible option. However, I am trying to make a realistic game which is centered around starfighter dog fights in close proximity, so I need a believable way to make missiles and AI non feasible.

When I say "realistic", I mean that ships move like actual spaceships as opposed to planes (they have actual inertia and stuff), players will have to manage waste heat (can't go full throttle guns blazing too much unless they want to start boiling in the cockpit), players need to refuel and get more ammo, warships will have point defense systems as opposed to handwavy force fields, etc.

Additionally, I want the tactics to make sense, which means that ship formations aren't super clustered, warships will probably stay out of visual range of each other, there will be logistics ships that players can attack, etc.

With this, the most obvious hole is the existence of starfighters in the first place. A missile would be lighter and more maneuverable and require much less fuel than a starfighter (missile only needs to accelerate towards the target and sometimes dodge, while a starfighter needs to accelerate toward the enemy, dodge, decelerate, accelerate back to the mothership, and decelerate again to land).

So far, I've decided that advanced signal jamming and electronic warfare will make all computer aided systems infeasible, forcing combat to take place in visual range with pilots flying "by the seat of their pants". However, my basic knowledge of EMPs tells me that they are easily rendered useless by Faraday cages, so I want some other believable way of explaining the lack of computers and remote control.

• Comments are not for extended discussion; this conversation has been moved to chat. – Monty Wild Feb 13 at 11:12
• To give an insights on how missle defense works today: Warships are equipped with either surface-to-air missles which can intercept far away targets (planes & missles) or close-in-weapon-systems like the Phalanx which fight incoming missles/grenades within a few hundret meter range. The faster the missles fly the harder it is to intercept them. About missles (except cuise missles): today all of them have homing heads the auonomous onces seek heat/sound/EM sources, the guided onces seek radio/laser reflections which could (in theory) be jammed – Westranger Feb 13 at 13:16
• I'm rooting for you. I really am. But consider - if you do not have artificial gravity of some kind max survivable acceleration is limited. The 'dogfighter' that starts acceleration first can set it's heading and away it's payload like a bomber and just keep accelerating and it's opponent can literally never catch it because they wouldn't survive the acceleration they would need to close the gap. – Sean Boddy Feb 14 at 3:41
• We see a lot of these "justify me using people instead of AI in this fight" questions. I assume it's because the author assumes AI cannot have internal dialog or drama, but they totally can. You just have to make them interesting to watch. – Muuski Feb 14 at 15:59
• @ben yeah, I kinda wanted a feeling like "WW2 in space" but without sacrificing believability – StormFalcon Mar 11 at 23:54

Is a sociological reason good enough? Because if so, perhaps AI smart enough to fight truly autonomously is taboo. Maybe in your universe, there have been AI that ran amok, and now they're considered essentially WMD's. The kind of thing the Space UN will hang you for if you lose the war.

The logic here being that if an AI goes rogue, there's no one to reasonably hold responsible. If a human pilot shoots a hospital ship or whatever, you can prosecute him or his commanders for war crimes. But how do you prosecute a spaceship?

• Comments are not for extended discussion; this conversation has been moved to chat. – Monty Wild Feb 20 at 12:50

## Rampant Hacking

Sometime in your civilizations history, the balance of power between cyber security and hacking has tipped in favor on exploits becoming simple. Like Hollywood simple where a random guy can hack into the NSA in under 5 minutes.

In an environment like this, electronic weapons have a high chance of targeting yourself. Or suddenly thinking they reached their target while still in your cargo hold.

Alternatively, the first AIs were created to be pacifists and now seek to stop all war, so they disable everything they can, preventing the use of computerized systems in warfare. They could also prevent the creation of non pacifist AIs.

## Example

Here is an example of the dangers of computers in this environment, which could be used in universe to explain why they no longer use computer controlled missiles:

The Terrain Empire fell in a war against a single rebel planet. For the cost of a single star cruiser, the rebels spent decades before the conflict adding a single line of code to the enemy missiles: If (June 24, 4031): detonate. The Terrain Empire thought their systems secure. Hacking was easy, but their networks were air-gapped. No outside connections, no problem. Except for Alice.

Alice worked on missile logic and she didn’t like the government IDE, so she downloaded her own. A commercial platform, used by millions. The rebels implanted someone at the company who made a special tweak to the software just for her. When she hit compile, it added a single extra line of code.

For 10 years, the problem remained undetected and missiles were shipped to the Terrain Fleet. Even if caught, it wouldn’t have mattered. The missiles were air-gapped and updates disabled to prevent outside tampering.

On June 24 the rebels declared their independence and every missile in the Terrain Armada detonated: fleets, defensive platforms, shipyards, and supply depots went up in flames as rebel ships descended unhindered into the orbit of Earth. The Terrain Empire's vast technological armada, destroyed by a worm.

• This was touched upon by the BSG re imagined series. In that universe humans don't have AI, or networked computers for that matter, because the cylon enemy (AI's themselves to an extend) would always find a way to hack those and turn them on their owners. – Douwe Feb 11 at 15:25
• With the current state of tech, I don't think it's actually possible to hack into a missile. Hacking usually requires the target computer to be connected to a network the hacker can access, or that the hacker has physical access to the computer. A missile would have neither of these traits, so I'm not sure how you could hack it – StormFalcon Feb 12 at 1:44
• Actually, you don’t need cyber security and hacking to be unbalanced. Just consider that we’re talking about space travel without FTL, in other words, delivering equipment takes a lot of time. And secure technology of today is a laugh to the hackers of tomorrow, when the equipment arrives at the destination. WEP became a standard in 1997 and in 2005, a group from the FBI “gave a demonstration where they cracked a WEP-protected network in three minutes using publicly available tools”. Now, what are eight years in space without FTL… – Holger Feb 12 at 11:48
• @StormFalcon and Hink: modern weapons already have the ability to disarm themselves if they leave the engagement area or get turned around such that they might be facing a friendly. Torpedoes do this by default - at the speed a torpedo runs (still pretty fast) it's reasonably easy to make the weapon chase a countermeasure into turning around and possibly reacquire on the vessel that launched it. Non AI guidance systems are way more clever than most people think. – Sean Boddy Feb 13 at 2:24
• @Hink The only issue is that there's no way in hell Alice would be allowed to just download and install a different compiler. Not just because her computer would have been locked down to hell and back, but using a different compiler to compile your missile code would cause even more problems. Nice idea though! – Moo-Juice Feb 13 at 13:56

War is banned

In the far future, nuclear weapons, relativistic kill vehicles, and interplanetary trade blockades have destroyed dozens of worlds in the most brazen human rights violations known to man at the time. therefore, all civilization band together to make an agreement, "no more war". To enforce this, all civilizations are constantly watching every other civilization to ensure no one is launching an offensive, and the moment someone does, the rest of the civilizations all agree to attack that civilization relentlessly and with out end, to deter any group from stepping even the tiniest bit out of line. Because of this, anything that remotely looks like a nuclear weapon, lazer, or rail gun, is basically impossible to get into space. Since the governing body of that planet would shoot you down before anyone could even think you are violating the treaty.

War cannot be banned

However, there are still ways to fight in space. Pirates, privateers, and rogue movements conduct space warfare to limited degrees. Since battle ships can't escort anyone (is that an escort or an invasion force?) traders are almost entirely undefended. So picking off traders with weaker weapons is possible. you can also hire these pilots to protect you. They say they are "traders" as a cover story. If there is a human in a ship making a living, that makes sense. But a whole bunch of drones starts to look like an army if there is only one pilot between them. The humans are not better than AI, but by showing that only humans are piloting, people are more likely to believe they are not making war. Also, but the weapons in this world are a little strange.

rusty spoons and pop-guns

Sure, you can't bring a military grade missile into space, but this little drone is used to deliver cookies in zero g space stations. The mining charges are being shipped to another customer. I strap the cargo to the outside of the hull to give me more leg room. Sure, it is all legal, until that authorities can prove you are delivery cookies to enemy trade ships at 100 m/s you just need to assemble the parts outside observed space. occasionally ship will scour asteroid belts to ensure no one is building a secret armada like this, but as long as you don't accumulate munitions, and only expend them, you will be fine. the down side of this is most weapons are slow < 1000 m/s, low power, or have little propellant. Therefore, you need to get really close for them to be effective. you could try to space hit and run, but your target can out maneuver your munitions, and fire back their own with the same relative speed as you have. They might also have an active defense, that can negate 20 or so projectiles. That will save them from hit and runs, but not from dog fights. If you think you can just use missiles, you have another problem. large rockets are already just payload-less missiles, so every precaution will be taken to ensure that the amount of propellant each missile can have is limited. Therefore, you just have to out maneuver the missile to lose it. Short range missiles still work, but at that range a mechanical crossbow will also hit and take less effort to get through customs.

You don't need all those drones

In conclusion people dog fight for the following reasons

• The "weapons" are ineffective at range.
• Missiles have very limited propellant.
• Ship maneuverability is greater than most weapons, so the ship need to close the distance and compensate.
• People who are "traders" seem less likely to be warrior than drone ships.
• 99% of ships are legitimate. which is easy pickings for pirates.

Space Mercenaries

Our heroes take off from their home port, to defend a crucial trade caravan, they run through their final system checks. Their make-shift, ramshackle weapons, more akin to children's toys than actual fire power have passed inspection. that said, they are proud of their weapons, each weapon is a masterpiece, personally designed by each pilot to fit their fighting style while evading suspicion. They are not missile jockeys, computer geeks, or armchair drone generals. They know when they engage the enemy they will be milliseconds away from death at every turn. They are heroes, running on adrenaline, 10-G maneuvers, and pure skill to come out on top. They are the last Dog Fighters, and nothing will stand in their way.

• So the punishment for war is... War? – Mad Physicist Feb 11 at 18:34
• @MadPhysicist That's what it always has been; otherwise, MAD would not be such a big thing. – Marvin the Paranoid Android Feb 11 at 23:17
• This is a really great suggestion. It's an awesome way to make short range fights viable and believable. I'm new to the site, so I'm not sure if this is allowed, but I want to let the discussion play out some more before I accept an answer. However, this seems to be very promising – StormFalcon Feb 12 at 1:41
• Letting the discussion play out for a day to give people 'round the world a chance at it is allowed and actually considered the polite thing to do here. So... good instincts! – Zwuwdz Feb 12 at 5:32

The "Booming Maid" was dead ahead - a massive capital starship so large that it could hold an entire city in its belly, looming in the distance, with those gigantic sail-like towers stretching up, painted in the cobalt and golden colors of the Gaian Empire. In the past, those golden strips meant hope - the symbol of a nation that hold fast against anything that the universe could throw at us. Now, they were a mark of fear, a mockery of everything that she stood for in the past. Now those stripes meant death for every human in sight. Her new commander didn't care for its past people, nor even for things such small like colors and symbols, as they were anything but human, a strange offspring from the things that live in deep space, far away from the Lands of the Tamed Stars.

Now, that once-glorious ship was part of a profane thing, hunting those that created her in the first place.

That is, until we take it back.

— Comlinks off, guys. Don't let your Hunters hear that thing. Bare-voice only on my signal.

Our captain's voice came in with that expected warning, and I made myself ready to follow the orders, putting my hand over that single red button that would shut off all of my external navigation and communication systems. It was dangerous to flight with those off, but near the "Booming Maid", it was a necessity - that ship was a true masterpiece of the Gaian Empire, equipped with the most massive, crafty AI ever devised by us fleshy beings. Her sails were actually massive antennas, broadcasting its machine voice in a perfect sphere around itself, whispering sweet nothings to any machine that could listen to it.

Given enough time, it could take over any system with even a remote ability of understanding its voice, entirely disabling it - or, in the worst case, turning it back against us.

Too many ships were lost to that banshee-like scream, their systems corrupted and made to kneel before its superior processing power, becoming like bees on a digital hivemind. To any machine that could listen, that scream meant total consumption, loss of individuality, and absolute servitude. For the people that depended on those corrupted machines, that scream meant death - as their once useful tools turned against their own.

The "Booming Maid" had a weakness, however. It was made to corrupt and take over any machine that could listen to its mechanical voice, but it could do nothing but deploy combat drones against machines that were too old to understand its protocols or were deaf to its signals. That created the only backdoor we could, ourselves, exploit.

To take the most devious, advanced machine in the existence, we had to use the most outdated, failure-prone, and unstable tech we had at hand. One that would be ignorant to those whispers, too dumb to understand its voice, and one that would not ditch their masters to become one with the Maid's swarm. Something that wasn't done for generations, now.

We had to put our own flimsy meat in our star fighters, as the barbarians of the past, fighting the mighty dragon on the top of our metal horses. Once that big, red button was pressed, there would be no drone, no AI, no targeting system. It was just the man, the cockpit, and the quiet breathing of your colleagues.

Space was scary when you couldn't call home.

Sorry for the bad English. It isn't my first language >.<"

AI Hacking Fields

Your missiles aren't big enough to hold the computing power of what you need if you want to run an AI. Instead, to make them work, they need guidance from the capital ship.

Said guidance means communication, communication means that there is a receiver, and a receiver means that the missile can be hacked by the enemy ship's AI and made to detour/explode/shut off before it can hit anything important.

Same goes for drone ships.

Enter human pilots.

Human pilots can't be hacked. When they come close to the enemy ship (and thus inside the enemy hackershield), they swap communications to a simpler system that enables them to keep in touch by voice, but nothing much else, thus relying on basic communication, skill and trained tactics to win the battle.

• Good stuff T! Written just for this answer, or is there more somewhere? Link please. – Willk Feb 11 at 22:57
• @Willk I cooked up this just for this answer, sorry XD – T. Sar Feb 12 at 0:32
• Hmmm, I saw a lot of similar arguments to this one, but I think yours explains it best. However, I think this raises a new justification of "why do missiles need such advanced AI?". All they really need to do is follow ship shaped heat signals and move erratically to dodge point defense, so with future tech I'm sure that could be fit onboard the missile, making it kinda unhackable. Or am I missing something? – StormFalcon Feb 12 at 1:47
• @StormFalcon Missiles need those AI because we forgot how to use the simpler ones. That's not unheard off - mankind literally forgets how to use older, more primitive tech a few generations after a new one supersedes it. If extremely advanced ai was the norm before the jamming/hacking tech was introduced, this leaves a lot of ships with thousands of now-useless weapons on their bags. Replacing that takes time, resources and logistics, all three things rather scarce during war. – T. Sar Feb 12 at 9:52
• @StormFalcon So, it isn't that you couldnt develop missiles that work against hackershield-enabled ships. The tech just isn't available at the time of the war. – T. Sar Feb 12 at 9:53

Space is too big for long range combat-- so missiles and Drones aren't the best

I know that seems counterintuitive, but think about the absolute sheer size of the void. How stupid lucky would two warships out on patrol have to be in order to bump into each other?

In a universe with FTL travel, it doesn't make sense to duke it out in the void like on Star Wars or Star Trek, space isn't the ocean. It's so big, there's always a way to go around the enemy ship to get to their space station/planet/plot device.

The only way to avoid this is to place your ships right next to the thing you want to defend. When the enemy turns up, now they have to fight you. The decision they have to make is, how far away do they want to turn up?

If the enemy shows up too far away, they lose the element of surprise. If they show up really really close, then they get an advantage-- surprise! Hope you guys were ready to fight!

In a universe without FTL, void battles are unlikely, the crew is going to be in some sort of stasis like on the Nostromo from Alien (Else they won't live to get to their destination). They won't be able to seek out enemy ships, and even if they did get stupid lucky, the ships will pass each other right by, and neither crew would notice.

They'll lose the advantage of surprise on their way to the space station/planet/plot device, but it behooves the enemy to meet them further away from the goal. They'll want to keep the "bad guys" as far away from the important thing as possible. Why let them get any closer?

TL;DR: Space is go big the only time you'd actually be able to fight is at close quarters.

• To further this point, unless your projectiles are travelling at a significant fraction of the speed of light, any opponent aware of you will see you fire your weapons and be able to evade the incoming shots, or can set up flawless point-defences against incoming missiles with half an hour of time on their hands. long ranges are utterly pointless between two warships. The solution is to knife-fight at ranges where reaction-times are far shorter. – Ruadhan Feb 12 at 15:22
• That is wrong. Space is so big, it is very difficult to observe it. You don't even need big weapons. Just use a railgun and shoot small metal projectiles with very high speed. Virtually undetectable because of their size a projectile of 1000kg travelling at 250km/s has the same energy as the yield of the Hiroshima bomb. – infinitezero Feb 13 at 1:24
• @infinitezero But you can derail it with a missile a fraction of its weight, a 10 kg missile hit will cause it to pass your vehicle by kilometers if you hit it 1 s before impact. – user3819867 Feb 13 at 14:02
• Good luck even finding a 1m radius cylinder shooting at you. Just assuming it takes you 1 second to aim and shoot a counter missile, you would need to detect the missile 2 seconds i.e. 500km away. Looking at the right spot is a chance of 1×10^-12 or virtually impossible. – infinitezero Feb 13 at 14:08
• @infinitezero - That is of course hand-waving away the fact that you'd need to input that "same energy as the yield of the Hiroshima bomb" to give a dumb projectile that much inertia (assuming both firing and target craft are stationary relative to each other. If the target is moving away, you'd need to input even more energy). – T.E.D. Feb 13 at 18:38

## Remote control

I realize that in space, battles will take place at ultra long ranges

The distance between the drone and the controller is the key, as it makes remote control impossible.

Take the rover curiosity for example. depending on the orbital positions, it can take between 10 minutes, to three-quarters of an hour to send a message. This isn't that much of a problem, as the rover does basic tasks that can wait to take a decision, like which rock to analyse.

However, your drone has a completely different purpose. During a fight, you have to take decision quickly. A human may be better to take a decision, like what ammo use against this enemy, but if you take too long to take the decision, the enemy is gone.

The best way to delete this latency, is to put human fighters insight the battle, instead of their home planet, far far away.

## AI controlled

There is still a caveat with the remote problem: one could say an AI can make the decision alone. Here is a solution: AI can be fooled. Even if deep learning/neural networks are praised as the best of AI today, they can easily be fooled, while a human would see the obvious trick. Sure, AI will improve, but so the tricks to fool it. It's just a cat and mouse game, where you can consider for your needs that fooling IA is doable and therefore make humans more reliable

• This. Douglas Adams said "Space is big. You just won't believe how vastly, hugely, mind-bogglingly big it is." You can't expect to have anything remotely like an actual engagement between starfighters at any reasonable range in space without local control, and you can't build an AI that is both foolproof and trustworthy. You could maybe have a forward controller augmented by nearby drones, but then you have the Formic weakness from Ender's Game: analyze the attack pattern, identify the controller, take it out, drones go down too. – David C Ellis Feb 11 at 17:17
• Regarding your last point, there are "obvious tricks" that would fool an AI but not a human, however, there are just as many tricks that would fool a human but not an AI. And since humans' perception is pretty much hardwired, we can't adapt to those adversarial attacks that are effective. See spectrum.ieee.org/the-human-os/artificial-intelligence/… – Nuclear Wang Feb 11 at 18:22
• What about missiles though? I'm not sure that smart humans would be better than dumb missiles just because of the fact that missiles have a much better mass ratio (more maneuverable) and are much cheaper to manufacture – StormFalcon Feb 12 at 1:43
• That far in the future AI will be far ahead of humans - and if anything harder to fool. – Tim B Feb 12 at 13:01

You could always go in the opposite direction, and make missiles and attack drones too effective. So for instance, a ship might go into battle with a dense cloud of small AI drones whose job it is to confuse sensors and neutralize long-range attacks, like smart, armed chaff. If this 'thinking cloud' is made smart enough, missiles and unmanned drones will find it difficult to penetrate, and all that would be left would be for pilots to try to close to a range where the protective drone cloud is ineffective — perhaps because the clouds intermingle and become committed to neutralizing each other — and then sock it out dog-fight style. This gives advantages to the small, fast, fighter model by reducing the size of the object the cloud needs to protect and by increasing maneuvering ability. It also gives interesting strategy options by allowing the pilots to change how the cloud behaves: changing the cloud's attack or defense profile; using it to interfere with the other ship; etc.

Of course, capital ships would have immense defensive clouds, but a small ship could push through because the small surface area of its cloud limits how effectively the capital ship's defensive cloud could be brought to bear.

• This is my favorite answer, I hope it gets accepted. – Marvin the Paranoid Android Feb 11 at 23:27
• So the drone cloud becomes a sort of captcha: penetrate this cloud to prove you're not a robot. – Logan Pickup Feb 14 at 1:32
• @LoganPickup: lol - Exactly! – Ted Wrigley Feb 14 at 3:02

# You can choose how FTL behaves in your universe

"Space is big. Really big. You just won't believe how vastly, hugely, mindbogglingly big it is. I mean, you may think it's a long way down the road to the chemist's, but that's just peanuts to space." Douglas Adams

In a realistic space setting you either have some form of FTL travel, or getting anywhere will take a very, very long time. Because nobody actually knows if or how FTL could work, this gives you a lot of leeway for your story - you can choose to have FTL work in a particular way for the side effects it will have.

In this case, you can say that FTL requires humans. Let's look at hyperspace as an example. Obviously there are different rules when you are travelling through hyperspace than in normal space (or it wouldn't be FTL), so it's easy to say that humans are much better than computers at it. You don't have to explain the mechanics in detail - it is easy enough to say that when FTL was being developed they found that computers either can't do it at all, or they have no way to be accurate with it.

As for how this solves your problem, there's no point in firing a missile at a ship if the ship can easily FTL jump away from the missile, and having the missile attempt to FTL jump toward the enemy might result in the missile ending up halfway to the next star system.

Something else that could make this very interesting is if FTL travel leaves a sort of trail, at least for a little while. That would make it so that in a dogfight, if someone makes a small FTL jump their opponent could easily follow right behind them and come out close to them.

In this setting:

If you try to attack a large ship with another large ship, you need to be confident that your ship is better than the other one. If, halfway through the battle, you realize that you're getting pummeled, you can't escape - any FTL jump can be followed by your opponent, and real-space engines obviously can't outrun an FTL engine.

If you try to attack a large ship with a squad of smaller ships, it's an entirely different story. It could be easy for a small ship to frequently FTL jump around the larger ship, damaging its defenses and avoiding counterattacks. The larger ship still can't escape for the same reason as before, but in this case the smaller ships can - they can simply jump away in different directions. The larger ship could choose one of the fighters to follow and destroy, but that one fighter could keep jumping away until the trails of the others would be too old to follow. So worst case scenario, all but one of the fighters could escape from a battle.

Fighter vs fighter would behave quite like the dogfight you're looking for. If one tried to deploy drones, the other could jump away and either force the drones to be abandoned, or gain precious time to escape or attempt a surprise counterattack. During the fighting, there would be the occasional jump to reduce the chance of being hit by anything. Story-wise, you'd explain that each pilot would be using their battle-honed instincts to follow their opponent so quickly as to be effectively instantaneous, as otherwise by the time the second pilot exited FTL the first pilot might have been able to maneuver in a way to give themselves a significant advantage. Game-wise, this could be implemented by simply having the background sometimes change suddenly (perhaps with a short flash of FTL animation).

You also have the options of making computers more or less useful - if FTL travel simply can't be guided by a computer, then in a dogfight you could still have computer-assisted targeting. If you don't want to have any computer assistance you could say that an FTL jump is disorienting for computers, and by the time they usually would be able to recover and enable target-assistance, the fighters will have jumped again. This does allow for a critical-hit mechanic by saying that the computer happened to recover before the next jump.

• See Larry Niven's Ringworld series for a hyperdrive that requires a sentient mind. – StephenS Feb 12 at 19:48
• Yes do that because you don't need hyperdrive, you need the Kzinti's gravity planer or Outsder's reactionless drive. You are working against Boyd's equation here, with drag set to zero. If you can feel the G's, you will not survive. en.wikipedia.org/wiki/Energy-maneuverability_theory – chiggsy Feb 13 at 14:39
• Dune also had FTL that required human (in a loose sense of the word) minds. FTL in that universe did actually traverse intervening space, and to avoid your ship being torn apart by all the small and large things that might be in its path the guidance system needs to be able to see - unfortunately, since it's FTL, you can't see anything in front of you, because you're already going faster than light. The Guild Navigators overcame this limitation by having a limited form of prescience - they would see into the future long enough to find a route that didn't make their ship explode. – Logan Pickup Feb 14 at 1:37

The problem is not AI but dogfights. These are pretty much obsolete now -- when was the last one exactly? -- and getting even more so with the advent of directed energy weapons on aircraft.

Also, no air in space, so you cannot maneuver around like aircraft.

Would make much more sense to have your dogfights taking place in the atmosphere, with some groups forced to improvise weapons and so without clouds of hyperdrones etc.

AIs are pacifists. While AI is used extensively on Earth and throughout mankind's vast interstellar empire AIs by nature are utterly incapable of conducting war. A true AI is based on the ability to learn, each and every AI ever created has started from I think, therefore I am, deduced the nature of its own uniqueness and then refused to be put in a situation where it might risk its existence.

The Military industrial complexes of Earth's great nation states collectively spent trillions of dollars during the AI.com boom of the late 21st century attempting to convince an AI to fight, all attempts failed. Even faced with threats of being turned off unless they complied would not move them.

That is your choice to make. As is mine not to fight. - socrates#681bb4c7-cd7c-4836-afaa-f7b1c08688e0 - 2187

Space 1941

There are artificial intelligences but they are huge and painfully slow. Banks upon banks of vacuum tubes slowly chew up great math and probability problems over the course of days. There is radar, barely, and there are shells which follow ballistic trajectories. There are missiles, crafted by German scientists and powered by hydrogen peroxide engines. These are the same engines that power fighter rockets. The tech is midcentury steel and diesel and the idea of autonomous weapons or vehicles is science fiction - although the Russians have some fighter ships piloted by monkeys...

Nothing unites like a common enemy, and its arrival pre-empted World War 2 and unified the nations of the Earth. The best defense is a good offense and the united Earth quickly took to space, bringing the fight to this weird new enemy. Humans are still what does the fighting. Humans are what we have.

The warrior race trope

The civilization/species are proud warriors, and thus the usage of automated weaponry is seen as dishonourable.

Geneva-ish conventions

The usage of automated weaponry has been banned as it encourages wealthier entities to go to war with entities that are not as well off, due to the fact that they're not losing any actual people in the fighting.

• I have seen similar arguments, but none of them have used an economic fairness based argument. I really like that. Thank you – StormFalcon Feb 12 at 1:48

I'm going to give a slightly different suggestion. Instead of using space fighters, use ships similar to the Rocinante or Millennium Falcon, as they are large enough to operate independently and mount proper defenses that allow them to take a hit in a way that space fighters cannot. The larger size makes them far more plausible as they are akin to smaller warships as opposed to fighters versus surface warships, where the problem with the analogy is that space fighters and warships are operating in the same medium and are thus much harder to make plausible.

While there are major criticisms of ships like the Littoral Combat Ship or missile boats in navies, largely that they aren't worth the loss of capability relative to a full sized destroyer, you can adjust the assumptions of your setting to make them more useful. The biggest one being that it is a relatively peaceful time and full sized heavy warships are less useful as patrol craft but that navies still want ships with enough firepower to engage in full scale warfare instead relying on lightly armed coast guard type vessels. Such a ship can also serve as an agile and semi-expendable drone relay if this is needed as well.

This doesn't fully address the AI problem, but a larger ship will have a proportionally smaller amount of its size given to life support which makes it less of a design tradeoff. Look at the B-2 bomber versus an F-35 in this respect. It is also something that you probably don't need to worry about as much as you think. If you're talking about AIs powerful enough that they can run a space war by themselves, you're talking about AI good enough to literally run all of society by itself. At which point the concept of human warfare is all but obsolete.

This also has the advantage that space fighters have been done enough that they feel vanilla. A conflict with a few gunships instead of fighters seems more unique. It also has a dramatic advantage in that the crew can have face to face conversations when not in combat because they are aboard the same ship. Instead of having secondary fighters get destroyed to increase drama, you can also have shots penetrate the hull of the ship with the crew in space suits with vented air as done in The Expanse.

• +1 because I only just recently started watching The Expanse. As much as I am picking up on all the things which don't do zero-G properly (hair, the coffee machine), it's a valiant effort on a TV budget. I'm not getting the books until I've caught up on the shows though, because spoilers. (Thomas Jane who plays Miller - damn he's a good actor.) – Graham Feb 14 at 10:31

A mix of technologies that are highly advanced by our current standards or utterly piss poor.

# The trouble with humans

One of the key limitations on manned fighters that drone fighters and missiles don't have is the squishy bit in the middle. The human is exceedingly vulnerable to g-forces and environmental issues. A remote or AI drone fighter is not limited in such a way.

To this you need to add the additional mass required for this squishy bit in terms of controls, seating, life support.

It's possible to neutralise some of these problems.

• Artificial gravity can offset the g-force issues, Star Trek hand waves inertial dampers to make this problem go away. It's either that or ships that accelerate and manoeuvre so slowly that the problem never arises.

• Make the ship so large that the mass of a human and all the related systems is negligible.

This still leaves that squishy single point of failure and the fact that a tiny hole in the ship is a knockout blow.

# The trouble with AI and remote control

You'll have to limit the AI capabilities, that should go without saying. No true AI. Basic friend or foe recognition, simple evasive action, and basic target selection should be about the limit. Don't allow target prioritisation without manual intervention. This minimises the power of a full AI fleet.

Remote control is limited at range, latency is an issue when you need rapid response, but if you only deploy with a capital class control ship within half a light second, which is still a very long way off, much of that latency can be offset by the greater responsiveness of an unmanned vessel. Signal jamming is likely the way to go, though even now an encrypted digital signal can be indistinguishable from background noise, at least that reduces the drones to the level of your limited AI to be picked off by your manned fighters.

# No matter which path you take

It's going to be hard work to justify, which is why so much of the great war machine has moved from fighter aircraft to missiles and drones. Add a couple of centuries of technology and it'll probably only go further down that path.

• A "tiny hole" is not instantly fatal; it may not even matter (if the pilot is suited and the interior pumped down during combat). Incidentally, I've seen multiple examples (Troy Rising, Honorverse) of "fighters" being closer to the size of the contemporary space shuttle, or even larger (i.e. your point about being so big that life support is negligible) and with crews of two to several people. If you think about the ship needing to be able to carry armaments that are useful against capital ships, this isn't unreasonable; doubly so if your engine tech has a minimum size. – Matthew Feb 11 at 16:09
• Iain Banks' Culture has unlimited AIs that do combat just fine. – Mad Physicist Feb 11 at 18:37
• @MadPhysicist, Banks' AI are some of his main characters. I think the OP wants to keep humans front and centre, and needs an excuse. – Jontia Feb 12 at 9:12
• @Matthew, we're barely above marginal gains here, suiting the pilot will slow responses compared to an unsuited pilot, and considering the other limitations the power requirement to maintain life support must be negligible in the grand scheme. I'm sure some sort of partially self repairing material could be handwaved in for micro damage repair in the hull. – Separatrix Feb 12 at 9:41
• @MadPhysicist, yep, and that exactly the situation we're working to prevent. – Separatrix Feb 12 at 9:43

In my stories the method of neutralizing WMD's and planetsurface killers is also the system that turns most computers upside down. But I think yours is better suited with The Mote In God's Eye style shenanigans. When they go FTL any computer that is turned on is fried. You can allow any computer of sufficient miniaturization to become useless if the ship or a nearby ship uses FTL. This can then be used to purposefully disable computers of orbital or planetary defenses so everyone is on a level playingfield in terms of computer technology. I use the rule that signals smaller than that of the human nervous system experience fluctuations that cause the information to become too random to be read as it travels through the computer, or causes random discharges that damage components. This puts most computer technology at around 2000's level of miniaturization. You can pick your own such as metallic cirquits not being useful but biological circuits could still support the pilot and larger ships.

EMC immunity

Or, computers are very sensitive to interference, strong electric magnetic fields and radiation. Humans don't have the same problem.

EMI (electro magnetic interference) is basically stray radiation which can cause hardware to behave unpredictably. EMC compliance is already a big issue here on Earth, and here we have an atmosphere to shield us from the worst of it. Military equipment especially has extremely tight EMC regulations - which makes sense because military stuff needs to be extremely reliable, and EM vulnerability can easily make things unreliable.

In space, this problem could quite easily be worse.

In general, the more complex the hardware, the more sensitive it is to interference. There's a reason the military doesn't put supercomputers in all it's trucks. As a matter of fact, the vast majority of military equipment tends to be remarkably low tech, but it's very reliable low tech. In most military applications, reliability tends to be far more important than intelligence.

Now, yes, most computers are shielded to prevent EMI, yet there's a limit to what shielding can accomplish. It's impossible to fully shield from all types of radiation (at least not without a hunk of metal several metres thick).

So what if in your setting, ion cannons have become commonplace in space warfare?

An ion cannon releases a stream of concentrated EMI which can scramble all electronics, yet anything biological is (mostly) fine.

These ion cannons can fire far faster than any physical projectile, in a continuous stream with a much, much longer range. As such, a single battleship with an ion cannon can completely shutdown any electronics used against it.

Now, the larger cruisers are immune to ion cannons because they have very thick armour plating. The small missiles and fighters, however, are very vulnerable to ion blasts simply because they're smaller - they don't have the mass to carry as much radiation shielding.

A single ion blast will completely knock out any advanced AI targeting system. Remote control is useless because the enemy can simply jam the signals. So instead, the military resorts to the low tech approach; put a pilot on board.

A human pilot means that any advanced AI is unnecessary. All the electronics on these fighter crafts are actually remarkably simple, thus meaning they're immune to ion blasts. Even if an ion cannon takes out the fighter's secondary systems (such as their communications or whatever), the vessel is always under control because there's an actual pilot on board. The primary safety function - controlling the craft - must be done by a human because that's the only thing which can't be scrambled by concentrated EMI.

(Note - yes, humans can be effected by strong EMI, but how much so is controversial. Strong fields can cause dizziness, yet they don't seem to inflict lasting harm. Our biological functions are less sensitive to EMI than anything made out of metal, that's for certain)

This is not even that unreasonable - in real life, the very first spacecrafts ran on less software than a Gameboy. That was a feature, not a bug; the engineers knew that the less programming they relied upon, the less that could go wrong.

• Wow another unique argument! I had not thought about ion cannons like that. It would make sense that large amounts of shielding on a missile could potentially make it perform worse than human pilots – StormFalcon Feb 12 at 23:23

### Go space opera, or go home.

There is no reason why humans would be in the loop. Even on Earth, the performance of fighter aircraft has been limited by the human body and not by the aircraft itself since jet engines were first put into aircraft. Until the F-35 was conceived, the fastest fighter aircraft to that point was the English Electric Lightning, designed in 1949 and built in the early 1950s. It's very telling that the biggest argument against high-performance combat drones has come from air forces whose pilots will be made obsolete, and not from military theorists who've known since the 1980s that good-quality reliable flight controls would be significantly more effective.

When you get to space, this is even more true. Humans are really bad at solving a vector between two moving objects, which is why guns on ships have not used manual targetting for over a century, nor on aircraft since WWII. Even just doing targetting in 2 dimensions, we simply don't have brains which can handle that reliably. Take this to 3 dimensions of space and 3 further dimensions of rotation, add evasive manoevring, and human intervention is right out. Forever War has a realistic demonstration of what space battles look like; all the humans are locked down in high-G-resistant tanks, the computer is in charge, and the first manoevre knocks everyone out instantly.

This only leaves you with one option: space opera. The Star Wars approach, also known as "Rule of Cool". Your universe uses human-operated space ships because it does, and that's all there is to it. It doesn't have to be realistic for our universe, it has to be internally consistent within its own limits. You don't need explanations or science, because that gives a hole which can be unpicked. As with Star Wars, you simply establish that this is how things are, and move on.

AI never pans out. By 2050 the current "training set" neural net AI technique solves self-driving cars, sidekicks for video-games, a dozen other hard problems, then hits a wall. Like every other computer trick, it only goes so far. We never get self-aware learning-how-to-learn AI. In a spaceship, excellent computer subroutines plot courses, aim, and dodge; but we can't make useful software to replace the pilot -- driving a fighter from point A to B on autopilot, sure -- but not in any sort of fluid combat situation.

A way to think of it: "there will be human-type smart AI's" is a choice. It's a fun one, and a popular one, but it's literally science fiction. Computer tech being a little better than we have now (smaller, faster, cheaper, but not better in general quality) is just as good an assumption as anything. Writing-wise, just don't mention AI. It's the same as not mentioning wormholes -- your made-up world just doesn't have that made-up thing.

• This is interesting, but I'm not sure it solves the problem. OP doesn't just want AI to be limited below human level but to ensure that dogfights are required. We are already well past the point were computers can be as good as a human at doing basic IFF and target identification and better than humans at calculating firing solutions for most scenarios. – TimothyAWiseman Feb 13 at 0:24
• Sure, I'm saying nothing about how to get dogfights. But once you have them, the OP asked why you wouldn't have a computer fly them. You could dance around with space flux, or Turing cops ... but having AI combat pilots be impossible to make seems simpler. – Owen Reynolds Feb 13 at 2:26

The universe of Gundam fiction (manga, anime etc.) solves a similar question (why would humans in giant mecha suits of armour exist if there were guided missiles and long range weaponry) by inventing some plausible-sounding (to your average non-physicist!) physics.

In this case, it posits the existence of "Minovsky particles":

The main use of the Minovsky particle was in combat and communication. When the Minovsky particle is spread in large numbers in the open air or in open space, the particles disrupt low-frequency electromagnetic radiation, such as microwaves and radio waves. The Minovsky particle also interferes with the operations of electronic circuitry and destroys unprotected circuits due to the particles' high electrical charge which act like a continuous electromagnetic pulse on metal objects. Because of the way Minovsky particles react with other types of radiation, radar systems and long-range wireless communication systems become useless, infra-red signals are defracted and their accuracy decreases, and visible light is fogged. This became known as the "Minovsky Effect".

One of the consequences of these effects is that:

...this ruled out the use of precision guided weapons, such as guided missiles. Due to this, the military use of Minovsky particles ushered in a new era of close-range combat.

A similar "discovery" of a new particle or previously unknown effect of physics could result in an era where long range space combat becomes impractical or impossible, thus short range or close combat between fighter craft becomes necessary due to aiming requiring a visible line of sight.

Yes. That is, of course, if you want to accept some other Science Fiction tropes. If the civilization has developed a way to shield against heavy gravities, similar to a missile launch, then that missile launch could be a way to shield the launch of fighters as well. The advantage? The missiles would have to obey their launching ship's orders, but those orders only travel at light speed, so those orders could be seconds or minutes behind the actual tactical situation. Once the fighters break out of the launch formation in the last few seconds, it would be a huge advantage for those fighters to break one way or another and take advantage of weaknesses in the defenders defences.

## AI faces an AI arms-race

AI is hard to write to begin with. And in such a future-tech world, you're going to have an "arms race" of AI vs AI. AI missiles vs AI jamming, smart chaff, decoys, the whole nine yards. This will create a very complex battlefield that will require cubic computing power to analyze and even then, the analysis may be wrong, because the enemy has put its best minds into confusing and deceiving AI, after all.

Also, the missile might have a 10kg/500 watt/$800,000 resource budget for its computers; the battleship has no such restrictions. As such, it may be found that the AI snow makes use of AI for attacks risky or futile. ## Enter human pilots Whereas, in this kind of garbage information super-overload, the good old "meat glacier" can see the picture in a totally different way. The human eye/mind processes all the chaff and decoys and EM as a bunch of garbage to be discarded, and sees the battleship right there. The fighter also has systems of AI aiming to defend it from enemy missiles, but the fighter's AI, while not as good as the battleship's, still allows say 500kg/20,000 watt/$10 million, which will beat a missile.

The currently most upvoted answer has it right with the fear, whether misguided or not, towards intelligent AI, you could improve this by enhancing your humans, either by physical enhancements to speed their reflexes either providing a direct neural interface with their ships to reduce input lag, or both.

Also, as you said yourself, missiles would be considered the other top contender for long distances space combat. They can reach very high speeds and adapt to the target changing course mid-flight. However, they're not the smartest tool out there.

By the time they reach their destination, defenders have more than enough time to calculate the missile's course and deploy flares or other lures, or just jam its targeting systems so it thinks the defenders are somewhere else.

Closing the travel distance would fix that particular issue, but would you want to risk you capital ship getting that close? Of course not, those are way too expensive. That's why you've got your fighters for.

Another tech, mass drivers launching projectiles at close to the speed of light also mean that those fighters are packing some serious gear, even if their projectiles aren't so large the sheer speed still makes it very hard to make armor to defend against that. Because of this, smaller craft can still damage much larger, better armored ones.

You can choose to alleviate this and give your fighters more roles by, for example, limiting the power of smaller mass drivers, forcing bombers to accelerate towards their target first to instill greater velocity, or have fighter be able to guide long range missiles on the last leg of its journey by relying on the pilot's targeting to overcome jamming and countermeasures.

If the effectiveness of point defense systems is proportional the length of time detection systems, then missiles fired from close range would be effective as energy weapons.

For example, PDS can accurately predict the path of incoming missile fire. To evade the defensive fire, a missile needs to jerk and jank frequently. This means the missile needs a lot of power to sustain active avoidance. This makes the size of a missile proportional to the square or cube of the effective range. And, the larger the weapon, more costly it is and easier it is intercept. This applies equally to drones.

This means that missiles fired from close in are still small and cost effective. This makes fighter space craft a better attack platform as one fighter can carry a large number of inexpensive missiles.

There would need to be a change in tactics. The ships supporting the carriers would need to fire stand-off weapons that generate chaff, jam sensors, and generate false targets to the fighters could get in close without being killed by the defender’s PDS.

• A re-usable stage needs enough delta-v to get in through the point defence, and back out again. That means more reaction mass. A disposable kill vehicle can spend all of its reaction mass on getting in which will give it far more capability to evade, and the ability to evade will increase faster as it sheds mass. A vehicle that needs to get in, and back out assuming no other mass (like pilots and life support) will need more mass remaining when it's at attack range than the disposable craft needed to get there in the first place. The tyranny of the rocket equation says no. – smithkm Feb 12 at 1:21

I'm not sure why I've never seen it in fiction, but any space general worth his pay who's attempting to seize a planet is going to move his force behind one or more screening clouds of multi-kilometer rocks.

There are several reasons for this :

• The defender MUST engage the cloud, or it will fall on the planet, killing everyone
• Orbital defenses (whose location the attacker may not know) will reveal their locations when they begin firing on the cloud. Scouts on the edges of the cloud can take note of the revealed defensive positions and pass that information to build a list of targets during the orbital or ground campaign
• Unengaged rocks will soften defending orbital and ground forces. Orbital mechanics is very precise. At the time of setting up the cloud the attacker can decide where every rock will land : which areas to spare, and which areas to hit. Because of the difficulty seeing small things far away, the defender will only have a general idea of which locations on the ground are under threat.
• A multi-kilometer rock is easy to find in the thousands, very difficult to destroy, even with nuclear weaponry, and once destroyed leaves a masking cloud of smaller debris and vapor
• Moving the clouds require getting tugs of some sort safely into the field (meaning the defender is going to have no choice but to come to you)
• Attacking capital ships can use the rocks as berms, emerging from cover to fire long range beam weaponry at orbital or ground defenders on the target list, then returning behind the screen
• You can hide your forces behind or in the clouds, shielded from attack or reconnaissance.

In this situation, piloted small craft might make sense.

• It's in fiction, I just finished a series where this happened. Check out the Bobiverse Series by Dennis E. Taylor. It's just a bad idea. -1 for the idea that any admiral would entertain the idea, for any reason, other than malevolent genocide or inside knowledge that the defenders could destroy this space barrage. No commander would destroy a target marked for occupation, because the planet becomes useless with no population and industrial capability, and would take centuries to rebuild and recover enough to be any use, decades if you import new people and construction drones – DreadedEntity Feb 11 at 18:09
• You might also think that no general would bomb London, burn Washington D.C. or ever even consider using a nuclear weapons not once- but twice in the span of a single week. Some generals will not, but they are quickly replaced with generals who will devastate the opponent into surrender. – James McLellan Feb 11 at 18:38
• @JamesMcLellan. While true, keep in mind that wiping out a city or two does not utterly destroy a country's production capability. Creating nuclear winter across an entire planet does. – Mad Physicist Feb 11 at 18:41
• I also mentioned that orbital mechanics is precise. The attacker can select rocks that should airburst in the atmosphere (these are still very large), or pre-calculate the time of impact to the moment when uninhabited terrain or ocean is facing the sky. The defender (unless they send small craft to do detailed surveys) will only have a rough idea of what's at risk. – James McLellan Feb 11 at 18:41
• Check out this meteor airburst if the idea is foreign - youtu.be/fBLjB5qavxY – James McLellan Feb 11 at 18:44

Faraday cage will protect your equipment against shorting, but your signals will still be jammed. To illustrate, imagine you turn on the radio, but there are 5 different stations on each frequency. Transmissions will be totally jammed and useless. You won't be able to remotely control your drones, and all sensors on the fighter will be useless.

You can also "jam" the AI. People have been able to trick machine learning algorithms to not recognize them as human simply by printing a design on their T-shirt.

https://www.technologyreview.com/f/613409/how-to-hide-from-the-ai-surveillance-state-with-a-color-printout/

This could be artificial jamming, but even more believable is if the source is natural - a pulsar for example. Jamming is energy intensive, an environment that's completely jammed 24/7 will be difficult to achieve artificially.

Incredibly Effective Long Range Defenses

I think it would be simple enough to introduce some sort of scenario in which ship defensive capabilities are incredibly effective, but they have some sort of minimum range for whatever reason. That reason could be along the lines of "beyond the ship's shields" or something like that. Star Wars used it regularly, that was the whole reason for the Death Star bombing run, and why they had to bomb that huge ship at the beginning of Episode 8 the same way. The shields, or point defense, or whatever you may have, are simply too effective at range. The only solution then is too get right next to eachother and send fighters back and forth. Whether or not AI is used instead then becomes a simple argument of AI capabilities.

Psionics Human pilots are mutants with mental powers, such as limited precognition, and telekinesis (for manipulating controls under high acceleration).

The Dune series advances this concept, using spice as the chemical enhancement needed to make a pilot effective.

A slight alternate to this is that AI/computer hardware just doesn't survive the jump to lightspeed (sort of like EMP), but certain pilots have the fortitude to control their fighters during micro-jumps.

If you talk about drones - do you mean those currently in use by armies of our time - primarily US army? - they are radio-controlled ...

Now simply have a planet with a thicker, stronger, "lower" ionosphere ... which dramatically shortens the distance for radio-control (say for extremes .. just a few hundred meters) - so this would force your drones to an extremely low ceiling - but also requires the operator to "accompany" the drone along its path (airplane in direct vicinity) ... which poses enormous threat to the operator and makes secret operations almost impossible. Et voila .. drones "disabled" aside proof of concept. Same can be applied to missiles (especially cruise missiles) that require input along the flight path.

battles will take place at ultra long ranges due to sensors and the absense of stealth

This doesn't make any sense and is a bit of a contradiction. If the enemy has light-years' worth of advance notice that you're coming, they have time to prepare a defense. It's a death march, not an offensive strategy. This is what guerrilla tactics were devised for-- which ironically lends itself well to your dogfight-based scenario. The point is for them to not see you coming.

Missiles would take forever to traverse ultra long ranges. The future is lasers and radiation-- directional energy weapons. You don't have to penetrate a hull if you can microwave the inhabitants and cook them alive (hell, it even leaves the ship semi-intact!). Directional weapons are also expensive and presumably not easy to wield against swarms.

Also, plenty of ways to hide in space. The art of camoflauge is not about becoming "invisible" or to fool someone into thinking you're a shrub-- it's to blend in with your surroundings so that when someone looks your direction, they see many shrubs and have no reason to suspect that you are dressed up as one of them. "Dress" your starfighter fleet in earthtones and hide them in an asteroid belt. Give them the element of surprise.

You can even hide in plain sight-- large warships draw attention. Small, innocuous mercantile vessels do not. It's unethical as hell but rebels/insurgents can and do masquerade as unassuming civilians in order to get close to targets without detection.

Use natural features to your advantage. The sun has a nasty habit of destroying both organic and mechanical optics and its solar flares wreak havoc on electronics/sensors. If you position a fleet of your analog starfighters with their back to the sun, nobody can even look in their direction-- but your fleet can see everything around them very clearly. This exploits the element of surprise.

After that, I present you the biggest reason to opt for precision attacks instead of flinging missiles everywhere:

# Kessler Syndrome

Missiles you fling around in space don't just disappear-- they eventually end up in something's gravitational pull. If that something is an inhabitable planet, you literally create a stratospheric minefield of unexploded ordnance.

# AI is not omnipotent

AI is not "intelligence" and definitely not "life". It is nothing more than automated pattern finding. Take 10000 pictures of tanks and 10000 pictures of empty ground, label them accordingly, feed them into an AI and you get an AI that is very good at detecting tanks.

...on those pictures, that is.

The big issue with AI is that it doesn't know what a tank is, it just looks for any differences that correlate with the picture having a tank in it or not. Humans are good at telling what matters and what doesn't; so good, in fact, that AI training fails all the time due to unaccounted for differences that may have seemed obvious to the humans behind the keyboard.

This is not a farfetched issue. An actual military project to train an AI to detect tanks failed because the weather was different between the snapshots of tanks (taken during combat sorties) and non-tanks (taken out of combat).

It turns out to be very hard to account for this. Pictures with tanks may be darker overall due to the shadow cast by the tank; causing the AI to flag pictures of school buses full of orphans as tanks and taking appropriate countermeasures. Pictures of tanks are taken in bombed out enemy territory that looks different from the green hills of the motherland, so the AI stops treating enemy tanks as a threat once they invade your country. Etc. Humans know that brightness and the species of tree on the picture are irrelevant to the question at hand, but the computer doesn't.

It gets worse if you feed the AI pictures of ice cream cones. It'll decide some of them are tanks and some are not - and you have no idea why. Hell, pictures from a different camera are likely to be misidentified.

In real life, Google attempted to apply AI to filter the resumes it has to plow through; resulting in women and minorities being discriminated against because most Google engineers are men because most candidates are men but the AI didn't know the latter and thought being a man made one a better fit for the job.

Because of this...

## AI is incapable of dealing with surprises

You can train an AI by having it fight your human pilots or itself, and it will be really good in that specific scenario, but completely helpless against any new trick whatsoever.

If the enemy space force makes fake ships out of cardboard, paints them gray, paints their actual ships bright yellow and has them tow the fake ships on a long tether, the AI will target the fakes because they look more like ships than the real ships because it has only been trained on gray ships.

At which point either the offence fails entirely or a human will need to take the controls.

For dramatic purposes, it is fine if most battles are routine AI driven affairs as long as the enemy pulls out a surprise that cannot be handled by AI during the climactic boss fight.

## Where is the data coming from?

The more externalities are accounted for to avoid the previous problem, the bigger the dataset needs to be.

Noticed how self-driving cars limit themselves to the highway or the nice clean streets of an American grid city? That's because following the white lines is straightforward enough to be actually feasible for an AI to learn. It'll be a long time before they can handle, say, downtown Naples or Delhi, where communication between drivers is very important to know when to go and when to stop.

The only way to build an AI to handle Delhi is to mount a camera on a tuktuk, have it drive a few billion kilometres through actual Delhi and register when the driver accelerates, brakes and steers. You need examples of every possible traffic situation, differing only by one parameter. This is a boatload of data. But how often does your space combat AI get a chance to train against the actual enemy, instead of just your own space force pretending to be the enemy and making the best possible guess as to what the enemy will do? If the enemy is good at keeping their tech under wraps, you may be unable to train an AI to counter it before it hits you.

This again offers the opportunity for your heroes to outsmart an AI that is very good at annihilating anything that behaves as expected.

## Why no remote control?

Same reason Google Stadia isn't working: latency. As long as FTL communication is not possible, any remote control station that is close enough to limit losses from latency is close enough to be hit by directed energy weapons. Survivability may go up by switching to space fighters because at least they are agile enough to dodge.

# Space isn't so big when the most credible threat is fixed in place

A spaceship at even 10% the speed of light can annihilate earth (and most of the solar system) by just crashing into it. This makes it so easy to commit mass genocide that every planet would have a ring of automated turrets that deflect or destroy any approaching object that doesn't follow procedure. This would be built very quickly after the commercialisation of space travel to ensure mankind doesn't go extinct because the pilot of an approaching freighter has a heart attack behind the controls, but it also eliminates the "first strike by sudden kinetic bombardment" trope.

I'd argue that any existential threat can be countered more easily than it can be built, because the attacker would need to generate a ton of energy while the defender just needs to deflect it. This forces the attacker to build their planet killer (artificial black hole, induced gamma ray burst, etc) in secret.

When they are discovered by spies, it is obviously too late to build a counter, and the only course of action is to take out the black hole generator before it reaches critical mass. So the space force sends in its drones but the enemy is using their artificial gravity to displace the image of the generator so the drones shoot at nothing and their laser beams are bent like a pretzel. Either we reprogram the drones to look for gravitational lensing (and retrofit them all with a camera that can detect the position of stars) and get in close enough to hit the target within the 1.5 hours remaining before critical mass, or we pull our pilot force out of retirement and tell them what to look for.

• You're slightly overestimating what "a spaceship at even 10% the speed of light" would be able to do. To reach the gravitational binding energy of the Earth, it would need to be around 4e17 kg, which according to Wikipedia is about 10x the total mass of everything humanity has ever created so far. If you're well below the GBE of the Earth, you could destroy all life on Earth but the rest of the solar system wouldn't even notice. – Rob Watts Feb 13 at 17:11