Stick to ship combat and no fighters.

[archimage]

I would broadly agree with Praetori.
Whether Fighter or Missile will depend on how point defence technology advances (and on the decisions made in face of those advancements).

Super effective PD would mean fighters either have to be shielded to overcome the PD and firing lasers, and generally be overly expensive, or be suicide darts. Missiles would just have to be produced in such numbers to overwhelm the PD. This has its problems, but they are necessary as there is no other recourse (ignoring lasers for the time being).

As PD decreases in effectiveness the usefulness of fighters increases in simple terms of logistics. You might get 100 (even 1,000) missiles from 1 fighter/bomber, but if you have to store, transport and deploy those 10^x missiles you have a massive supply line which needs to be defended by even more missiles. Manpower still needs to be fed and watered, and the few hundred tonnes of food required to keep a ship spaceborne is a logistical issue BUT it stays relatively static. If you have your missile boats and fight a serious battle you suddenly need to replenish 10^x+3 missiles. Those missiles need to be manufactured, which means factories given over to them, which requires more missile defended logistic targets. Are we running a total war in peacetime? Are we leaving factories empty? Are we spending rampantly to keep missiles flowing just in case of a battle?
Then consider 'sensor ghosts', or hit and run tactics. If you can fool the enemy into launching their barrage then you have just created a dozen opportunities for yourself, even if they have another 10/0 barrages left.
Also, just gonna put it out there....Reloading? Are we putting everything on Alpha Strike (in which case is this missile boat just a one trick pony, destined to explode after firing its salvo) or does firing 10^x missiles happen every few seconds, with the attached logistic nightmare/mechanical system that is operating in battle field conditions and might just jam/leave a dozen paths to your magazine for a lucky shot (Looking at you Jutland).

Fighters, on the other hand might survive their initial encounters. They might just need to be restocked and sent back out.
It all depends on how effective the PD is.


As to whether Fighter/Bombers should be in the game...There is no reason not to include them. We aren't going to get a 'realistic' game where one multipurpose fighter takes down a Station someone spent 50 years building, or where 4 men, a dingy and a block of plastique take down a battleship.
But options are important. So many games give you 5 ship classes, which are just scaled up versions of each other, and call it a choice.
I don't want Stellaris marred by Missile Boats Today, Missile Boats Tomorrow, Missile Boats FOREVER. (Same with Laser boats, or Mass Driver boats)

Give it ebb and flow. Give one side an advantage in PD, then let the other side get ahead in Fighters/Missiles, and by jo give us events and flavour text that bear that out. Give an event for sudden tech up:
'Your new technology gives you an added edge on the battlefield +5%damage for 3 months.'
'Your technological advancements are proving problematic for old style ships, halving the benefit for 2 years for ships built before X'
'My Leader! The enemy has developed new technology that we cannot defend against. Our Admirals are sounding the retreat! A) STAND AND FIGHT! (+30% damage taken) B) Well...ok. (All units lose 20% Organisation).
(Obviously with better effects)

 

[Teleros]

If it has sensors and any form of computer making decisions based upon it ugly things can happen. People have hacked the bloody mid-term exams by writing SQL code on the paper formula headers (dumping the DB when read by machine, pretty resourceful) and human-written viruses can hide in a photo/picture. Imagine a malicious AI specialized in that type of warfare and finding a single gap in your security systems. Or spies working with sabotaging the AI production inserting malicious code to be used later. With no human controllers within a parsec or so the mayhem your very own AI ships could cause makes those fleshling-pilots look less of a bad option.
Fleshlings can always be disciplined through propaganda, loyalty or sheer fear (ebil alienz will eat us all or, defect and we kill your family etc) but they are individuals not swayed by code and self-reproducing on their own compared to AIs to be built into drones.

Hence why I said to separate the sensor systems from the AI running things. Imagine sending a virus via radio pulses towards a modern-day, human-operated radar: even if you successfully send your code to it, it won't affect the operator except insofar as it appears as background static or similar.

Now, inserting code outside of battle is another matter (eg the factory or w/e). I'm specifically talking about after you've built your AI-piloted starship or whatever.

Fiction is fiction and debating stuff based on technologies that doesn't even exist yet is just personal preference and bias and won't result in anything.

Wrong attitude IMHO. Start with what's realistic (which admittedly is usually boring and not fun), then tweak to get your desired outcome.

Battlestar Galatica style fighters the size of a sports car zooming about and dogfighting like it's WW2 might not be a realistic representation of the future when we today have BVR combat between modern fighters the size of buses but it's still FICTION. No one knows why they even have fighters in star-wars when most capital ships seem to fill every role the fighters do (except maybe implanting a torpedo down a ventilation shaft, which could be handled by a sufficiently clever drone) but it's part of the story.

BSG I'm not really sure about. Star Wars it's at least plausible given how their shields work (ie you can fly under them), and how limited their AIs and droids are.

A LOT of Sci-Fi works have fighter-style craft in them

Doesn't mean it's good sci-fi, just that it's popular .

Yeah, backdoor hacking in military is not only possible but already being used, think about how Iran's nuclear project messed up by Stuxnet.

Stuxnet if anything reinforces my point: the Iranian facilities in question were completely isolated from magic in-combat Cylon wireless hacking, so the people behind Stuxnet had to get in another way - in this case, via a USB stick or similar physically taken into the facilities.

Now apply that to a combat situation: getting a spy / saboteur on board the enemy ship to suborn their AI Stuxnet-style is fine & dandy (if difficult). Phoning them up and sending a virus that way is a sign that the guys on the receiving end are terminally stupid and have unrealistic computer security.

Getting sensor or communication hacked or jammed is fatal to any combat aircraft, manned or drone.

Hardly.

1. Think about how radar works on a fighter plane: the radar sends out radio waves, they bounce back, and appear on the monitor. Not much you can do to hack that.
2. The communications are another matter, in that you can in principle decrypt them and listen in on orders etc, or possibly send your own orders etc out (until someone catches on & implements whatever protocols exist precisely to avoid this). After all, there's a reason the British in WW2 just listened in on Enigma-coded messages - and even then they couldn't always act on that information because it would have given away the knowledge that they'd broken the code.
3. Jamming is a double-edged sword. You can, like the Iranians, focus a powerful radio transmitter on a particular aircraft or drone, and thus blanket the real signal (hence how they downed a US drone a year or two back)... but that's a very big signal for anti-radar missiles to home in on. Or you can blanket everything on a particular frequency (or frequencies), in which case you prevent your own people using those frequencies too.

Generalist fleets and specialist ships should be better than spam fleets of the same ship.

Indeed.

Personally, I'd like to see carriers used as patrol / security ships: hunting down pirates, policing the high orbitals and asteroid mining operations, that kind of thing. Very different from the modern conception of the carrier as the heart of any big offensive fleet mind you. Still, you wouldn't want to skimp on them.

 

[Teleros]

Apologies for the double post.

Whether Fighter or Missile will depend on how point defence technology advances

Space is not an ocean. The problem isn't whether your fighter can be seen (it will be, magic cloaking aside), or whether it can be shot down (it can be, magic cloaking aside), or even whether it can hurt a capital ship (debateable, depending on shield tech & such). The problem is that it's an inherently inefficient design.

Let's start over from the basics...

1. Mission: Deliver lots of BLAM to the target.

2. Limitations: Delta-v, delta-v, delta-v, life support, delta-v, delta-v, delt- oh you get the idea. Oh, for those who don't know, "delta-v" means "change in velocity" - there being no air in space to bank against like fighters can on Earth, all your changes in velocity have to come from firing rockets (whatever Star Wars would have you think ).

3. Optimum Design: An AI-piloted missile loaded with submunitions / multiple warheads.

Point #1 I think we can all agree as being the basic raison d'etre of a starfighter.

Now consider point #2. You need delta-v to accelerate to the action, delta-v to slow down, delta-v for any combat manoeuvres, delta-v to accelerate away from the action after expending your weapons, and possibly delta-v to slow down at your mothership's rendezvous. All this means you need to carry lots of propellant and lots of fuel / a big power supply. This affects the mass of the ship, which in turn affects the amount of delta-v required, and so on.

Second, your pilot needs life support. This requires mass: you have the pilot himself, his space suit, his acceleration couch, air supply, fighter controls, and so on. Again, this will affect the mass of your starfighter, affecting its delta-v again. Joy.

Which leads us to point #3. Your missile doesn't need to return to base, so you can cut out some of the delta-v requirements right there. Its AI core is almost certainly less massive than a human pilot etc, so you can cut out mass there too. Then, it can likely survive higher accelerations, making it more manoeuvrable in combat than a human pilot. Finally, it can cheerfully blow itself up against an enemy vessel when its ammunition is all gone. You can also opt to dump some of these advantages in order to gain others: for example you could balance out the life support mass by making up the difference with shielding, if that makes our missile survive longer in combat than a slight edge in delta-v.

+

Now, how would I justify fighters? I mean, I know I've just said how horribly unrealistic they are, but they are cool. So...

Option 1: Cultural reasons. Your species in question is willing to use inefficient means of warfare if it also allows them to place noble, duelling warriors (in SPAAAACE).

Option 2: Policing reasons: You don't use fighters in fleet actions, but you do use them as police vessels, eg for convoy protection, anti-piracy work, etc. Under such circumstances, you may prefer having a human being in the decision-making loop, rather than a missile AI. When that civilian liner is hijacked, being able to do a fly-by and such can be invaluable.

Option 3: Pure handwavium: Capital ship shields are a good distance from the hull, letting fighters fly underneath instead of having to try & use pop guns to penetrate said shields. Armour also sucks (no unobtainium hulls etc), so shields are the main defence against all threats. AIs have just never worked out in practice, so humans are still competitive / superior to them as pilots. Inertial dampening tech scales up badly, so fighters can be really manoeuvrable, unlike the big battleships etc. Cloaking tech only works on the smallest hulls, as cloaking tech isn't perfect and is too easily defeated by decent sensors.

+

So... as you can see, it is possible to justify having some fighters, but you have to work hard at justifying them. Sure, you can just say "I'm having space fighters JUST BECAUSE", but that's poor writing, and we don't want that.

Looking at you Jutland

Royal Navy battlecruisers exploding at Jutland had more to do with lax safety protocols than anything else. Those ships (even the battlecruisers) were very tough provided you didn't leave powerful explosives lying about like the Royal Navy did. Not the best decision ever made .

 

[Thrake]

Your AI that is capable of advanced decision making, independent action and complex strategy which can also resist the best hacking attempts that your opponents have is great. However one might argue that such an AI deserves its own rights and can refuse to be put in an expendable fighter ship. Does a true-AI piloting a ship actually count a being unmanned?

When the AI rebellion happens you certainly don't want to be the space empire which uses mostly AI-controlled ships.

You don't need something complex and sentient. Just something able not to crash into other ships, and able to target. Orders can be given by men and you can then imagine however you like that an admiral can give broad orders to a thousand drones or just say one man control ~20 drones each (destroy turret something,...). The rest is barely data processing.

Will they now? Do you even know we'll ever have AI? As for how advanced they are at a certain point in time and the strenghts and limitations to them, those are also unknown.

Some programs are being tested as a mean to provide answers that only an expert in one field would as of now (it scans vast amount of data on the Internet to create a database which allows it to answer complex questions). Also as mentionned, AI piloting is already present/tested; probably not perfect but I see no reason why it couldn't be improved; improving computing capabilities as time goes on will definately help too. Simply put, an AI piloting a ship is much more likely than just FTL. Eitherway, an AI can only be better at things like targeting than an human; its just a matter of reactivity and computing (physics) and an AI already does it better than any human could.

 

[viola]

Seriously? There is still people arguing for realism?
Do they even know what fiction is? Do they even know what Science Fiction is? They got no problem with FTL travel and tentacle aliens but Space Fighters is what they deem ridiculous?

Seriously, what's with Stellaris and attracting all the people who can't separate reality from fiction? If you can't tolerate to look at something that doesn't make sense in reality or is "implausible" somehow then this game is not for you. Please stop spamming the forums with these threads that completely miss the point of the game.

mod edit

 

[Kuschelflummi]

Seriously? There is still people arguing for realism?
Do they even know what fiction is? Do they even know what Science Fiction is? They got no problem with FTL travel and tentacle aliens but Space Fighters is what they deem ridiculous?

Seriously, what's with Stellaris and attracting all the stupid people who can't separate reality from fiction? If you can't tolerate to look at something that doesn't make sense in reality or is "implausible" somehow then this game is not for you. Please stop spamming the forums with these threads that completely miss the point of the game.

*snickers* You are going to be ignored or, much worse, will regret this post.

 

[TheDungen]

*snickers* You are going to be ignored or, much worse, will regret this post.

Well he comes to the right conclusion for the wrong reasons, of course willing suspension of disbelief is a thing but honestly when talking about tehcnology 500 year sinto the future we have about as much knowledge of what that will be as people 500 years in the past had about today.

 

[Praetori]

Hardly.

I'm pretty sure the quote was from Just a Joke and not me but nevermind.

1. Think about how radar works on a fighter plane: the radar sends out radio waves, they bounce back, and appear on the monitor. Not much you can do to hack that.

You'd be surprised.

Spoiler

Even though not hacking in the traditional sense it's actually what modern ECMs do. All active sensors like lidar or radar are dependent on the pulse repetition frequency for differentiating between the echoes and allow for calculation of multipathing or filter propagation issues.

To increase the PRF without losing range you adopt some smart tricks by modulating and encoding your pulses (it's basically data packets that enables us to figure out from which microwave element the specific pulse originated and when), it also helps us see through some noise jamming or other unfavourable conditions by actually knowing which pulse is ours due to the encoding (your cell-phone works along the same principle).

Modern ECMs however are an adaptation of this technology and ultra fast processing along with data collected through SIGINT (or even worse through spies) has enabled techniques in ECMs that emit modified returns stronger than the cross section of the target but with different properties in frequency (simulating a different doppler-shift and thus course), timing (throwing us off distance and angle) along with some other neat method to spoof a probing sensor. Some of them even mess up with the signal processing filters causing discarded calculations or CPU lag which can have some pretty nasty consequences in integrated shared solutions such as on some missiles which for good reasons share hardware resources for guidance and sensors (ICs).

Larger sophisticated sensors of course have methods of noticing foul play and often adapt and/or notify the controllers but putting the same type of technology and processing-power into a missile that will destruct on impact is both expensive and dangerous (as a single undetonated one will give your enemies top-notch intel on your advanced sensors.

Some sensors even have functionality to detect an incoming Anti-Radiation Missile and shut down transmissions to protect the sensor, by fooling the sensor system that that's about to happen you could theoretically make a sensor "blind" without even being nearby.

3. Jamming is a double-edged sword. You can, like the Iranians, focus a powerful radio transmitter on a particular aircraft or drone, and thus blanket the real signal (hence how they downed a US drone a year or two back)... but that's a very big signal for anti-radar missiles to home in on. Or you can blanket everything on a particular frequency (or frequencies), in which case you prevent your own people using those frequencies too.

True and jamming is effectively limited by the available power you carry along. Narrowband and beamformed is a high-tech power-efficient way of blanketing a transmitting sensor on particular frequencies you still want to utilize yourself. Wideband and omnidirectional requires a huge powersource and a very powerful jammer but it can be pretty crude although it, as you stated, also prevents you from utilizing the same band (as well as being big fat target for ARMs).

Hence why I said to separate the sensor systems from the AI running things. Imagine sending a virus via radio pulses towards a modern-day, human-operated radar: even if you successfully send your code to it, it won't affect the operator except insofar as it appears as background static or similar.

The AI must have some way to interface with its sensors though or it's blind (even if they're running on different hardware) and it's the buses or connections that create the vulnerability. As long as there is hardware or software at some point looking at the data there could be ways in (such as Stegosploits). I would never trust weaponry of any meaningful calibre to machines without human controllers. Not for the fear of rouge or malicious AI like the classic sci-fi trope but rather because it COULD enable system wide vulnerabilities for a hostile intelligence (human or other) to utilize and turn friendly weapons into foes.

Stuxnet if anything reinforces my point: the Iranian facilities in question were completely isolated from magic in-combat Cylon wireless hacking, so the people behind Stuxnet had to get in another way - in this case, via a USB stick or similar physically taken into the facilities.

Or space dust strewn out across a system that are really tiny sized micro satellites filled with malicious code. Like biological weapons of old. And AI supposedly working for you but without any means of you communicating with it is kinda pointless. We are now again at peace with empire X but since or AI's have a strict no-comms protocol we can't really tell them to stop.

 

[EntropyAvatar]

Now consider point #2. You need delta-v to accelerate to the action, delta-v to slow down, delta-v for any combat manoeuvres, delta-v to accelerate away from the action after expending your weapons, and possibly delta-v to slow down at your mothership's rendezvous. All this means you need to carry lots of propellant and lots of fuel / a big power supply. This affects the mass of the ship, which in turn affects the amount of delta-v required, and so on.

This is a strong point, assuming that ships in the Stellaris setting use reaction drives for in-system movement. A lot of settings feature some kind of "magic drive" for sublight, to go along with the magic FTL. It seems to me to be a choice whether, in the Stellaris setting, FTL technology is basically a one-trick-pony and we use currently-understood physics for everything else, or if some of the magic of FTL also leaks into sublight travel.

If you allow for some weird physics for sublight travel in the game setting, then a reusable delivery platform could make a lot of sense. Say that there is a minimum size on an efficient sublight drive, and that they are expensive relative to their fuel requirements. Now, to some extent this is an example of "working hard to justify fighters", but I don't think positing a sublight drive where the delta-V realities aren't so punishing is that much of a stretch in most SF settings. It's more like the default. If the capital ships have to operate by normal delta-V rules, combat would be very very different than most people are imagining, and kind of hard to play with ("Sir, we destroyed the enemy raider force, but you didn't leave enough reaction mass to cancel our out-system vector, so now we are bound for deep space").

 

[Kuschelflummi]

Well he comes to the right conclusion for the wrong reasons, of course willing suspension of disbelief is a thing but honestly when talking about tehcnology 500 year sinto the future we have about as much knowledge of what that will be as people 500 years in the past had about today.

Just because someone is right, its no guarantee, that everybody else sees it the same was. Especilly in the internet.

 

[Praetori]

Just because someone is right, its no guarantee, that everybody else sees it the same was. Especilly in the internet.

 

[Praetori]

Let's start over from the basics...

1. Mission: Deliver lots of BLAM to the target.

2. Limitations: Delta-v

3. Optimum Design: An AI-piloted missile loaded with submunitions / multiple warheads.

A well and structured post though it's as you say a matter of justification and making assumptions on AI capabilities while not doing it for engines is a bit strange.

1. Mission: Deliver lots of BLAM to the target:
Can be done in a myriad of ways but only smaller craft enables you to do so in many places at once at remote locations without exposing your precious fleet. The sensors and computing-power of a smaller craft is not blown up along with the warhead as in a missile.

2. Limitations: Delta-v:
Assuming that the engine and power technology favors larger engines and larger masses (there are plenty of technical effects today that work well with low masses but worse with larger ones).

3. Optimum Design: An AI-piloted missile
That assumes AI is viable, reliable, cheaper than fleshlings and less corruptible. The reason why not all soldiers today have a remote controlled tank isn't because they don't exist but because soldiers are cheaper.

A 50 tonne $1bil craft with advanced sensors, navigation and guidance that can be used multiple times to fire $100k "semi-smart" missiles or a 20 tonne $10bil missile that can only be used once. If you don't care about economy the second option might sound tempting but if you and your enemy have $1000bil to spend each the choice is not so simple any more.

If you buy 100 AI-missiles and your opponent buys 100 fighter-crafts and 9000 "semi-smart" missiles you might be in trouble, especially in a game of "saturate the PDS".

 

[TheDungen]

Just because someone is right, its no guarantee, that everybody else sees it the same was. Especilly in the internet.

Peoples opinions matter little when it comes to right and wrong.

 

[valergain]

Yeah... the most powerful military force in the world got also handed his ass a few times by lesser forces..(Vietnam, Irak, Canada) But thats OK, because its the USA.

You, dear sir, are absolut not able to compare the usefulnes of weapons.

I can assure you the reason the US military lost in those cases was not because the enemy had knives. And again the appropriate analogy here is not using knives in modern combat it's using mounted cavelery in modern combat,

Also all wepeoens are useful in a certain context. It's just question of whether or not they are worth investing in.

 

[Kuschelflummi]

I can assure you the reason the US military lost in those cases was not because the enemy had knives.

No they lost because they were to stupid to use their weapons right!

And its not. Cavalry and Fighter have not smiliarity. Soley because unlike knifes, Horses are animals. They have to be trained for war.
A Spacecraft on the other paw, not unlike a knife, is build for battle.
Also I used that anaolgy, because some people here think, they KNOW without ANY DOUBT that Space Fighters are useless.

 

[valergain]

No they lost because they were to stupid to use their weapons right!

Which in no way contradicts my statement.

And its not. Cavalry and Fighter have not smiliarity. Soley because unlike knifes, Horses are animals. They have to be trained for war.
A Spacecraft on the other paw, not unlike a knife, is build for battle.

I disagree. A horse in this case is a vehicle, used for transportation. The reason I bring them up is because I think fighters in space are a similar case that the method of transportation. You see space is big, you may think it's a long way to reach Paradox Development Studio but that's just peanuts to space. And battles is space are likely to take place over massive distances. Fighters would to be able to accelrate and deacelrate at much higher rates than normal ships. At which point why not make the normal ships that fast? But if the normal ships are that fast then the fighters become redundant.

Also I used that anaolgy, because some people here think, they KNOW without ANY DOUBT that Space Fighters are useless.

As I think I said every weapon is potentially useful based on the situation. And as for fighters being useless in space it certainly seems that way right now based on what we know about the future of space combat.

 

[Kuschelflummi]

You know nothing, Jon Snow.

Alway wanted to quote that

 

[IronXZ]

3. Optimum Design: An AI-piloted missile
That assumes AI is viable, reliable, cheaper than fleshlings and less corruptible. The reason why not all soldiers today have a remote controlled tank isn't because they don't exist but because soldiers are cheaper.

Why do you guys think we would need super advanced AIs for the missiles to hit? MODERN smart missiles do their jobs pretty well, and they are used in an atmosphere. It's much easier to detect and hit things in the vacuum of space.

 

[EntropyAvatar]

Why do you guys think we would need super advanced AIs for the missiles to hit? MODERN smart missiles do their jobs pretty well, and they are used in an atmosphere. It's much easier to detect and hit things in the vacuum of space.

The only reason I can think is the possibility of some kind of ECM arms race. Like a specialized ECM AI that is orders of magnitude smarter than the missiles could maybe figure out how to play them like a fiddle.

 

[Teleros]

You'd be surprised.

Point about the computers used in radar. Still, you can keep them separated from the main AI.

The AI must have some way to interface with its sensors though or it's blind

Humans don't need a direct interface though, we just look at our displays. Now, a direct interface is likely to be more efficient than being physically isolated like humans are, but if getting hacked via the sensors is a realistic threat for your firewall tech etc, then it seems a sensible thing to do, no?

This is a strong point, assuming that ships in the Stellaris setting use reaction drives for in-system movement. A lot of settings feature some kind of "magic drive" for sublight, to go along with the magic FTL. It seems to me to be a choice whether, in the Stellaris setting, FTL technology is basically a one-trick-pony and we use currently-understood physics for everything else, or if some of the magic of FTL also leaks into sublight travel.

Indeed. I'd love to see Doc Smith-style inertialess travel, but I suspect we'll have something more conventional.

If you allow for some weird physics for sublight travel in the game setting, then a reusable delivery platform could make a lot of sense.

Snipping the rest, but... basically I agree.

making assumptions on AI capabilities while not doing it for engines is a bit strange.

Well, if it's a reaction drive then it needs propellant to work at all, and if it's a reactionless drive then congratulations on breaking physics . Though you're right that I am making assumptions about such things, but I think they are reasonable ones from a hard science perspective.

1. Mission:

Deliver lots of BLAM to the target:

Can be done in a myriad of ways but only smaller craft enables you to do so in many places at once at remote locations without exposing your precious fleet. The sensors and computing-power of a smaller craft is not blown up along with the warhead as in a missile.

Yep - that's in part why I suggested carriers as patrol / escort vessels.

2. Limitations:

Delta-v:

Assuming that the engine and power technology favors larger engines and larger masses (there are plenty of technical effects today that work well with low masses but worse with larger ones).

That's pretty reasonable though, no? Big ships in space tend to be slow because if nothing else pulling multi-g manoeuvres in them will be much more likely to break things. Smaller craft can typically withstand higher g-forces for this reason.

3. Optimum Design:

An AI-piloted missile

That assumes AI is viable, reliable, cheaper than fleshlings and less corruptible. The reason why not all soldiers today have a remote controlled tank isn't because they don't exist but because soldiers are cheaper.

Well, there are also lots of jobs a remote-controlled tank can't or won't be able to do.

A 50 tonne $1bil craft with advanced sensors, navigation and guidance that can be used multiple times to fire $100k "semi-smart" missiles or a 20 tonne $10bil missile that can only be used once. If you don't care about economy the second option might sound tempting but if you and your enemy have $1000bil to spend each the choice is not so simple any more.

I'd be surprised if the missile costs 10x as much as a fighter, complete with skilled pilot and all that aboard. Even leaving aside all the pork in modern defence contracts, fighter pilots and their aircraft tend to be some of the most expensive objects you can buy - doubly so if you think in terms of money per tonne of hardware bought.

If you buy 100 AI-missiles and your opponent buys 100 fighter-crafts and 9000 "semi-smart" missiles you might be in trouble, especially in a game of "saturate the PDS".

Recall that my idea was a missile with its own submunitions (the semi-smart missiles in your case). Think "kamikaze AI-piloted fighter" - it acts like your meatbag-piloted fighters, but can probably act/react faster, pull higher-g manoeuvres and expend more delta-v in said manoeuvres (because it isn't expected to re-arm or resupply), and after firing off its own missiles can then blow itself up on a convenient target (even if it has no warhead of its own, an object impacting at ~3km/s will deliver kinetic energy roughly equal to its own mass in TNT).