Fighters in space

[NecronLord]

assuming it is obviously illegal,

And so therefore you want to go into little tricks involving collateral damage. The responsibility is still the officer's.

[Xon]

[quote="NecronLord]
AS for hacking it the answer is simple. Have loyalty to the state hardwired into the processor.[/quote]
Thats beign too complicated, just hardwire the damn thing not to accept anything that would change its programming unless you have physical access. One time pads(fitted at launch), can be used to allow drones to exchange intel & cordinate themselves and allow controllers to still issue commands to the drones.

[Marcus]

This may be a cop out, but what tech level are we dealing with, here?

If computer technology is, say, todays... Go with the human, for now.
Given the computer technology of your typical, mainstream Sci/Fi universe... definitely go for the Human. Id be willing to bet that Moore's law will carry the real world well beyond most Sci-Fi computers in terms of speed and raw calculating power within my lifetime.

Given open ended computer tech? Well, then your 'Drones' are crewed by 'People' that just happen to not be made out of Carbon Compounds, and your back to the original question.

Ultimately the question is one of offense vs defense, and of the nature of propulsion technology and sensor technology.

[BenRG]

An example of AI confusion. There are so many inputs (threats, communications, countermeasures, the blinding flashes of energy from power core breaches, etc.) coming in at such a rate that the AI locks up because its' 'buffer' is full. The advantage of a human in such a situation is that we rarely wait until we have received all or even most of the sensory information about our situation before we react instinctively to any given threat. Basically: "Oh dear too many things happening! Who cares? Start shooting and get the hell out of here!"

An AI, even if it has a near-infinate number of programmed responses to any given situation, is still limited to only those responded. A truely sentient pilot (be it human, clone, alien or even a brain-in-a-vat) can generate completely new behaviour patterns when faced with an unexpected situation.

[SWPIGWANG]

god, if only 1/9999 percent of anti-bioship setiment is directed here, as here is a prime example of humans thinking too much of themselves


Anyway, if this keeps up we'll need a self aware bullet to fight wars as well as a total break in the command structure.

So it is better to have a army that does not follow commands....

[Xon]

An example of AI confusion. There are so many inputs (threats, communications, countermeasures, the blinding flashes of energy from power core breaches, etc.) coming in at such a rate that the AI locks up because its' 'buffer' is full. The advantage of a human in such a situation is that we rarely wait until we have received all or even most of the sensory information about our situation before we react instinctively to any given threat. Basically: "Oh dear too many things happening! Who cares? Start shooting and get the hell out of here!"

So basicly, your saying that if an AI has too keep track of a few billion factors, it will go flubber? Considering people normally go flubber with < 1000 things occuring around them simultaiously I'm still voting for the AI.

As for a limited buffer, WTF! the is no such thing as a limited buffer size. For modern computers there is only ~ 4billion address bytes of memory(barring memory paging). 64 bit base computers, can address ~1.844*10^19 bytes of memory.

Lets say each event takes 100MB, and the code for all the managment is 100GB, that still allows for ~2^36 or 6.87*10^10 unique events. Which is about 68.7 billion unique events.

The events would probable be processed simulationsly by parallel processors, as that is the most efficent way to handle huge volumns of information.

An AI, even if it has a near-infinate number of programmed responses to any given situation, is still limited to only those responded. A truely sentient pilot (be it human, clone, alien or even a brain-in-a-vat) can generate completely new behaviour patterns when faced with an unexpected situation.

This is so fucking wrong it isnt funny!

Do a google on a topic call 'neural networks'
What is a Neural network

Learning != sentient

[BenRG]

So basicly, your saying that if an AI has too keep track of a few billion factors, it will go flubber? Considering people normally go flubber with < 1000 things occuring around them simultaiously I'm still voting for the AI.

I'm saying that an organic brain wouldn't bother to go 'flubber' (such a strange word). It would just disregard all input and act on instinct. A computer would try and process all input from its' sensors, taking up processing time, slowing response times and maybe even causing a kind of electronic 'traffic jam' as incoming data has to be held until it can be processed.

This is so fucking wrong it isnt funny!

Don't shout at me on the subject, prove it instead.

Do a google on a topic call 'neural networks'

(...)

Learning != sentient

Actually, sentience is more than just learning. There are other factors to be considered, but that isn't relevent to the topic.

The point is this: Even a true neural-net AI cannot create completely new behaviours, just modify its' existing program for new circumstances (a quantum leap, true, but still limited to an evolutionary progression). Humans have a tendency to do stupid/unprecidented things. "Hey, instead of flying around this enemy ship, why don't I fly through the landing bay that goes its' entire length...?" Probably suicidal, but so unexpected it could work. An AI would never even consider such a manoeuvre, unless its' programmer was a bit... strange.

[Hotfoot]

So basicly, your saying that if an AI has too keep track of a few billion factors, it will go flubber? Considering people normally go flubber with < 1000 things occuring around them simultaiously I'm still voting for the AI.

I'm saying that an organic brain wouldn't bother to go 'flubber' (such a strange word). It would just disregard all input and act on instinct. A computer would try and process all input from its' sensors, taking up processing time, slowing response times and maybe even causing a kind of electronic 'traffic jam' as incoming data has to be held until it can be processed.

So you completely ignore having drone operators giving the drones orders from the ship/station that they were launched. As for confusing the AI, you assume that the drone would be unable to determine the importance of all the data in takes in, and thus would look at every bit equally? Even were this to be true, how much data could there possibly be attacking the processor at once? Why not use dedicated processors for each function, so as to improve performance and add redundancy?

This is so fucking wrong it isnt funny!

Don't shout at me on the subject, prove it instead.

Do a google on a topic call 'neural networks'

(...)

Learning != sentient

Actually, sentience is more than just learning. There are other factors to be considered, but that isn't relevent to the topic.

He just said that. != is a boolean meaning "not equal to", thus, learning is not equal to sentience. For future reference, here are further booleans:

< less than
> greater than
<= less than or equal to
>= greater than or equal to
== equal to
!= not equal to
|| or
&& and

The point is this: Even a true neural-net AI cannot create completely new behaviours, just modify its' existing program for new circumstances (a quantum leap, true, but still limited to an evolutionary progression). Humans have a tendency to do stupid/unprecidented things. "Hey, instead of flying around this enemy ship, why don't I fly through the landing bay that goes its' entire length...?" Probably suicidal, but so unexpected it could work. An AI would never even consider such a manoeuvre, unless its' programmer was a bit... strange.

The ability to do something stupid or unpredictable is limited at best. Nevertheless, should there be a need for "human ingenuity", ther drone operators can simply give the drones orders to fly through the landing bay that goes the entire length of the ship. Drones are far more likely to obey this order even if it is suicidal because they have no fear of death. However, ask yourself, how many battles are won because of some stupid or unpredictable stunt, other than in the movies? How often are these stupid or unpredictable stunts even worthwhile? Chances are they'd simply be suicidal. If you're comparing this to evolution, remember that most mutations are not beneficial, and that many are in fact fatal.

You're assigning unrealistic limitations to the AI, while giving overly too much credit to the human pilots. Let's go through the advantages of each one more time:

Drones

• Higher acceleration/velocities than human-piloted craft
• Higher accuracy with weapons than human-piloted craft
• Better and more efficient piloting than human-piloted craft
• Obeys orders better than human piloted craft
• Utterly disposable with no loss of human life

Fighters

• Can think of stupid or unpredictable things to do on the fly. (more of a negative, but that's the best I can think of)

[BenRG]

So you completely ignore having drone operators giving the drones orders from the ship/station that they were launched. As for confusing the AI, you assume that the drone would be unable to determine the importance of all the data in takes in, and thus would look at every bit equally? Even were this to be true, how much data could there possibly be attacking the processor at once? Why not use dedicated processors for each function, so as to improve performance and add redundancy?

Ah! I didn't know that this was an option available. I thought that we were talking about completely autonimous AI fighters, not remote drones.

Of course, the biggest problem for remotes is the need for a secure com link to the mother ship or control station. Given the sort of electronic environment of most space battles I've seen (charged particle blasts going everywhere and lots of active jammers), I would have thought that a remote control option would be eliminated pretty quickly.

I'm no expert in computing, but even if you seperate each function to its' own processor, surely there is still a certain amount communication of overload from one system to another. Remember, for example, the flight control subprocessor will need information from the navigational sensor subprocessor before it can begin manoeuvres. I'm not suggesting that this is a fatal error, but I'm just trying to demonstrate that there will be problems in high-information-density situations, no matter how much you decentralise the control system.

[Knife]

.......why fighter when you can missile?

and is there any good reason for space fighters to exist without having physics that make newton roll in his grave?

All would depend on the firepower/armor/range of the ships in your universe. If enough firepower can be loaded into a missile to penetrate a capship's defense's, then a fighter/bomber type vessel would be handy(human or droid pilot withstanding).

It would follow a model no all that unlike the modern Navy and its carriers use. A large ship that can hold small fighter bombers or attack craft, arrives near a target and launch its attack craft. The attack craft moves towards the target and the mother ship does not have to put itself into danger to destroy the target. The attack craft would come into range and launch their weapons and depending on the situation use multiple attack vectors, come up real close to minimize ECM, and use other tactics.

Since you have realitively small manuverable craft out there that can deliever a weapons payload that is a threat to capships, using another small, quick, manuverable craft as an intersepter to destroy the attack craft would be handy. Thus you would have fighter/bombers and figher intersepters stationed on the mothership.

If the attack craft cannot deliever a payload that would be a threat, the the use of small fighter like ships would not be that probable.

[NecronLord]

An example of AI confusion. There are so many inputs (threats, communications, countermeasures, the blinding flashes of energy from power core breaches, etc.) coming in at such a rate that the AI locks up because its' 'buffer' is full. The advantage of a human in such a situation is that we rarely wait until we have received all or even most of the sensory information about our situation before we react instinctively to any given threat. Basically: "Oh dear too many things happening! Who cares? Start shooting and get the hell out of here!"

An AI, even if it has a near-infinate number of programmed responses to any given situation, is still limited to only those responded. A truely sentient pilot (be it human, clone, alien or even a brain-in-a-vat) can generate completely new behaviour patterns when faced with an unexpected situation.

The human brain has a buffer it is called short term memory. Look at a page of a book and try to recall the first sentance ten seconds later, Then do the same without repeating the words over and over or other memory techniques, 40 seconds later, you will have much more sucess. In actuality a drone would have far better memory capactiy than a human.

For the second part need i remind you that by the time the human has generated a new behavious patter (which will often get them killed) They are a rapidly expandig debris cloud. As mentioned above in the same time a neural net processor would have come up with many possible soloutions, simulated written the best ones to memory and begun excecuting one.

I'm saying that an organic brain wouldn't bother to go 'flubber' (such a strange word). It would just disregard all input and act on instinct. A computer would try and process all input from its' sensors, taking up processing time, slowing response times and maybe even causing a kind of electronic 'traffic jam' as incoming data has to be held until it can be processed.

Need I remind you that the proposed drone already has instinct. To protect its citizens and kill their enemies, as opposed to frankly useless instincts consitisting of screaming urinating and soiling. As for a drone attempting to process everything it doesn't work that way, a computer processes one task at a time. It just does them at a blindingly fast rate.

To conclude a drone can, in the time it takes a human to think up some on the fly solotion (not to mention overcome his/her fear of dying if it doesn't work) have developed simulated and tested potentially hundereds if not thousands of potential soloutions.

I'm no expert in computing, but even if you seperate each function to its' own processor, surely there is still a certain amount communication of overload from one system to another. Remember, for example, the flight control subprocessor will need information from the navigational sensor subprocessor before it can begin manoeuvres. I'm not suggesting that this is a fatal error, but I'm just trying to demonstrate that there will be problems in high-information-density situations, no matter how much you decentralise the control system.

Answer is that you have a heirachy to direct overall funtions down to subprocessors controlling things like stabalisers. Each one you give a stupidly large (by our standards, say about a gigabyte) cache memory of its own in addition to the main memory.

[Hotfoot]

So you completely ignore having drone operators giving the drones orders from the ship/station that they were launched. As for confusing the AI, you assume that the drone would be unable to determine the importance of all the data in takes in, and thus would look at every bit equally? Even were this to be true, how much data could there possibly be attacking the processor at once? Why not use dedicated processors for each function, so as to improve performance and add redundancy?

Ah! I didn't know that this was an option available. I thought that we were talking about completely autonimous AI fighters, not remote drones.

Even human pilots in fighter craft receive orders from the carrier that they launched from. Why would you assume that you would not be able to issue orders in a similar fashion to a drone? The drones would have AI capable of making decisions, but wouldn't do anything it wasn't ordered to do by a commanding officer (in this case, a drone operator).

Of course, the biggest problem for remotes is the need for a secure com link to the mother ship or control station. Given the sort of electronic environment of most space battles I've seen (charged particle blasts going everywhere and lots of active jammers), I would have thought that a remote control option would be eliminated pretty quickly.

Not a significant problem, really. Encrypted communications are much easier and more secure than most Science Fiction would have you believe. Even if quantum computers are commonly used to break incoming codes on every ship in the fleet, they still can't break one-time pads, which are becoming more and more easier to use thanks to the advances of technology. As for actually getting the message to the drones, a communications laser could be used to ensure a direct link to the drones, and then back to the ship. You could easily design the drones in such a fashion that they will distribute orders to each other as they get them, so that even if half of the drones can't establish a direct link to the parent ship/station, they can still receive orders so long as a fraction of the drones receive them and can transmit them to the other drones nearby.

In any event, it becomes a somewhat moot point if specific orders for a mission are loaded prior to launching the drones. The drones could handle enough decision making within the parameters of a predefined mission. Obviously, there would be a failsafe state for the drone should the shit hit the fan, which could be defined before the drone was launched. If things are really so bad that a drastic change of plans is required and the parent ship can't contact the drones and the drones don't know what to do, you're FUBAR anyway, and having human pilots probably wouldn't make the situation any better.

I'm no expert in computing, but even if you seperate each function to its' own processor, surely there is still a certain amount communication of overload from one system to another. Remember, for example, the flight control subprocessor will need information from the navigational sensor subprocessor before it can begin manoeuvres. I'm not suggesting that this is a fatal error, but I'm just trying to demonstrate that there will be problems in high-information-density situations, no matter how much you decentralise the control system.

Again, we return to algorithms determining what data is important and what data is junk. Decentralized processing would make the processing of the data much faster with specialized processors (think of it like how a really advanced 3D graphics card has its own processor and memory seperate from the normal computer). That's the hardware aspect. Now you're focusing on software, how it decides to process the information. In space, you don't need to worry about many of the things you have to consider when flying an F-16. There is no ground to crash into, no major gravitational forces immediately acting on you, no worries about drag or lift, you can make snap 180 degree turns at high velocity with no significant stress on the frame of the ship, and so on. There are just a few navigational hazards that the AI would have to worry about:

• Asteroids
• Planets
• Moons
• Stars
• Comets
• Other ships
• Incoming missiles/projectiles

In all of these cases, incoming information via sensors is not likely to be overloaded. Active sensors might be jammed, but passive sensors would still function normally. Basic program would be to manuever without colliding with any of the above objects. Not terribly difficult to do, with the exception of incoming fire. Here's how a sample engagement could happen:

Code: Select all

Sensors: picking up unknown contact at 100km, relative velocity 800m/s and accelerating at a rate of 60m/s.  Is not broadcasting valid IFF or civilian codes.
Communications: notify command of event, proceed with standing orders until told otherwise
Manuever: move to match velocities and intercept target
Weapons: calculating firing data, ready to open fire once we are in range.
Sensors: contact silhouette identified, civilian shipping.  Hold fire unless fired upon or friendly ship is threatened.
Communications: Updating command of status, attempting to open communications between contact and command.
Manuever: target intercepted, holding position at 800 meters from target.

There, the drone could be ordered by command to destroy the ship anyway, but since it was identified as a civilian ship, it asked for clarification. Should the supposedly civilian ship have opened fire, then the drone would have returned fire. But let's assume a more military situation.

Code: Select all

Sensors: picking up unknown contact at 100km, relative velocity 800m/s and accelerating at a rate of 60m/s.  Is not broadcasting valid IFF or civilian codes.
Communications: notify command of event, proceed with standing orders until told otherwise
Manuever: move to match velocities and intercept target
Weapons: calculating firing data, ready to open fire once we are in range.
Sensors: contact silhouette identified, military vessel on a combat alert.  It is launching drones and missiles.
Communications: Updating command of status.  Coordinating targeting data with friendly drones.
Manuever: target intercepted, beginning evasive manuevers and dropping countermeasures to evade incoming missiles.
Weapons: target priorities calculated, drones currently have higher priority than parent vessel.  Targeting nearest drone and opening fire.
Manuever: target locked, manuevering into firing position as requested.

Could the system be overloaded? Sure, if the processor isn't fast enough or the algorithms eat up too many cycles, but it's not terribly likely. Even if it were to happen, and there was too much going on for the drone to keep track of at once, it would have priorities as to which data gets processed before anything else. It doesn't need to be able to track every asteroid in the asteroid belt between Mars and Jupiter if it's doing manuevers between Neptune and Uranus. Even if it were in the asteroid belt, it would only need to track the asteroids that were close enough to be significant threats to manuevering.

[Joe Richter]

A star fighters main limitation is always going to be its pilot; low resistance to aceleration, slow reaction times, low multi tasking capability, suceptability to fatigue/stress/fear, the list goes on. Even if you have to hamstring a drone to an INCREDIBLE level, it is still going to outperform any fighter of simelar tech base.

As a final example, we have a simelar situation in the present day with cruise missles and the like, the only difference being that they destroy their targets by "ramming them till they give up" rather than shooting from a distance. It is still a remote drone that is programmed, launched, manouveres automonously to its target, bypassing any intervening terrain and does everything in its power to anihilate its target. they DON'T get confused by flying through a raging battle with explosions. They DON'T get hacked by the enemy and turned on their owners. They DO outperform any fighter available. They DO commit "suicide" when told to.


And this is all with our present day technology. Technology will continue to advance. Propulsion technology, weapons technology, sensor technology, all this will advance and drones will be on the forfront of this, making best use of the technology, while manned fighters will always limited by the frailty of their pilots. Always limited to the same maximum aceleration. Always limited to the same amount of data processed. Always suceptable to fatigue/stress/fear.


Anyone who thinks fighters will be viable in any way other than a public relations way in as little as one hundred years is deluding themselves.

[tharkûn]

Alot of people have complained about the physical limitations of fighters ... so what? Most of the time you aren't going to be trying for high g turns or maximum thrust. Can missiles outperform fighters? Yes. Do fighters carry missiles for this reason? Yes. Realistic space combat would only be close range if jamming is an issue ... most encounters would involve lobbing missiles far beyond the range of the human eye to discern targets. You use fighters because they give the maximum range with the maximum decision making capabilities.

As far as AI control goes ... do you REALLY want your WMD's controlled by a string of ones and zeros? If you have millions of scenarios preprogrammed what are the chances that the code will be COMPLETELY bug free? In the end the integrity of the code comes down the humans who wrote it, no computer program can tell if another is completely bug free unless it is bigger and written by humans also.

And of course there is always the fun thing of what happens if the enemy gets a complete copy of your AI source code, its every action is completely PREDICTABLE. Even it's response to the enemy's apparant foreknowledge is predictable (because it is all in the code). Once the enemy has your source code they can plug it into a bigger computer (planet based) and then see exactly how your drone will react to a handful of scenarios and find ways to beat it. It is bad enough when the enemy figures out human your SOP's, but at least they cannot predict your response. With AI it is ALWAYS predictable (if you want perfect control) or it is randomized (and I don't mean "randomized" like modern computers using complex but ultimately predictable functions, but ones hardwired to random QM events like radiodecays), but then you no longer have a perfectly obedient AI.

The weakest link, as I noted earlier, is the PROGRAMMERS. What happens if your enemy bribes/coerces a programmer into introducing a single fatal flaw that doesn't show in 999,999,999 out of 1,000,000,000 cases. What happens if the programmers include a backdoor that allows them to take control of the system? You are now in the situation where a handful of programmers ultimately control huge masses of weapons with no one else able to countermand their actions. Think about the possibility of a military coup, the top brass bribes the programmers and testers to let the next generation of weapons go with the ability to perform illeagal orders (like say targeting the prime minister) and take over. Instead of having to subvert an entire military they need only go after the programmers.

The fact of the matter is drones are vunerable. The run through binary logic and are completely predictable (or not completely controlable), the enemy needs only to grab the code for a relatively small computer on a craft and plug it into a bigger land based computer and work out how to defeat the AI in a few specific scenarios. And this of course makes your programmers the weakest link in the chain; from coding errors to design flaws, to active sabotage they can do whatever they want and only other programmers can stop them. As the size of the AI increases it is going to be harder and harder to double check all your programmers work in entirety.

Joe: Cruise missiles CAN'T deal with a raging battle the entirety of a cruise missile is flight control. It's a glorified autopilot, it doesn't have great FoF recognition, it doesn't react if the target you told it to nail was hostipal, etc. Likewise the missile doesn't have to pick its own target (do I kill the SAM site or the weapons factory first?), it doesn't have to deal with targets who surrender, and it doesn't have to deal with enemies equally equipped.

You are also forgetting something else: those cruise missiles need a LAUNCH PLATFORM, be it converted battleships, subs, missile frigates/destroyers, or aerial fighters. How do you deploy your missiles in space? As stand alone units? Forget about it, the enemy need only build one warship capable of blowing your missile before it hits (be it AMM, lasers, high speed mass drivers). Clustered in groups? Not good either as the enemy can just bypass the cluster, or make a run at it with an area effect missile. The reason the carrier group dominates is it is strong enough to withstand attack, it can dominate a large radius of approach. The only way to take down a carrier group is to swarm it and overpower it (especially if it has multiple carriers).

Now the whole system could go AI (if you trust your programmers and you have that level of computer tech), but you still do not find a better bang for your buck than fighters equipped with missiles. The missiles have all the inherent advatanges of not having a pilot. The fighters give your launch platform the maximum range with the minimum target size. The carrier group gives you the largest operational radius.

It's called combined arms. Effective use of destroyers, cruisers, carriers, fighters, and missiles will beat the crap out of any single unit strategy.

[Xon]

Please, the AI that would be used wouldnt be ripped off a stupid RTS game.

The AI I would propose would be based of a combination of neural nets & expert systems, which closly mimics the human brain(but is really really fast).

With Modern AI, the AI is not purely code but as includes a dynamic data element.

There is a different been constantly using a feature and needing the feature at a critical time. Plus with the limitations of humans, modern fighters are critically hamstrung. Most of the time a fighter would be in the idle state or getting to a combat location, but once in combat it sure as hell needs those high gee turns & maxium thrust.

One critcal reason fighters dont outmanoeuvre fighters, is becasue of human limitations. Plus the use of proxy detontated explosives make narrow misses a moot point.

[Joe Richter]

tharkûn, I fail to see why a fighter would have a greater maximum range than a drone, what with the necesity for greater radiation shielding, life support and the need to limit the acceleration (a unit that can reach a point in space quicker has less chance of being intercepted). As for needing a launch platform for these missles.....well, Duh! What do you think aircraft carriers are? Thats right, plane transportation and launching vessles.

On the topic of Cruise missles you are right in all your points, in fact you make a good job of pointing out the limitations and crudity of them, yet they are STILL usefull! They are so basic in fact that I am unsure wether they should be classed as drones or autonomous missles at all (for the purpose of this thread) but they are still very usefull and efective. Just think, the following generations will have better performance characteristics; better friend or foe recognition, limited ability to change
mission priorities on the fly and suchlike, but the fighter will still be limited by its pilot, still limited to the same reaction and input management speeds, still limited to the same low accelerations, still being more mass and space intensive.


Refering to the initial thread topic, autonomous missles will have greater range than a fighter (less mass due to no lifesupport and less rad shielding needed), greater acceleration and manoverability, much faster reaction time and decision making capabilities, be phisically smaller and so will be harder to spot, quicker to produce (no pilot training time needed), more survivable (redundant key systems can be built in, you aren't going to put in a redundant pilot are you?). The list goes on.

In fact, your only cogent argument revolves round the vulnerability of this system to enemies compromising the programmers. This argument is good as far as it goes, but what realistically can anyone do when a barrage of missles have their targets hard coded into them at the time of launch, again have hardwired friend/foe recognition tables and do not recieve or respond to signals once launched (all basic and obvious steps against such an ocurance)?

[tharkûn]

The AI I would propose would be based of a combination of neural nets & expert systems, which closly mimics the human brain(but is really really fast).
Yes and that is still quantifiable. Take the source code, dump it into a yet larger and faster computer and the AI is completely predictable. Neural nets do not return randomized values, they return fairly uniform values. So I take your source code, dump it into a a computer a few orders of magnitude closer, pick a few scenarios I want to win at, and then run your code in a virtual environment until I know everything your AI will ever think to do. Data can be simulated as needed.


There is a different been constantly using a feature and needing the feature at a critical time. Plus with the limitations of humans, modern fighters are critically hamstrung. Most of the time a fighter would be in the idle state or getting to a combat location, but once in combat it sure as hell needs those high gee turns & maxium thrust.
AtA missiles already shoot over the horizon, this trend will continue. The big thing you need to do is make your fighters have low mass with high burn ... mainly to get to combat ranges. Once you get to combat ranges you will be using optics to track the opponent ... by the time his missiles get to you, you will have point countershots in flight. No matter how fast your drone is, it is still going to be slow as hell compared to a tracking laser. Let's say it does .1c, by the time your drone (or missiles) get within range of the fighter, the fighter has deployed counter measures with similar velocities. If laser weapons become viable then your drone will get shot out of the sky ... nothing outruns a laser.

One critcal reason fighters dont outmanoeuvre fighters, is becasue of human limitations. Plus the use of proxy detontated explosives make narrow misses a moot point.
Sure but what range of engagement are you talking? Only at short ranges of enegament do the drones have a signifant lead. In most cases your biggest delay is not reaction time, but ToF. Proxy detonations work both ways, you can take out a fighter with one, but a fighter can take out your missile with one.

The era of dogfights is long gone. Today AtA combat occurs beyond the horizon, you get the enemy in radar, send a smart missile and hope his countermeasures are not sufficient. It will be the same in the future (unless laser systems become viable weapons). The limits of the human will be minimal. These are basic givens:
1. The missile is always faster than the launching platform.
2. The longest delay is ToF, not reaction time, not maneuver time ... its all about ToF.
3. Optics allows you to indentify threats with a fraction of the ToF said threats will have.

So your drone comes within 5 minutes (at maximum acceleration) from my fighter. I get optics on your drone and begin taking counter measures. Sometime 2 minutes later my countermeasures hit your drone about 2/3 -1/2 the initial engagement distance. Your drone can either defeat my counter measures, and I'm screwed, or your drone is dead. This doesn't change if it is a fighter on the other side. The fighter comes up to 5 minutes ToF, launches its missiles, I detect them, launch counter meausres against them and launch a counterstrike. He detects my attack, launches his own countermeasures. First one to get a shot through the countermeasures wins.

Your drone gets minimal advatanges, it can turn and accelerate faster ... fine but short range missiles can do both faster still. It has faster reaction times, fine but the limiting step is ToF.

It has the added disadvatantages of having its code being a hideous weak point. If I get the code I can run it in a larger, faster computer and run a few million (billion) scenarios and predict exactly how your AI will react 99% of the time. I can then egineer my attack forces to take advantage of this foreknowledge and ensure victory.

Remember your AI has limited computational capacity, as dictated by the size of the drone. A military R&D complex where I take the source code to make predictions on your AI is not subject to the same limitations.

This completely ignores the extreme security risk in the programmers.

[SWPIGWANG]

Remember the following

Fighters have less range than missiles (given same mass)

Fighters have less firepower than missiles (given same mass)

Fighters have less survibility than missiles (more mass to life support, pilot, fuel...etc)

Fighters have less effectiveness generally than a drone due to limitation of the crew and weight of additional systems

It has the added disadvatantages of having its code being a hideous weak point. If I get the code I can run it in a larger, faster computer and run a few million (billion) scenarios and predict exactly how your AI will react 99% of the time. I can then egineer my attack forces to take advantage of this foreknowledge and ensure victory.

Remember your AI has limited computational capacity, as dictated by the size of the drone. A military R&D complex where I take the source code to make predictions on your AI is not subject to the same limitations.

RANDOM NUMBER GENERATOR
Just throw a few qubits into my computer and my system becomes effectively impossible to predict. *Ah the wonders of quantum randomness*

And you aren't getting the code.

This completely ignores the extreme security risk in the programmers.

You have magnitudes less programers than pilots, nuff said.

After all, a pilot leaking comm. frequency can be just as harmful as source code leak.

[SWPIGWANG]

GAHHHHH STUPID NO EDIT BUTTON

The small font is RANDOM NUMBER GENERATOR

[tharkûn]

tharkûn, I fail to see why a fighter would have a greater maximum range than a drone, what with the necesity for greater radiation shielding, life support and the need to limit the acceleration (a unit that can reach a point in space quicker has less chance of being intercepted).
The fighter is an extension of the carrier. You suggest replacing fighters with drone missiles from a carrier, correct?

So what is bad about this? The carrier is engaging the enemy directly. Two carriers come up alongside each other (maybe thousands of km) and begin lobbing missiles/point defense shots. Assuming parity there are 3 possible outcomes:
1 wins, mostly intact, the other is obliterated.
both die
they both run out of weapons (it can happen).

In the first shot it's great to win, sucks to lose. The enemy may have taken out a cap ship for only the cost of missiles and counterfire. Your investment in the capship is blown with minimal damage to the enemy.

In the second case nobody is better off. You gain nothing, you lose nothing.

Same with the last.

Now let's say you have carriers with fighters, they come into range of each other, fighter patrols are sent out and two fighter wings come into contact with each other. Fighters then exchange missiles. Here you have more options, you may lose, but you may take 3/4ths of the enemy with you. If you die you lose only the fighters, not your capship. Even if the enemy completely routes your fighters, they are far enough out your carrier has room to run. This is why fighters carrying smart "short" range missiles are normally the best option. The present the cheapest target, with longest range, and when you lose it isn't all or nothing.

Other fun things that make fighters superior to carriers. A carrier can only be in one place at one time. If you have a missile carrier and I have a carrier with a swarm of fighters I can send half the fighters to dick with you, and the other half the run around you (say to bombard a planet). You can only be in one place at one time. My fighters can be in two. Divisibility of firepower is a good thing.

Just think, the following generations will have better performance characteristics; better friend or foe recognition, limited ability to change
mission priorities on the fly and suchlike, but the fighter will still be limited by its pilot, still limited to the same reaction and input management speeds, still limited to the same low accelerations, still being more mass and space intensive.
Except that the fighter will be carrying cruise missiles. Modern AtA missiles (high end ones at least) are damn close to cruise missiles. The thing is the only time maneuvering means a damn is when the weapons are within kill/damage range of each other. If your cruise missile has a long ToF, it's big and less maneuverable. My short range missile can outmaneveur it and kill it. Short range AMM's are ludicriously cheaper long range cruise missiles.

Refering to the initial thread topic, autonomous missles will have greater range than a fighter (less mass due to no lifesupport and less rad shielding needed), greater acceleration and manoverability, much faster reaction time and decision making capabilities, be phisically smaller and so will be harder to spot, quicker to produce (no pilot training time needed), more survivable (redundant key systems can be built in, you aren't going to put in a redundant pilot are you?). The list goes on.
Which is why you give your fighters short range autonomous missiles. and engage at optimal distance. With optics long range missiles are trackable long before they get close enough to hit. Numerous short range missiles can be launched to intercept your long range shot LONG before it is in kill range.


their targets hard coded into them at the time of launch, again have hardwired friend/foe recognition tables and do not recieve or respond to signals once launched (all basic and obvious steps against such an ocurance)?
And when the enemy gets ahold your FoF table and uses its computers to find a way to trick the AI? There is no way to make a perfect FoF protocol, it can always be cheated. With humans error means you can't count on them to follow the FoF protocols exactly, they might still fire on you even if you have perfectly matched some obscure friend classification. With computers that is different. If the tables say you are a friend then you will be a friend. Knowing how the opponent will react means you can egineer the encounter to your advantage. Defense has to be ready for anything, offense only for what it plans to do.

Hard coding still does not take care of the problems with:
Coder error. Bug free software doesn't exist, putting a computer in charge of WMD's and then praying that somewhere in this gargantuan code somebody didn't flip an "and" to an "or" or slip in an extra "!" ... not my cup of tea. You also need to be sure your pysical systems are up to snuff, no stray charges ever dicking with the critical bits, no misalignments. A single flaw buried deep in the code can bring the whole thing down.

Programmer compromise:
All the programmer has to do is turn over the source code to the enemy, or introduce a minor, minor glitch that is hideously exploitable. Look at the number of hacks that occur because some programmer did something stupid that nobody caught, now imagine if the programmer was trying to do that on purpose.

Coup d'etat:
For normal military coups you need HUGE numbers of people willing to back you. Military coups generally do not occur without a good bit of popular backing (at least in the rank and file). With drone weapons you just need the top brass and the programmers, everyone else is useless.

Realistically you will see carrier groups with fighters (maybe missile boats) that allow your military maximum operational radius (due to the carrier), minimum target profile (both physical and in terms of target value) due to using fighters, divisibility of force (again due to using fighters), all the advatanges of drones by using them as your fighter's weapons.

[tharkûn]

RANDOM NUMBER GENERATOR
Just throw a few qubits into my computer and my system becomes effectively impossible to predict. *Ah the wonders of quantum randomness*

And you aren't getting the code.

What good is a completely unpredictable system? When presented with a threat you want your AI to respond with some semblance of logic. Of the bazillions of actions your AI could take due to the random numbers from quantum mechanics, only a few will happen. If I run your source code (feeding virtual input) one million times I have a DAMN good idea how your AI will react. Run it a few billion and you are toast. You have to program it to deal with some gawd-awful number of scenarios. I only have to work out a handful, and then equip my short range drones with your basic capabilities, plus a few insights gained from your source code ... I win.

You have magnitudes less programers than pilots, nuff said.

After all, a pilot leaking comm. frequency can be just as harmful as source code leak.
Except it's ludicriously easier to switch comm frequencies (not to mention encryption) than source code, besides which there is no reason a pilot needs to know his comm frequency (just give him choices of a-m). The fact that there are fewer programmers is a BAD THING. It concentrates power so that the enemy only has to compromise a few people to take out huge fractions of your force.

So let's say the enemy compromises a fighter pilot and he leaks comm frequencies. You get raped for two weeks and lose 40% more planes. You then swap frequencies and life is fine again.

Now let's say you a programmer who leaks code, you have to recall all of the ships, reprogram them sufficiently to take care of the problem. Especially given everyone's assumption of hard coding you have NO ability to fix them in the feild.

Fighters have less range than missiles (given same mass)

Fighters have less firepower than missiles (given same mass)

Fighters have less survibility than missiles (more mass to life support, pilot, fuel...etc)

Fighters have less effectiveness generally than a drone due to limitation of the crew and weight of additional systems
Remember the following:
The fighter is a platform to launch missiles from.


They are used because they are cheaper to loose than cap ships. They are used because unlike a cap ship their firepower can be split into multiple groups.

Remember in space there is an optimal engagement range. That is based off of how much time the target has to react (due to the long ToF) compared to how fast the missiles go. Engaging beyond that range is not viable against a target with AMM's. Sending drones from Earth to Venus is just going to give the guys on Venus time to mount a killer defense.

The real question you need to answer is:
Where do you launch your missiles from?

[Hotfoot]

Alot of people have complained about the physical limitations of fighters ... so what? Most of the time you aren't going to be trying for high g turns or maximum thrust. Can missiles outperform fighters? Yes. Do fighters carry missiles for this reason? Yes. Realistic space combat would only be close range if jamming is an issue ... most encounters would involve lobbing missiles far beyond the range of the human eye to discern targets. You use fighters because they give the maximum range with the maximum decision making capabilities.

Human-piloted fighters would be torn to shreds by drones/drone fighters, see the above.

As far as AI control goes ... do you REALLY want your WMD's controlled by a string of ones and zeros? If you have millions of scenarios preprogrammed what are the chances that the code will be COMPLETELY bug free?

Is a database any less stable the more data it has stored in it? Does it develop bugs when you have 10 scenarios stored or when you enter 1,000 scenarios?

Also, little bugs, should they appear, would not be a serious problem. Something which would cause serious errors, however, would be worked out before the drone AI would even be released. This is also where the redundancy comes into play.

In the end the integrity of the code comes down the humans who wrote it, no computer program can tell if another is completely bug free unless it is bigger and written by humans also.

If I was talking about, say, something like an Operating System or a program designed to run on an operating system, where there are tons of variables and things that could go wrong, I might agree with you. However, we're talking about something that would not have to worry about the dangers of feature creeping like most programs do. Turn on a normal calculator. Do any equation you can do with the input allowed to you. Try to make it crash, try to make it give you a wrong answer. Are there any bugs in it? Anything that would cause it to return the value of 4.6094 to the integer equation 2+2? I don't think so. Even the more advanced calculators don't have crippling bugs in their core programming, as your argument would insinuate.

And of course there is always the fun thing of what happens if the enemy gets a complete copy of your AI source code, its every action is completely PREDICTABLE. Even it's response to the enemy's apparant foreknowledge is predictable (because it is all in the code). Once the enemy has your source code they can plug it into a bigger computer (planet based) and then see exactly how your drone will react to a handful of scenarios and find ways to beat it. It is bad enough when the enemy figures out human your SOP's, but at least they cannot predict your response. With AI it is ALWAYS predictable (if you want perfect control) or it is randomized (and I don't mean "randomized" like modern computers using complex but ultimately predictable functions, but ones hardwired to random QM events like radiodecays), but then you no longer have a perfectly obedient AI.

Hah! What garbage. A drone would not be any more or less completely predictable than a human pilot in any significant way. Face it, the options available to any human pilot are ultimately limited. They can only do so many things in a fighter. Figure out what each of those options are, then set up a simulation based on every possible thing a human pilot could do in a fighter. A lot of those things will be stupid, like ejecting 2.305 seconds into the battle without a vaccuum suit (hey, who would expect someone to do that in a battle?), many will be just plain bad (never shooting at anything the whole battle), some will be decent, a few will be good, even less will be excellent, and the chances of doing something so incredibly spectacular so as to win the battle at the last minute are so rare, unlikely, suicidal, or improbable so as not to make a massive difference. So if the enemy can do simulations with the AI so far as to beat them every time, they can do just the same thing with humans, only there's more to the subject than your example allows.

So you know every possible response the drones could have to a situation. How does that help you in a battle? If we're talking about a turn-based competition like Chess or Go, then knowing that might help, but in a battle? Let's see, first the drones are given orders for the mission they are about to be deployed for. This is done on the ship, by the drone controllers. You'd have to know what sort of mission they have in store, and what their specific orders would be. Since you can't possibly know that before they do, you're stuck reacting. You also don't know how each drone will approach, when they will fire their weapons, exactly how they will chose to evade incoming fire, and so on. Also, allowing for randomly-determined actions would not make the drone any less obedient to orders. Why? Because there are a slew of possible actions available to carry out the orders. Selecting one at (quasi-)random makes them unpredicatable and thus less likely to get destroyed. It doesn't make them any less obedient in the slightest.

Now, you could say "well, then we know which possible actions it could take, we could lay down a spread of fire to cover the entire range of its possible evasion." Well, duh. You can do the same thing with a human-piloted fighter and waste less ammunition. Why? Human-piloted fighters are slower to change velocity, and thus more predictable. Also, the way humans are trained, they might commonly only learn, a few ways to evade incoming fire, and in the heat of the moment, remember only one or two. Much more predictable than an AI with potentially thousands of randomly selected evasion algorithms. In any event, it's still much harder to shoot down a drone than a human-piloted fighter, not to mention much harder to defend against, as the drone will always be far more accurate with the weapons given it than a human pilot could ever be.

The weakest link, as I noted earlier, is the PROGRAMMERS. What happens if your enemy bribes/coerces a programmer into introducing a single fatal flaw that doesn't show in 999,999,999 out of 1,000,000,000 cases. What happens if the programmers include a backdoor that allows them to take control of the system? You are now in the situation where a handful of programmers ultimately control huge masses of weapons with no one else able to countermand their actions. Think about the possibility of a military coup, the top brass bribes the programmers and testers to let the next generation of weapons go with the ability to perform illeagal orders (like say targeting the prime minister) and take over. Instead of having to subvert an entire military they need only go after the programmers.

Problem is, you'd have to be able to bribe not just one programmer, but dozens or more. That's an obvious concern, and one which is very easy to counter. First, obviously, would be a method of classifying the project. Only programmers that could be trusted not to be compromised would be hired for the project. Secondly, the code itself would be double-checked to see how each function works. Third, if a captain wanted to kill the Prime Minister with the drones on his ship, he'd be perfectly capable of doing so. They are weapons. They do what they are told. Just as a gun will fire a bullet when you pull the trigger, so will a drone destroy a target if you give the order. The problem would come in if you tried to subvert a number of drones into ignoring the normal chain of command to obey orders from some third party. This can be squashed easily through cryptography. Adding in a back door through that would be very easy to find and eliminate unless the whole programming team was on the take, in which case you've got a serious problem anyway. What if I bribed one of the members of the design team on a human-piloted fighter so that he made some of the systems purposefully faulty? Like if you hit it with a certain amount of a specific type of radiation, all the electronics shut down and the fuel tanks are dumped?

The argument cuts both ways, and the only way you can make it support your position is by arbitrarily making the drone designers idiots and traitors, while not applying the same limitations to your own position.

The fact of the matter is drones are vunerable.

No, they are not. Proper security measures (based on things we can easily do today, mind you) and a competant understanding of programming would prevent the absurd "vulnerabilities" you have claimed drones would have.

The run through binary logic and are completely predictable (or not completely controlable), the enemy needs only to grab the code for a relatively small computer on a craft and plug it into a bigger land based computer and work out how to defeat the AI in a few specific scenarios.

Nonsense. Regardless, even assuming you could predict every single possible action a drone could make, how would that help in a heated battle if you are using human piloted fighters which are slower moving, slower to react, less likely to hit the target, and prone to making stupid mistakes? You'd have to have to conceed that human-piloted craft are inferior and be using drones of your own to even begin to make the fight even, and even then there's no way you could use the data effectively.

It comes down to game theory at that point. Think of the scene from The Princess Bride where the Sicilian is attempting to determine which cup the poison was in. There are, by the way, only four possible solutions to that problem. Granted, he only thought of two at the time (a common, human mistake), but the point stands. You are presented two cups. Which cup does not contain the poison? Is it cup A, cup B, both cups, or neither cup? Say a computer chose at random which cup to put the poison into, and let's say, for the sake of argument, that it would leave at least one cup free of poison. We now have three possibilities, cup A, cup B, or neither cup. Which do you choose? How can you predict which cup does not have the poison, thus ensuring your victory (and through that, your continued existance as a living person)?

The simple answer is that you can't. There is no way to give yourself a better chance of survival. Your ability to predict the actions of a computer limited to three choices is not improved by knowing how it came to the decision than before, either.

And this of course makes your programmers the weakest link in the chain; from coding errors to design flaws, to active sabotage they can do whatever they want and only other programmers can stop them. As the size of the AI increases it is going to be harder and harder to double check all your programmers work in entirety.

Oh, gee, only other programmers can stop them. And I suppose they wouldn't...why? As above, just bribe an engineer to make a subtle yet devestating design flaw in the fighter, it would produce the same effect. Only other engineers can stop him, so you run the same exact risk both ways. This does nothing to support your position.

Joe: Cruise missiles CAN'T deal with a raging battle the entirety of a cruise missile is flight control. It's a glorified autopilot, it doesn't have great FoF recognition, it doesn't react if the target you told it to nail was hostipal, etc. Likewise the missile doesn't have to pick its own target (do I kill the SAM site or the weapons factory first?), it doesn't have to deal with targets who surrender, and it doesn't have to deal with enemies equally equipped.

If you decide to surrender after the point where the trigger is pulled and before the bullet hits you, it's too late.

You are also forgetting something else: those cruise missiles need a LAUNCH PLATFORM, be it converted battleships, subs, missile frigates/destroyers, or aerial fighters. How do you deploy your missiles in space? As stand alone units? Forget about it, the enemy need only build one warship capable of blowing your missile before it hits (be it AMM, lasers, high speed mass drivers). Clustered in groups? Not good either as the enemy can just bypass the cluster, or make a run at it with an area effect missile. The reason the carrier group dominates is it is strong enough to withstand attack, it can dominate a large radius of approach. The only way to take down a carrier group is to swarm it and overpower it (especially if it has multiple carriers).

How about drone carriers? Regardless, anything that would be able to destroy drones easily would be able to destroy human-piloted fighters even better.

Now the whole system could go AI (if you trust your programmers and you have that level of computer tech), but you still do not find a better bang for your buck than fighters equipped with missiles. The missiles have all the inherent advatanges of not having a pilot. The fighters give your launch platform the maximum range with the minimum target size. The carrier group gives you the largest operational radius.

Hence AI controlled Drone Fighters. Concession accepted.

It's called combined arms. Effective use of destroyers, cruisers, carriers, fighters, and missiles will beat the crap out of any single unit strategy.

And removing human pilots from the fighters makes them even better. For the record, I don't see any reference made here to using drones as the sole unit of the military. That would be stupid. The point was that they would be better than human-piloted fighters, and that point still stands.

[NecronLord]

The AI I would propose would be based of a combination of neural nets & expert systems, which closly mimics the human brain(but is really really fast).
Yes and that is still quantifiable. Take the source code, dump it into a yet larger and faster computer and the AI is completely predictable. Neural nets do not return randomized values, they return fairly uniform values. So I take your source code, dump it into a a computer a few orders of magnitude closer, pick a few scenarios I want to win at, and then run your code in a virtual environment until I know everything your AI will ever think to do. Data can be simulated as needed.
A neural net source code bears no resembelance to the compiled and implemented systems. Each drone is differently compiled. All the source code would allow you to do is simulate a "newborn" drone. not a real help to anyone you'd have to do that with each drone. Avoiding whatever countermeasures the drones have.


There is a different been constantly using a feature and needing the feature at a critical time. Plus with the limitations of humans, modern fighters are critically hamstrung. Most of the time a fighter would be in the idle state or getting to a combat location, but once in combat it sure as hell needs those high gee turns & maxium thrust.
AtA missiles already shoot over the horizon, this trend will continue. The big thing you need to do is make your fighters have low mass with high burn ... mainly to get to combat ranges. Once you get to combat ranges you will be using optics to track the opponent ... by the time his missiles get to you, you will have point countershots in flight. No matter how fast your drone is, it is still going to be slow as hell compared to a tracking laser. Let's say it does .1c, by the time your drone (or missiles) get within range of the fighter, the fighter has deployed counter measures with similar velocities. If laser weapons become viable then your drone will get shot out of the sky ... nothing outruns a laser.
Congratulations this is an argument for missiles in what way? Dont make me de an ass and sat neutrinos

One critcal reason fighters dont outmanoeuvre fighters, is becasue of human limitations. Plus the use of proxy detontated explosives make narrow misses a moot point.
Sure but what range of engagement are you talking? Only at short ranges of enegament do the drones have a signifant lead. In most cases your biggest delay is not reaction time, but ToF. Proxy detonations work both ways, you can take out a fighter with one, but a fighter can take out your missile with one.

The era of dogfights is long gone. Today AtA combat occurs beyond the horizon, you get the enemy in radar, send a smart missile and hope his countermeasures are not sufficient. It will be the same in the future (unless laser systems become viable weapons). The limits of the human will be minimal. These are basic givens:
1. The missile is always faster than the launching platform.
2. The longest delay is ToF, not reaction time, not maneuver time ... its all about ToF.
3. Optics allows you to indentify threats with a fraction of the ToF said threats will have.

So your drone comes within 5 minutes (at maximum acceleration) from my fighter. I get optics on your drone and begin taking counter measures. Sometime 2 minutes later my countermeasures hit your drone about 2/3 -1/2 the initial engagement distance. Your drone can either defeat my counter measures, and I'm screwed, or your drone is dead. This doesn't change if it is a fighter on the other side. The fighter comes up to 5 minutes ToF, launches its missiles, I detect them, launch counter meausres against them and launch a counterstrike. He detects my attack, launches his own countermeasures. First one to get a shot through the countermeasures wins.
by the time you have finished looking at the display my hundred drones for the cost of your training and fighter and pay and pension etc. have annihalated you. The drones will make better descions than you in around a thousanth the time you do.

Your drone gets minimal advatanges, it can turn and accelerate faster ... fine but short range missiles can do both faster still. It has faster reaction times, fine but the limiting step is ToF.

and can think faster and carries more weapons to mass unit and has a better range of Counter measures.

It has the added disadvatantages of having its code being a hideous weak point. If I get the code I can run it in a larger, faster computer and run a few million (billion) scenarios and predict exactly how your AI will react 99% of the time. I can then egineer my attack forces to take advantage of this foreknowledge and ensure victory.

I've produced a hundred odd drones for that each one using a neural net system. You can defeat one drone. ONE I have a hundred per fighter you have.

Remember your AI has limited computational capacity, as dictated by the size of the drone. A military R&D complex where I take the source code to make predictions on your AI is not subject to the same limitations.

[/b]good for you. How much good is that when each drone's software is individual to it? but wait you have the source code! congratulations on your drone which knows nothing at all.

This completely ignores the extreme security risk in the programmers.

Strategic Division of labour. The extreme security risk in your programmers is worse. I've bribed them to add a piece of code to prevent your fighters deccelerating after a cetain date. You just got X metal coffins. Or to open the hatch of your fighters after a certain date. I can make many more drones per unit than you can fighters and pilots.

[NecronLord]

^ {make me be an ass and say neutrinos}

Remember the following:
The fighter is a platform to launch missiles from.

And a drone can't launch missiles

[Darksider]

If you can pack them whith enough missiles they can engage smaller enemy capital ships like corvettes or gunships and eliminate enemy scout units also if you can make an areospace fighter that can fly in atmosphere it is a safer alternative to orbital bombardment (when you want to capture resources citys comm centers ect.) they can also eliminate squads of enemy bombers that are packing heavy guns to take your capships down