Loroi Ship Design

[Mjolnir]

Wouldn't that require the AI drone to be in prolonged combat for it to be able to pick up those "habbits"? And it would have to tie those habbits with a specific craft.
Now in the next sortie, would it be able to recognize that same craft? Or that it's not piloted by the same pilot?

It wouldn't require a given drone to be in continuous combat...data from previous encounters from the same pilots could be taken from other drones and from warships. A system of the sort I describe could easily be extended to comparing actions with previously encountered pilots, testing for the sort of consistency that indicates another encounter with the same pilot. Drones could even mimic specific biological pilots to throw off enemy drones...such a behavior might even arise spontaneously, though it'd probably be a preprogrammed tactic.

None of what you're saying makes sense. Computers ONLY respond as their programing allows. They cannot exceed it or go beyond it. Humans, can go beyond and do make irrational and illogical decisions that a computer never could. The only things a computer controlled fighter has over a human piloted one is g tolerance and nearly instant reflexes.

Wrong. Computers can easily be programmed to perform randomized searches for novel solutions, or to come up with approximate solutions based on insufficient data. Computers running genetic algorithms and similar approaches are in fact well known for producing inexplicable, but still functional solutions to a problem. Computers can be programmed to postulate and test theories, they can even be wrong.

It's true that humans do have preferences, but guess what? So do computers. Where humans excel at though is we can be truly random and no computer can have intuition based guesses like a human can. Why? Because humans do not think logically. We make random decisions and our thought processes are definitely odd. How often have you found yourself getting on a completely different topic that had nothing to do with what you started talking about? And there is emotions too. Those alter how we do things as well. Emotions no computer can ever have.

Again, randomness is actually something humans are very bad at, while being trivial to incorporate into machines. This is simple fact, easily tested...you can't even randomly pick heads or tails.

And you still haven't given a single example of how human intuition makes for a superior pilot in this environment, rather than an impediment. Intuition's nothing but estimation and instinct...machines can do estimation, and those instincts were evolved for a completely different environment. And yes, emotions alter our behavior in predictable ways...and a decent machine pilot will probably exploit this, choosing actions to provoke human opponents. One with a good enough model of the opposing pilot might even start doing this spontaneously.

[discord]

well, to be in the middle humans can be truly random(aka bug-eye crazy) but the kind of random you want for evasive movement in space is NOT something humans are good at, for starters rather few humans have good grasp of movement in 3d, proven by games like descent....or well at least indicated.

that's the funny thing about evasive movement in space, it should be non-randomly random, it should take into account known threats and choose a path these have problem with(non-random), but NOT with a simple algorithm so it gets predictable(random) and it should preferably give the evasive target a more advantageous firing solution(again non-random) the combination is rather difficult for humans to do in the fractions of a second it has to make it, machines however CAN make it in the fractions of a second...and despite being pretty predictable be near unstoppable for humans.

it will avoid fire, it will try and get 'behind' and fire upon hostiles, but the path to do so is not certain.

[Mjolnir]

well, to be in the middle humans can be truly random(aka bug-eye crazy) but the kind of random you want for evasive movement in space is NOT something humans are good at, for starters rather few humans have good grasp of movement in 3d, proven by games like descent....or well at least indicated.

A decent example of and attempt to counter this was made in Ender's Game..."the enemy's gate is down". Forcing this orientation sets the preferred plane perpendicular to the enemy, greatly weakening preference for one particular approach. An individual might still have a tendency to go left or right more than up or down, and prefer one direction over the others, but their orientation around the "vertical" axis will be less predictable...they're less likely to act like they and the target are sitting on a planetary surface or flying through a nearly 2D layer of atmosphere.

that's the funny thing about evasive movement in space, it should be non-randomly random, it should take into account known threats and choose a path these have problem with(non-random), but NOT with a simple algorithm so it gets predictable(random) and it should preferably give the evasive target a more advantageous firing solution(again non-random) the combination is rather difficult for humans to do in the fractions of a second it has to make it, machines however CAN make it in the fractions of a second...and despite being pretty predictable be near unstoppable for humans.

The coin toss example illustrates the basic problem. Humans will tend to alternate heads and tails far more than random tosses will give. A human pilot is likely to go right after going left, and after going left and right a couple times is likely to go up or down for a couple times instead, and is likely to keep their forward acceleration relatively constant. You don't need to exactly predict their actions to greatly increase your odds of hitting them.

[discord]

mjolnir: i am mostly on your side here.

the flipside however that is the advantages of humans in that command loop is mostly orders, and interpretation of orders, 'hey that is just crazy, why should i fire on that civilian hospital?' is not something a computer would ask, nor can you just tell a computer what you want, voice recognition and understanding is rather complicated and prone to error.

optimal is probably a hybrid system, some human inside the command loop(CnC fighter thing) controlling a group of pure drones...even if the g-forces involved might make the human somewhat less than perfect at the command part, but you really want a human in the command loop, if for no other reason than to be able to pull the breaks if something fucks up, long range communication is after all prone to jamming.

[Overkill Engine]

Or worse yet, interception of comms and substitution of "less than optimal" orders.

I would hope a human group leader would be harder to "hack" via broadcast comms than an AI group leader would be.

[junk]

Or worse yet, interception of comms and substitution of "less than optimal" orders.

I would hope a human group leader would be harder to "hack" via broadcast comms than an AI group leader would be.

The AI would be as hackable as the ship itself. Keep in mind that a pilot of a spaceraft will have almost just as many computer systems on his ship.

To be honest I certainly think that humans still have a role in future warfare - in part due to the batshit crazyness we exhibit. Not random behaviour but very odd.
This could potentially make human commanders very important.

[Mjolnir]

Or worse yet, interception of comms and substitution of "less than optimal" orders.

I would hope a human group leader would be harder to "hack" via broadcast comms than an AI group leader would be.

That is really going to be essentially impossible in either case, with it being trivial to use encryption strong enough that a planet sized computer couldn't crack it in time and with it being rather clear where a transmission is coming from. It's unlikely either the human or the robotic pilot will even accept communications carrying commands from the enemy...ignoring such signals without even processing their contents would be a straightforward way to prevent exploits. And as Junk points out, the human piloted version still has computers running everything...human pilot or not, if the enemy can get into the computers, you've already lost.

[Overkill Engine]

The....security offered by encryption is proportional to the amount of tech and resources invested in it, and even then....it just slows down the process, not halts it. It's basically just a padlock on a vault...it slows a thief down, but does not stop one sufficiently determined. People trust it too much. The infrastructure for crypto even in today's military communications is immense as is, and I shudder to think what a mishap would cause in something utterly AI reliant.

Make it worth the effort of relaying an encrypted broadcast signal back to a planet sized computer for processing, and it will happen.

The best security for information is making sure the opposing force cannot even get hold of it even in encrypted form, especially by not broadcasting it via a real-time telemetry feed that can be compared to observed field behavior.

Related link: http://news.cnet.com/8301-1009_3-10417247-83.html

Granted, in this case there was a critical link left unencrypted....the scary part is with that raw data they could ( ) have done far more than watch the feeds. The real reason for this failure was that we were too confident in our tech superiority and did not properly evaluate the resourcefulness of the enemy.

I wish encryption was an end all answer to infosec, but it unfortunately it is not. Sorry....I get on a rant when people say "just encrypt the signal".

[Mayhem]

Encrypted transmissions are fine as long as the "duration for which the information must remain secure" is less than the "time required for a hostile party to decrypt the information".

When controlling drone in a combat environment the priorities are preventing the enemy from:

1. giving the drone orders
2. knowing what the drone is going to do before it does it
3. gaining additional knowledge of the precise capabilities of the drone (or other of your assets in the area).

As long as the enemy cannot decrypt the transmissions during a battle, issuing the drone with new encryption keys at each launch will take care of the first 2, and carefully design communication protocols & directional transmissions will help for the third.

[Overkill Engine]

4. Pray that the enemy never salvages a drone.
5. Design them to blow up when disabled.
6. Pray the enemy never figures out how to prematurely detonate them...
7. FFFffffffffffuuuuuuuuuu.......

Let's just say there is a reason one of the procedures for soldiers is to destroy any equipment that you can't readily carry with you when vacating a position that is about to be overrun.

[Mjolnir]

The....security offered by encryption is proportional to the amount of tech and resources invested in it, and even then....it just slows down the process, not halts it. It's basically just a padlock on a vault...it slows a thief down, but does not stop one sufficiently determined. People trust it too much. The infrastructure for crypto even in today's military communications is immense as is, and I shudder to think what a mishap would cause in something utterly AI reliant.

The planet-sized computer was just an example of an unachievably powerful computer. It's a straightforward matter to increase encryption strength to the point that a computer capable of breaking it can't be constructed with the available matter and energy in the universe. And in truth, slowing them down is enough...if it takes a year to crack the keys, by the time they're done the drone will be destroyed, recycled, or at least using a new set of keys.

This is assuming the usual approach to encryption of using a key much smaller than the encrypted message. If the drone uses a one-time pad for commands, it'll never be broken...never, ever, not with a computer the size of the universe and an eternity to work on the snooped command messages, because those messages are just random data without the pad to correlate against, and even if combined with the decoded messages, contain no information about the unused portions of the pad.

I wish encryption was an end all answer to infosec, but it unfortunately it is not. Sorry....I get on a rant when people say "just encrypt the signal". :ugeek:

You haven't supported your position. As you admit, the issue with the Predator was a signal that wasn't even encrypted.

4. Pray that the enemy never salvages a drone.
5. Design them to blow up when disabled.
6. Pray the enemy never figures out how to prematurely detonate them...
7. FFFffffffffffuuuuuuuuuu....... :lol:

Don't pray, take measures to prevent them from gaining useful information from a salvaged drone. And don't make their self destruct packages controlled via an unsecured link.

[Zakharra]

Again, randomness is actually something humans are very bad at, while being trivial to incorporate into machines. This is simple fact, easily tested...you can't even randomly pick heads or tails.

And you still haven't given a single example of how human intuition makes for a superior pilot in this environment, rather than an impediment. Intuition's nothing but estimation and instinct...machines can do estimation, and those instincts were evolved for a completely different environment. And yes, emotions alter our behavior in predictable ways...and a decent machine pilot will probably exploit this, choosing actions to provoke human opponents. One with a good enough model of the opposing pilot might even start doing this spontaneously.

We can do it better than any machine. Do we have preferences, yes, but that can be trained out to a degree and you are forgetting one very important thing. Right now no one. NO one is really trained in true 3d space fighting. The best pilots we got can do their best in a planetary atmospheric fighters. We we get into space and develop space fighters/ships, there will be people trained and those that have the gift, to truly excel at that form of flight. Just as there are those pilots now that have a gift for flying/combat, so will there be space pilots that have a gift for flying in space.

Also, in a high tech environment, humans can operate when communications are jammed. A computer controlled ship that relies on constant communications is effectively useless when those communications can be jammed. Hells. Even radar can be jammed, flares can draw off heat seekers. How would a computer be able to detect what is the real target and what's the fake ones?
Human emotions can have an unknown effect too. It's not always predictable in how it affects us. Anger, rage, hatred, can enrage someone, make them like a wild animal, but it can also make the person a LOT more focused and intent, and let them exceed their normal performance.

[Absalom]

We aren't quicker than 'any machine', we aren't as obsessively devoted to the details of the orders, we aren't as gee-tolerant, and we do a lot of things that are unnecessary for the role (like breath). Humans will be the captains and commanders, we'll be in the capital ships, the 'combat boats', and the little bitty shuttles, but unless you treat 'fighter' as a JOB (like 'skirmisher' or 'observer') instead of a mixture of job and size, humans won't be aboard. And even then we probably won't be piloting them in combat, we'll be giving the piloting computers destinations and behavioral parameters. Let the machines polish out the twitch-level details, humans SHOULD just to set the goals in these situations; the ability to delegate (and train) is the TRUE mark of a good manager.

[Mjolnir]

We can do it better than any machine.

It's unclear what you're talking about. Human pilots do have a real advantage in ability to interpret orders...but by the time it's an issue, we'll probably have speech recognition and natural language systems working well enough to take complex verbal orders without needing a full-blown AI. For just about anything else related to space combat, however, it's clear that we really can't.

Do we have preferences, yes, but that can be trained out to a degree and you are forgetting one very important thing. Right now no one. NO one is really trained in true 3d space fighting. The best pilots we got can do their best in a planetary atmospheric fighters. We we get into space and develop space fighters/ships, there will be people trained and those that have the gift, to truly excel at that form of flight. Just as there are those pilots now that have a gift for flying/combat, so will there be space pilots that have a gift for flying in space.

And none of them will ever approach the capabilities of a machine pilot, with the difference being so great that even for manned craft taking full manual control will likely be a very unusual thing to do.

Also, in a high tech environment, humans can operate when communications are jammed. A computer controlled ship that relies on constant communications is effectively useless when those communications can be jammed.

And one that doesn't, isn't. Yup, a poorly designed system is poorly designed.

Hells. Even radar can be jammed, flares can draw off heat seekers. How would a computer be able to detect what is the real target and what's the fake ones?

Same way a human would, by looking at the various sources of data and trying to pick out conflicts and find the most probable correct solution. Computers can do this faster and more reliably than humans, looking at actual data rather than a simplified representation rendered suitable for human consumption, taking more sources of data into account than any human could pay attention to, and performing more complete analysis than a human's quick estimation.

Humans are also afflicted with a variety of odd perceptual quirks that could be exploited to make things difficult to see or make a false target look more real. Machines may have similar glitches, but they'll depend on the software and hardware installed. This is more of an issue planetside, though, where exploiting it is a simple matter of a particular paint job...it'd be rather difficult to influence the pilot's instrument displays in a way that gives an advantage.

Human emotions can have an unknown effect too. It's not always predictable in how it affects us. Anger, rage, hatred, can enrage someone, make them like a wild animal, but it can also make the person a LOT more focused and intent, and let them exceed their normal performance.

Emotions can easily make people more predictable, making a pilot fixate on a particularly irritating target and make simpler, more direct maneuvers in pursuit. The power of rage won't transcend physics to give humans an advantage, and may even be quite easy for machines to exploit.

[bunnyboy]

Usually anger or any other emotion makes us stupid.

Also humans are fully cabaple to any unhumanity like machine. Hospitals, civilians, cities, airplanes, dams, neutral foreign forces, allies, etc are only only targets, if you follow orders.
Proof: Any war fought ever.

[LegioCI]

Whether to use AIs or pilots has always been one of those weird subjects for me. Objectively, I know that AI's will generally be better than a human pilot in any given combat situation 99% of the time, but my gut just knows that an AI getting all the credit for taking out the enemy flagship is disappointing.

This causes me go back and forth between been brain and gut until I remember the BOLO stories I've read, and I know that if humanity's AI's are done like BOLOs, I would be completely OK with that.

[Overkill Engine]

I wish encryption was an end all answer to infosec, but it unfortunately it is not. Sorry....I get on a rant when people say "just encrypt the signal".

You haven't supported your position. As you admit, the issue with the Predator was a signal that wasn't even encrypted.

Actually, my position is that overconfidence is what caused that breach. Someone didn't give the "primitive sand dwellers" the proper credit and we got burned for it. However, what should have happened is that ground link was encrypted at a minimum, along with additional measures to ensure the encryption gets the chance to do its job too. Relying on crypto alone is also overconfidence though.

4. Pray that the enemy never salvages a drone.
5. Design them to blow up when disabled.
6. Pray the enemy never figures out how to prematurely detonate them...
7. FFFffffffffffuuuuuuuuuu.......

Don't pray, take measures to prevent them from gaining useful information from a salvaged drone. And don't make their self destruct packages controlled via an unsecured link.

I wouldn't even make it solely controlled via crypto link. I'd just worry that I'd make the self detonation conditions so paranoid/temperamental that I'd waste a lot of resources when I could have gone with something inherently less exploitable by the enemy.

Whether to use AIs or pilots has always been one of those weird subjects for me. Objectively, I know that AI's will generally be better than a human pilot in any given combat situation 99% of the time, but my gut just knows that an AI getting all the credit for taking out the enemy flagship is disappointing.

This causes me go back and forth between been brain and gut until I remember the BOLO stories I've read, and I know that if humanity's AI's are done like BOLOs, I would be completely OK with that.

Never read that series of books, but I do worry about the probability of the AI units going Skynet on our asses approaching 1 as we keep refining AI and robotics technology.

Pseudo-related link: http://www.botjunkie.com/2009/10/20/har ... obot-bees/

These could be militarized by replacing the pollinator with some form of toxin injector. ROBOT SARIN BEES!!!! I'm reasonably sure they could carry a 0.5mg dose per "bee". Granted, not exactly a space warfare weapon, but you got to start somewhere.

[Mjolnir]

Actually, my position is that overconfidence is what caused that breach. Someone didn't give the "primitive sand dwellers" the proper credit and we got burned for it. However, what should have happened is that ground link was encrypted at a minimum, along with additional measures to ensure the encryption gets the chance to do its job too. Relying on crypto alone is also overconfidence though.

As I recall, hardware capable of encrypting/decrypting streaming video was hard to come by at the time it was being designed, and there were cost/reliability/power consumption/other issues that prevented it from being used. It definitely should have been upgraded, but the reasons it wasn't were more resistance to change than overconfidence. A fair bit of it was resistance to the very idea of drones and unwillingness to give their development and deployment the needed resources.

I wouldn't even make it solely controlled via crypto link. I'd just worry that I'd make the self detonation conditions so paranoid/temperamental that I'd waste a lot of resources when I could have gone with something inherently less exploitable by the enemy.

It's pretty straightforward. Arm the system like existing weapon systems, by distance flown or time from launch, disarm with a secure handshake on return. In between, you're freely flying in vacuum at tens to hundreds of km/s relative to the enemy, at ranges from thousands to hundreds of thousands of km. Reliable tamper protection should be extremely easy.

Never read that series of books, but I do worry about the probability of the AI units going Skynet on our asses approaching 1 as we keep refining AI and robotics technology.

The system I have in mind is a lot simpler and more controllable than anything like strong AI. The environment and problems are really ideal for machine control, vastly simpler than even aerodynamic flight, with none of the hard problems like recognizing and manipulating interacting objects in a cluttered environment, using tools, etc. Most of the tasks the human brain has been optimized for have no use in space combat.

[bunnyboy]

ROBOT SARIN BEES!!!! Granted, not exactly a space warfare weapon

Then you haven't seen genemodified cyber killer flies.

[Overkill Engine]

I could swear I've seen that comic in a Heavy Metal magazine before, it looks damn familiar.


----

As I recall, hardware capable of encrypting/decrypting streaming video was hard to come by at the time it was being designed, and there were cost/reliability/power consumption/other issues that prevented it from being used. It definitely should have been upgraded, but the reasons it wasn't were more resistance to change than overconfidence. A fair bit of it was resistance to the very idea of drones and unwillingness to give their development and deployment the needed resources.

Probably cost. I know I had equipment bulk encrypting video conferencing feeds in a facility where I worked clear back in the early 2000's, that was about the size and weight of a DSL modem.