I'm wondering, Kavik_Kang, if your AI knowledge is based on trying to make a proper "wingman" AI in a military sense. Because that is not what SCO needs or wants. Step away from the military picture of this, and instead imagine 3 friends playing super-melee against 3 other people. Coordinated flight in formation? Nope. Coordinated firing patterns? Nope. Just 3 kids having fun and blasting the other guys' ships out of the sky (and possibly doing their moms).

Obviously there is a single-player AI; which already clears up most of the challenges (navigation, using abilities, knowing "your" ship abilities and when they apply, knowing "enemy" ship abilities and when they apply and how to avoid, etc). Those AI aspects are all difficult but obviously need to be (and are) addressed because a single-player AI is necessary.

A multi-player AI adds additional layers of complexity, but the only important one for SCO is acting in concert. The inputs and outputs of SCO are relatively simple and few. I would expect a (recurrent) neural network develop a good grasp and play at a challenging level given enough time to train. Any flaws in the AI can then be balanced out by making the game more asymmetrical (3v5, 3v7, etc).

Solving the AI for the real-world is a completely different proposition than solving it for the game.

There is a big difference between a wingman, or allied, AI and enemy AI.  Enemy AI in this is fairly easy.  Like in the original Star Control where the AI largely consisted of just going evasive whenever the players guns were pointing at it.  It's when you have a "glorious vision" of allied ship fighting along side of you, and having that make any sense at all, that things become next to impossible.  Even the enemy AI in these games is always going to be some kind of trick.

I have a very good AI for this top down combat that I had come up with for "AI Encounters" in Pirate Dawn that is based on a concept from SFB called the Oblique Option point.  I use 4 different Option Points, combined with the kind of AI such games already have, to make the opponent "seem more human"... but in the end it's still just an slightly more advanced trick that will make the enemy AI seem more intelligent.

Then, in 3D, I have an all out game design magic trick I use to create 3D starship simulators.  In that... what the player thinks is happening is not actually happening, so I can do anything I want.  That is limited to 1v1 though, and it will never be possible to make it work more than 1v1.  

So what I think Stardock should avoid attempting is having any kind of AI wingmen, AI that is on the player's side.  You might do this in a limited way, a ship in the distance you come to rescue or something, but trying to do "wingmen" will just detract from the game.  It could be done, but it's never been done well before and would be a lot of work to do right.  You would need separate AIs, combat and formation, and the AI would need to know when to use each, which would be the hard part.  Like many things, it's not that it can't be done... it's that it isn't worth it to anyone to do it.  It's too much time and work for what you get out of it.

If it's future dlc or content for an expansion, what does it matter if they attempt it as long as it doesn't impede the original release?

Look, I agree that for a good while still we'll have the highest level of challenge when we face smart primates rather than an AI. And that's okay. I don't expect the 1v1 AI to fool me more than once. I don't expect a 3v3 AI to redefine the way I look at computer games.

But I do play Dota2 frequently, and I regularly play bot matches (5v5, with me being the only human). The AI is predictable. The AI can respond with super-human speed to situations (thus fails to be human-like). The AI can be abused. But I enjoy it regardless. I get to practice combos in quasi-real situations, I can experiment with builds and heroes. I find it fun. The vast majority of my games are with-and-against humans, but that doesn't detract from my enjoyment.

I think that the AI proposition for Dota2 is higher than for StarControl super-melee (although I think they are both in the same league of complexity). That is to say, if SCO can have a multi-ship AI on the level of Dota2, then my needs are satisfied.

Like Volu said, having something like that in limited places just to have it might be cool.  I just wouldn't want the experience ruined by the game trying to have multiple satellite ships come with you to fights.  It is a tempting thing considering the game, with half-a-dozen or so ships on a mothership.  But it won't work well no matter what they try to do in the time they have to do it.  Having a couple ships locked in place with you, that are obviously not separate ships but really just one ship wouldn't be an improvement on one ship.  And trying to have them "escort" you through some AI will just have them doing a lot of wiggling, not firing when they should, and generally appearing to be deaf, dumb, and blind... because they are.

It's really mostly a matter of immersion.  With necessarily stupid AI wingmen you are constantly reminded that "this is broken" as they will constantly do things that remind you of that.  Without them... you are in control of the only ship on your side, it only does what you do so it is all "correct" in your mind.  That is really the point I have been trying to make.  

hmmm... have you considered that immersion is a layered concept? Yes, poor AI breaks sense of immersion. But from a different angle: if you know you have 10 ships but you're only using 1 to fight... that also breaks immersion.

Personally I'm against wingman AI. That is like picturing super-melee like World War 2 naval battles. Fleets in formation doing coordinated bombardment. That's not the kind of battle I'm seeking. I picture ace pilots pitted against each other, more akin to World War 2 aerial battles. Formation flight is really just intended to take you from A to B as a group. Once combat is engaged things become a freak fest, with desperate maneuvers and slick skill shots.

Specifically with Star Control, it is a simple game with a simple "Champion" based combat system.  So 1v1 is especially appropriate for SC in the main single player game.  That's why I've suggested "system defenses" like minefields, defense satelites paired with minefields, and then both of those possibly even combined with bases... which all cover each other in a layered defense.  And then the base itself is a layered defense within that layered defense.  And you can add in a single ship with AI to protect the base, which gives in a goal in life that is easy to meet... defend a stationary target.  So it is easy to create that "wingman AI.  And this adds the variety that the original SC didn't have, in a way that not only works but is particularly fun.  "Peeling the onion" is very fun, and I actually created quite a potential onion for SC that can just keep getting more difficult to peel every time you encounter it.

Very early there are minefields.  Then there are defsats within minefields, covering them ("Overwatch"... for you X-Com players, haha).  Then there is a "base station" at the center of the minefield and it is just like a single big ship.  Then there are defsats in the minefield around the base station.  Then a battlestation inside a minefield, it is like 3 big ships and is it's own layered defense all in itself.  Then there are def sats in that minefield around the battlestation.  And along the way in any of these, a single ship may or may not be there with a "defend this location" AI.  So the player gradually learns how to strip these defenses until by the end you have this overwhelming 5 layer defense of a defending ship (1), minefield (2), defsats (3), battlestation (4 & 5) that the player has learned how to take down little-by-little over the course of the game... so they can pretty easily do it even though it is actually a pretty hard thing to do.

This gets monotonous if you make them do it too often, but every once in a while will add a lot to the game.

Yea, that's a little bit akin to the final boss battle in SC2. And I do hope that that type of thing will be more common in SCO.

^ It was never confirmed to be in in the first place. Just sayin'.