What does the AI know?

[Aquohn]

Ergh, this is probably a pretty dumb question, but...well, till know, I've never been able to figure out what the AI 'knows'. What does it take into consideration when planning things? Does it possess perfect knowledge of everything the players do/have? I mean, they don't have scouts/intelligence units like we do, so how does the AI gather intel? This has quite a major impact on tactics and strategy, right?

[Shrugging Khan]

Ooooh, good question. I never thought of that!

[LaughingThesaurus]

I remember hearing something about how the AI actually had to poke into your planets occasionally with ships, but I think this was like really really old information.

[Toranth]

I'm pretty sure the AI has perfect information about what is in any given system.  Exowaves, for example, always know where your Fabs, cities, etc, are.  The AI also always knows the accurate strength of units in each of your systems (including Cloaked units).

[Aquohn]

Hmm. That closes off quite a few interesting tactical dimensions, doesn't it? I mean, as Sun Tzu put it, "All warfare is based on deception."

[TechSY730]

We would need dev confirmation to be sure, but I think the AI always has, at a minimum, "scouted planet" intel on all planets (what stationary things you have on each of your planets, and where they are), and strength counts of any system they have a thing or adjacent too. Beyond that, it is hard to tell. For example, I don't think the AI always knows mobile unit composition and poaitioning for a planet unless they are on it, even if they know the total strength by just being adjacent. (One important exception, they know where exactly fallen spire "what you need to escort" stuff is at all times during their respective stages in the mission.)

[LaughingThesaurus]

I'd kind of like it if the AI had to actually drop a ship on a planet in order to know the strength *at the time* so it knows whether it's smart to attack or not by actually gaining the knowledge instead of just automatically. I could totally see at lower difficulty levels, the AI just flies right into your planet and then flies out if it starts dying.

[keith.lamothe]

The AI has complete vis, and there's not much to be gained in the current model from it having less than that.  It doesn't exploit that vision very much, really.  In the example given of exos targeting fabs and whatnot, bear in mind that you'd also have full knowledge of any irreplaceable structures you'd ever had a command station on, too   You wouldn't know where the enemy golems were all the time, of course, but it's kind of an edge case.

War is largely deception, yes, but an enemy that thinks it has all the information is easier to deceive.

[Aquohn]

War is largely deception, yes, but an enemy that thinks it has all the information is easier to deceive.

Er...how?

[keith.lamothe]

War is largely deception, yes, but an enemy that thinks it has all the information is easier to deceive.

Er...how?

Because that makes it easier to use the information it doesn't have against it.  Your plans, for example.

Also, it tends to make an enemy much more predictable: to predict you have to know what they know (accurate or otherwise).  If they know everything (about the terrain, force levels, etc), then that's much simpler

[Aquohn]

So the AI doesn't try to guess your plans or anything?

[keith.lamothe]

So the AI doesn't try to guess your plans or anything?

To some extent it predicts you based on where your forces are and how they compare to neighboring AI forces, etc.  It also "predicts" that destroying that advanced factory would really, really annoy you, etc

[Aquohn]

Just another little thought: what if we had some kind of hack that affects the information the AI has about the players' forces? Might be interesting.

[Toranth]

The AI has complete vis, and there's not much to be gained in the current model from it having less than that.  It doesn't exploit that vision very much, really.  In the example given of exos targeting fabs and whatnot, bear in mind that you'd also have full knowledge of any irreplaceable structures you'd ever had a command station on, too

Indeed, it's no worse than what the Human player can get.  Hardly unfair.  The only minor complaint I've ever had was that cloaking things doesn't remove them from the AI's strength calculations.
I'd love to be able to ambush the AI that way someday.  Of course, I'd expect the AI to do the same right back at me...

[keith.lamothe]

One of the hack types I'm hoping to add is one that makes the target planet look like it has zero human military presence (but irreplaceables still register as targets) for maybe 5 minutes.  I may let that one be useable on human planets too, we'll see.

The two uses that come to mind are:
1) Clearing out a ton of AI threat by tricking them into your chokepoint (though you'll be subjecting your chokepoint to hack spawns before the actual trigger of this, which may or may not be contraindicated)
2) Getting the AI special forces (and threat, to some extent) to ignore one of your attacks long enough for you to accomplish your objective and get out.