Project image
)}
$93,168
pledged of $299,400pledged of $299,400 goal
2,434
backers
Funding Canceled
Funding for this project was canceled by the project creator on Nov 10 2016
$93,168
pledged of $299,400pledged of $299,400 goal
2,434
backers
Funding Canceled
Funding for this project was canceled by the project creator on Nov 10 2016

My favorite story snippets about the AI in Classic -- the first game.

Posted by Chris Park/Arcen Games (Creator)
20 likes

Chris yet again! ;)  One of the most-posted-in topics in our forums is the The AI is awesome thread, which was started by liq3 in 2009. 

Here are some of my favorite bits about AI War Classic, pulled out in chronological order.  Maybe I should have put this in a different order, because some of the best stuff is near the end, but as it stands this is a really cool progression.

(Where needed, I'll translate jargon for those not familiar with the first game, clean up grammar, shorten, added explanatory notes, etc; but the original links are all right there, too.)

Forcefields of Doom

liq3: I'm assaulting the AI's planet, got a beach head set up.  I'm building my own command station, all is going well.  Then I notice 2 Mark III Force Fields drifting my way.  I laugh. What are those puny things going to do?

They slowly make their way to my station. I have nowhere near enough forces to kill them. Now, get this: they drive over my station... AND THEN THEY BRING IN ATTACKING UNITS. I was like "OMFG". The AI provided a shield for it's small amount of units (only like 6!) to kill my comm station. I was flabbergasted. Not only that, but they were blocking the wormhole, so I couldn't even retreat!

THEN while I was trying to get out, IT SENT A WAVE AT MY HOME STATION. I can't remember, but I think it may have even killed it.  This AI is just amazing.

Bringing Home A "Friend"

ldlework: Near the end-game CBWhiz and I were consolidating our our forces and making sure all of our defenses were beefed up.  We noticed that a few AI planets were mounting massive fleets on but were otherwise docile.

We figured this wouldn't last long, and so to stifle any future problems these bulky planets might cause us while we were focusing on the enemy homeworlds, we decided to raid the planets and destroy their guard posts to prevent them from becoming any larger.

The enemy planet had about 3500 Mark III units and an assortment of others. I generated a fleet of about 2000 myself and sent them in, way-pointing the posts, then way-pointing the wormhole home. It was to be a quick raid -- and it was...

Bu the massive AI fleet followed me on my retreat -- destroying each of our planets on the way all the way to my homeworld.  I successfully fended off the remains at the very end. When my homeworld was finally cleared all I could think was a mix of "Holy fraking raptor $%$!" and "Oops!"   The AI's are vindictive heartless #$%$%s. Prod with care.

"Friends" Bringing ME Home

liq3: Alright, so some more crazy antics from the AI.  I had a ton of my own cruisers sitting near my home station, but there were 200 enemy cloaked ships hiding somewhere on my planet.

A wave of AI bombers arrives and starts heading towards my home station. The hidden enemy ships decloak, and I discovered that they were etherjets -- little guys with tractor beams, which were suddenly grabbing and running away with my cruisers! 

Thanks to some fancy trickery on my part I managed to wear the bombers down and survive!

Falling for the Feint

SonicTitan: I just got schooled by the AI for the first time - not a loss, but a huge tactical error that couldn't have been anything other than deliberate.

The AI was amassing a 300 ship cross-planetary attack.  I pulled most of my fleet back to my home planet and went on buffering defenses while I waited.  Just as the timer on the CPA went off, I saw that a base next door was being attacked by about ten ships. 

Like a nimrod, I throw my whole fleet into the wormhole, thinking that this is where the attack is coming from. I mop up the ships and that's that. Until 150 ships warp into my home planet and start shredding my defenses. 

Meanwhile my fleet is next door, as far away from the wormhole as I can get. I'm a little spooked, actually.

Using Its Own Bugs Against You

vonduus: This AI even uses its own bugs against you: I had a case with AI ships dancing in a circle -- some ships could not decide if they wanted to attack me or run away through the nearest wormhole. [NB: Bug was fixed in 2010] 

I thought: "Ha, the indecision bug" and went after a blob of ships in the other end of the star system. Whereupon the dancing ships stopped and went for my Zenith construction site. I almost lost the whole game.

So yes, kudos to Arcen for making this fine piece of seemingly sentient software.

Not Bothering To Lock All The Doors?

Mánagarmr: The AI just taught me the importance of protecting ALL wormholes, no matter how inconspicuous the planet on the other side looks. My home planet had three exits: Two fairly uninteresting Mark I planets (but one resource-rich that I wanted to take), and one Mark IV.

I put up a massive turret blob around the Mark IV wormhole and massed a fleet in front of the rich planet in preparation to take it, but left the last wormhole largely unguarded.

The AI kept sending Raptors and Minipods through the unguarded hole, waited until he had a few, and went on a harvester rampage.  I ended up having no resources because he constantly killed my harvesters. 

That last wormhole is now protected too. So guess what? He now sneaks the ships in via one of the other wormholes.

I hate this AI. And I love it. #%$$% evil thing.

Exploiting Human Emotion

cukeaddict: Latest game I have turned the difficulty down and... the AI masses 400 ships each on 3 worlds near some of my key borders, so I decide to swing by one of those garrisons after I capture some goodies first.

This results in huge losses for me and a 2-pronged counter-attack that destroys all of my science labs before I know what happened.

A couple hours have passed in that campaign and now the AI is taunting me with 600-700 ships on another border world, just waiting for my fleet to wander too far away -- or for me to attack out of fear and trigger my doom.

I think it's great that the AI can exploit human emotions since it gives them a strategic and tactical advantage when I focus on the thing that makes me mad instead of the critical objectives.

Turning Out The Lights

shortstar: A huge and complex multi-pronged battle plays out, with them flanking me and a whole bunch of other stuff.  I take about 75% losses on that fight, but weaken them enough I should be safe for a bit.

I build back up and then head out with my fleet to kick some butt.  I'm heading home, and then... what is this...  a brownout... ohhhh noooooooo, they snuck in and got my power plants... so my turrets are offline...

I did manage to save myself, and most of my homeworld, but the moral of the story is no matter what you do, the AI is stronger and faster. I thought my homeworld defense was uberstout, and it was, so the AI turned off the light switch. 

I'm still playing the same game now, and the AI has fought me tooth and nail for every inch of ground.  It seems it knows what I am thinking, and it found my weaknesses. It's probably reading this right now...

Tricks of the Trade

TechSY730: The "fun" the AI has with mobile tractor units is a great example of specially-coded tricks that go along with the emergent bits, but there are many others.

  • Cloaked AI ships stay in hold-fire mode (and thus remain cloaked) until they can attack their target or if they are revealed.
  • Cloaked ships that are trapped on your planet after a failed attack... if their target is still too well defended, and it is not safe to retreat... they'll just lurk until the time is right.
  • Ships that can pass through force fields will (correctly) not consider force fields part of the defenses of their target, and thus the AI is far more likely to "bum-rush" their target even it is well defended.

And many others. And these only deal with ships that have special abilities. The AI has really smart behaviors for ships in general.

Combos

Vihermaali: Massive attack on my homeworld.  My shields are holding and suddenly...my home command station just dies. With all the shields around it still holding.

I reload my save and this time I watch carefully for ships that can shoot through forcefields and I see... raid starships that the AI puts under shield bearers of its own and then RUSHES them towards my home command station!  The rest of the battle is basically a distraction.

Daaaamn, I didn't know AI could be this clever.

Focus and Sacrifice

Corsaer: It was really early in the game and I needed to take a MkIV planet holding an important objective that also happened to be directly in my path of conquest. I had only one MkIII ship type unlocked at the time, but the AI only had a few hundred total ships and an Ion Cannon on the target planet. 

I'm new so I wasn't sure exactly how large of a force I would need, so I decided to err on the safe side and send my entire fleet. 

As soon as my fleet finished warping through, a large group of AI ships that had been lurking on the target planet made a beeline past my fleet and into the wormhole I came through, quickly destroying my undefended planet before hopping over to assault my homeworld.

EMP Guardians

Nalgas: EMP guardians are all kinds of evil when they catch you off guard -- disabling lots of your ships as they self-destruct. Like the literal train of them that ran through multiple systems in this game a few months ago. 

I don't think I would've been able to hold that off without the Fallen Spire stuff I had captured, that was immune to EMPs. In a regular game, the AI sending EMP guardians to multiple systems in a row after already having used one to wipe out a good chunk of my fleet would be more than enough to take any remaining "defense in depth".

Unlikely Hero

Coppermantis: This made me rage. I was fighting off a huge exogalactic strikeforce, took out almost everything with a martyr unit of mine, but an enemy regenerator golem survives.  Then it goes and DESTROYS MY HOME COMMAND STATION.

Yes, I lost to a lone regenerator golem. It's not quite brilliant AI, but it was still interesting.  I was all "Oh, regen golem, what can it possibly-YOU HAVE LOST"

Keith (one of the devs): And the water boy scores the touchdown!

I'm Not Falling For That

Philo: This AI is indeed amazing. We tried to outsmart it by baiting it into attacking a weak planet of ours. It had a huge exowave coming for us, in that area, and we left almost nothing in defense directly on that planet.

But! We put two thirds of our forces and a nuke 2 hops away where the AI can't detect them, ready to jump in when they attacked the trap planet. We'd pop the nuke first, killing most of their stuff, then send in the rest to mop up.

The AI instead took the long route through its own territory, ignoring the bait, and went straight for our most heavily-guarded planet. Our nuke was useless since it would only hit our own troops there -- and our other troops were too late to get back and save the day.  

Lost again. :)

Don't Push Me, Kid

RCIX: I was messing around with my own Raid Starships way in enemy territory, sniping off Ion Cannons that were in my way of my future expansion. All of a sudden I get the message:

The AI Is Gathering Threat in response to your deep strike on Michkill

Weeeeeee, I didn't know it would do that! Thankfully I have my fleet and a turret complement parked securely, and a solid turret system on my home world.

Just Tuesday

Faulty Logic [one of the best 10/10 players of Classic]: "I have a Fallen Spire fleet! And a massive chokepoint! I am invincible!"  

Minutes later:  You have lost  

AI taunt: How stupid do you think we are?

On a completely different note, threat fleets are terrifying. Within thirty seconds of me withdrawing my own fleet to attack something of theirs, the TF is on the move. About a minute later, there is a smoking wreck where my command station was. 

They attack, target, position themselves, and even retreat with either excellent intelligence or ridiculous luck.

After Three Years Of AI Upgrades

Vihermaali: Did the AI suddenly get more intelligent? 

  • AI troops don't suicide anymore.
  • I've seen 1k+ ships just wandering around AI systems (NOT my border systems)
  • Hit-and-run attacks (destroy one system, retreat)
  • Ships from neighbor systems will come to help when I attack a system with my champion.
  • The AI doesn't just blindly pile up ships on wormholes... they retreat. 

I'm still playing with the same 7/7 AIs as always. This is awesome.

Learning On The Job

Alex Heartnet: I just found out what Neinzul Bomber Clusters and EMP Guardians do. At the same time.

After destroying the bomber cluster and seeing all those bombers pour out, I order my forces to retreat thinking that I would just wait the 4 minutes for the bombers to kill themselves through the typical Neinzul self-attrition. I had sufficient defenses. Then the EMP guardian entered my system as well and everything stopped working.

Analysis Complete

TheVampire100: Not a story, but the best part of this AI is that it's not predictable. In most games you can always tell what action of yours triggers what action in the AI, and you can start cheesing it.

In AI War, however, the AI analyzes your behavior and and realizes when you try to use the same tactic. I've never seen such a good AI in any game.

Thanks For Reading!

That was a nice change of pace from just asking you to Tell Everybody about the bloody kickstarter.  Rereading all this just makes it so that even more I can't wait to get back into this again.

-Chris

Chris Phillips, Anthony, and 18 more people like this update.

Comments

Only backers can post comments. Log In
    1. Chris Park/Arcen Games 2-time creator on

      Yeaaah... sorry about all of the updates lately. I'm trying to not be annoying about that, and will not be hitting that frequency going forward.

    2. Patric Owen Thomas on

      As a person who has never played the original. It seems like a lot of posts. I guess better more conversation then not enough.

    3. Chris Park/Arcen Games 2-time creator on

      @Mario: Good idea! I'm also going to use a scaled-down version of this as our first "promoted post" on Facebook ads. I think that's the thing that a lot of people could really connect with.

    4. Chris Park/Arcen Games 2-time creator on

      @Fenrakk: Yep, that one was an emergent thing. In a lot of cases, the AI winds up seeming to make a very distinct and intelligent decision, but it's not based on central rules or whatever; instead it's analyzing a lot of smaller data, making a series of smaller unrelated decisions, and things come together in some fantastic fashion.

      To be perfectly honest there are plenty of times it results in it doing something stupid or underwhelming, too... but we as humans in general are "programmed" (so to speak) to ignore a lot of that sort of thing. We attribute those things to just mistakes -- humans make those sorts of things as well, of course -- or a minor lapse in data crunching or whatever else. "No AI is perfect," we're trained to think.

      But the difference here between this AI and a lot of other AIs is that it has room to pull off some stunning stuff, and when that happens, our brains turn that into a narrative (because that's what human brains do). Even I find myself anthropomorphizing it and attributing emotion to it (it was "angry" about something I did in the past) or intent where there is none (certain kinds of combos).

      A huge majority of the things listed above are of the organic/emergent variety. Things like the etherjet tractors logic I programmed in explicitly. The "dancing" bug was a series of emergent behaviors colliding and causing the bad sort of emergent behavior, and both Keith and I chased aspects of that for the better part of a year. He got most of the pieces, I got a couple, and ultimately that prevented some undesirable stuff.

      Some of the logic on things like going the long way around ("I'm Not Falling For That"), or rushing through the wormhole to grab your planets after you go through ("Focus and Sacrifice") are things that I can trace back to multiple specific code rules. It's still emergent based on semi-related underlying logic, but it's easier to trace.

      Something like the EMP guardian coming with "Learning On The Job" is a mixture of luck of timing as well as it just doing raw numbers-crunching. The one in "Combos" is just absolutely pure dumb luck, but it's actually more a testament to the richness of the gameplay simulation and thus things that can happen by luck rather than any AI programming. But it still is the sort of thing that happens with regularity, and we start thinking of that as being "smart AI."

      It's not really that it's smart in that circumstance, but it sure does SEEM smart. The bit on "Exploiting Human Emotion" is probably the best example of all: that is absolute and complete anthropomorphism of simpler underlying game mechanics.

      And THAT is the coolest part about the AI in AI War, and my original realization from years ago: "it's not about what the AI does, but rather about what the AI seems to be doing." In other words, this AI is not intelligent in a true sense. It's perhaps the best magic trick in gaming AI, though. I came to realize that all gaming AI is either just systems-based or a magic trick, and I had been reading a variety of psychology texts about brain heuristics shortly before programming this, and so that was my insight that led to this stuff.

      Then after about 2010, Keith went to town and really perfected it using a lot of new small rules that he introduced, and larger overarching systems he introduced, and more.

    5. Draco18s on

      Fenrakk: Almost all of it is emergent. There's not a true neural network or anything that complex, but there's a whole lot of little rules that result in the AI doing amazing/terrifying things.

    6. Mario Malnar on

      Awsome. Love this. This should go into ai classic update patch announcement section.

    7. Fenrakk Mendelevium on

      That very first story has me really interested in what's happening under the hood there. Using a forcefield to block your retreat? Is it actually noticing where your fleet is, realizing it can trap you there and deciding to do so in order to attack somewhere else? Or is it a consequential series of smaller decisions that build up to that emergent situation?