AI: You can’t see it in a screenshot. But it matters

but... but it's gold :3 so you can release it on Impulse for beta players ~~~ 

The AI is what bring the value six months to a year down the road.

Really, GC2 has been the only game I can think of in my lifetime where I felt the AI at balanced settings was a legitimate opponent.  Hopefully Elemental brings more of the same eventually.

A lot of games rely on multiplayer for challenging opponents, but for TBS finding human opponents can be such a challenge that the AI does need to be up to scratch.  I wish I was AI coding right now too...

It's just too bad it's so hard to get AI to do things that are "technically legal" like players do.  An AI that could start legitimately out-maneuvering me would be a wonderful thing to behold.

Can you provide any insight into the sort of things we can see from the AI? Do they have anything special up their sleeves that we may not have seen before?

Thanks for all your work Brad. I've been following Elemental for awhile now, and finally took the plunge on the Limited Edition. I'm looking forward to the release of all your hard work in 10 short days. Bravo sir! Keep that AI code coming

tbh thats multiplayer, but ok

thats impossible

that would be a real I not an AI

the problem is humans LEARN

there is no way a static ai can be competive in the long run cause ppl play more and learn, while ai doesnt (unless it get constantly upgraded including brand new strategies etc)

Remember there is such a thing as "Machine Learning". Not quite up there with human learning yet but still .

..as I heard the AI in Galciv2 - ToA was excellent, so I have high hopes with regard to the quality of the EWoM AI...in Froggie we trust! :)

Great. The lack of competition from the ai in the beta was the most painful thing in said beta.

Code it up to the point it can kick ass, Brad.

Some games just give the AI extra resources compared to the player.  I hope folks are talking about the way the AI uses resources, not just extra food, gold, etc.

I am not an AI programmer, nor do I play one on tv...but I wouldn't think it would be impossible for the AI to initially have different strategies be the same weight and then as it wins and loses games to modify those weights and write them to an xml file. Not the core xml file for the AI, but similar to a modded file. Basically as it wins and loses it mods itself toward successful strategies and away from less sucessful strategies. Then, when players sign on to the Reactor they have the option of uploading that file back to Stardock...where it gets synthesized with others and worked into the next update. Think of it as "Would you like to make your experience better by allowing us to collect non-identifying usage data, blah blah..."

And thus Skynet is born.

And thus Skynet Microsoft is born.

Mods, actually

tbh it's the game itself being good - which also includes things you can't put on the back of a box.

Because if it's not good, then who will still play it in MP?

And as far as human learning - how do humans learn how to get better at games? Better build orders, better strategies, seeing what works and what doesn't.

Who's to say AI can't do that, especially with the whole idea of using Reactor and having the AI pull on what players do? Who's to say the AI will just be this static thing that can never ever ever ever change it's strategies? Even old time AIs can change their strategies if they are built half decent. Simply countering what the human has for units is changing it's strategy, a lot like how human players do.

I mean, when I modded AI for Kohan 2 - I gave it builds the human players were doing and company compositions and combos good human players do.

Reactor could be doing the exact same thing - except the AI will do it for itself.

There are ways to have the AI play like a quality human player. It might never beat the top players, but that doesn't mean it can't compete with the majority of players.

Which would mean that MS would be learning .. I have my doubts

Hope for a none cheating AI too Hope it can somehow keep up with weird experiments from the modding community

I couldn't disagree with you more.  Take Starcraft 2, the AI in that game sucks compared to the original.  My friends and I still played the original up until the release of Starcraft 2.  Due to to the TERRIBLE AI we are back to playing the original.  Like another poster said I have never gotten the sheer value for dollars paid to hours played out of any other game like I had with Gal Civ 2.  Multiplayer is what companies add if they are too lazy to program decent AI.  Good AI means I can play this game in 10 years.  I very much doubt multiplayer will help a game achieve the same goal.

Well, NWN 1 is still going strong because of MP...but otherwise I would agree with you.

In my opinion, what keeps a game going is several different variables. The AI is always important for those of us who prefer a reliable (the comp can't drop out mid-game) challenge. Mods are also probably the greatest extender of playability in a given game. As for MP, that too is vital for a games longevity, however no one will care if the core game itself is terrible. As was used in a post above, NWN1 wasn't a terrible game. It was all kinds of awesome.

Also, as far as the AI's learning capability goes, there is some research into the "newer" forms of computer AI in robotics that takes its ques from the human brain being a literal "learning machine". So instead of having scripts and triggers and such for events, these new programs learn the same way the human brain does, by trial & error. These programs actually form a type of neural network in the same way humans do. Quite fascinating stuff really. Albeit all still in the early stages of development, it is advancing rapidly. Though I do seriously doubt any skynet-like program being created due to this. I'd put that thought in the same league as the LHC causing "strange matter" that would suck up the earth...But I digress, I'm quite certain Brad will come through with some awesome AI. The thing I love most about Brad's work is that the AI doesn't hate on the player him/herself. The most painful example in recent memory that I can come up with for AI that pisses me off to no end is C&C Red Alert and Tiberium War/Kain's Wrath...I can go in and play a large FFA against the computer and each time one of them gets a super weapon up, it always aims for me. Maybe I should reconsider the skynet comment..

Not everyone can, or will, or has the opportunity to play multiplayer.  What if I am on my laptop in the north woods of Minnesota with not much signal?

It is there and then, in situations where connecting to the Internet is either not practical, not possible, or not desired, that a strong AI shines. 

I appreciate the efforts put into AI, therefore...

I think AI good teacher for players, before they can show real skill in multiplayer. It good move - look how computer opponent take start, think why it so, and try repeat, or maybe improve. Maybe not immediately when you get game, but after some number of games it get you some new expirience.

So when you code AI, take in mind important role of study players

And good to see several different AI types, based on different advantageous strategies (different ways explore technology tree, different balance betwen produce units/rise economy, even maybe different strategy of build new settlements depend on different type resource, which AI want to get first). Because different number of different type AI in each game - give you big number various situations, one not like one another. So it significantly increase replayability of game, even if years past since last addon.

Rath, in Red Alert theres a toggle in one of the .ini files labeled "AI Paranoid" that is default switched to on.  When it's on, the AI's all ally against you around the 5 minute mark.  If you turn it off, the FFA's really were FFA, made it much more fun.

My favorite AI so far is the one from AI War.  It manages to do the darndest things.

I've played Elemental in multiplayer. And at the risk of hurting sales, I'll tell you that I don't find it fun.

I can think of ways it might be made to be fun in the future but it will require making it quite a bit different from the single player game.

Some things that are fun single player are just painful in MP.  

So I'll say it up front: If you're buying Elemental mainly for MP, you're going to be disappointed. It's a single player game that has multiplayer.  If you're playing it with friends on the LAN or whatever it's not bad but against strangers? No way.

What's not fun about the multiplayer? I don't buy deep grand strategy games for multiplayer anyways, I didn't even know this game had multiplayer until much later.