Will AI be designed and maintained to be competitive?

One problem with TA/SC AI was that they always seemed to be too limited in their ability/ambition and so the only way to scale them up to stay competitive was to allow them to cheat. Knowing that Sorian is only board as well as others, does this mean that AI is going to be a much bigger investment?

Perhaps like most things about PA, this will be something Uber expose for the modding community to take forward, and I can't think of a better person to have "on the inside" than Sorian if this is the plan.

Personally though I think of the hundreds of hours I played TA/SC, about 99% was against AI either skirmish or co-op. Most of my friends are in a similar position, so I'm just hoping that Uber take AI more seriously than TA/SC ever did.

Sorian is working on the AI.

As in is currently working on the AI. I see the occassional Facebook status from him where he gloats about all the fun he's been having making PA's AI do new things.

Just to be clear, I am not criticising the current Alpha AI, probably posted in the wrong forum. Nor am I questioning Sorian's ability. I'm more interested in Uber's stance and for a lack of a better word, budget, they are willing to throw at AI.

https://twitter.com/merobbins/status/364574378131849217

"Managed to get the AI building on metal spots today. Nobody can complain that AI cheats anymore." -- Sorian, 7:31 PM - 5 Aug 13

I am already sure by Ubers design and update style, combined with sorian, should make a good initial ai, and regularly balance it along with actual unit power.

They also claim to aim to do just that, go look at initial kickstarter questions.

Loved the hashtag #theaistillcheats!

One problem I've noticed about AI in most RTS' is that the AI has no will to live. If their base is being wiped out and their best bet is to run and rebuild elsewhere, they'd rather just sit there and die. I really hope uber makes their AI want to live.

If the AI was smart enough to understand self-preservation, wouldn't that make it harder to turn off?

The slippery slope to Skynet is what awaits those that meddle in such affairs.


---

seriously though, self preservation seems like a very strange concept for a computer to wrap its transistors around.

That's because the transistors don't move

Also, self-preservation is easy for a computer to understand. The AI's goal is to win, and if dying gets in the way of that goal, then it's bad and should be avoided. Anyway, we're not talking real AI in a game like this, we're talking more about adaptive algorithms... AI implies that it's actually creating the ideas on it's own.

Ok, then it's hard to code in those algorithms. I can tell you that as fact because if it were easy it would have already been standard.

I've actually had ai in SC2 cross the map on me twice in a game even tho I was damn sure to guard all exits, flank, and all those neat tactics. And that's with the ai set to easy :shock:

I expect Sorian to do better :ugeek: :mrgreen:

This shouldn't be an issue if you are using a map of influence, right?

The decay mechanic should ensure that uneconomic standpoints (like a lost base) are given up if they can't be hold.

Although you might need to modify the expansion behavior over the default implementations as dense, dedicated bases tend to be much more efficient than the wide spread, monolithic presence which AIs based on such maps usually produce.

That's also something which always bothered me with the AIs in SupCom and SupCom 2, the AI would never consider building a real, dedicated secondary base. You could always rely on the AI building a mostly monolithic base with unprotected outposts consisting solely of unprotected mass extractors and even the main base had only soft borders. Which in return meant, that you only had to set up artillery vs the starting position as it did render the AI powerless.

If a secondary base can be established in a remote location, DO IT. It's not like the principles of limited multitasking capabilities would apply to the AI, even less so with PA and shared armies on the player side. The goal is not to imitate real players, but to push the game towards the limits. The AI has still enough ground to loose, when it comes towards the concept of interplanetary expansion.

This is most likely to come naturally when you take into account having to planet hop anyways. Or at least I imagine that would be the case.

Ahh, I love to hear that. (I didnt like in SC2 when I blew up 80% og the opponents mass extractors and he was totaly unafected and still pumped out fatboys.)