Anyone remember Kasparov vs Deep Blue?

If you are trying to remember a GvG battle then you are thinking about the wrong game. In the mid to late 90's, IBM created an AI program that they had challenge the world's top-ranked chess player. So what's that got to do with GW?

I'm suggesting that Anet make an AI guild that will have a team of 6 or 8 bots (depending if it is going into HA or GvG). Each bot would then have to pick it's class and skills but then coordinate with the other anet bots. After programming the bots, anet would then turn them loose on PvP. After that, anet could only touch the bots during the release of new expansions or skill changes to let the bots know about the skill changes.

The bots would, through a combo of randomness, skills that depend on other skills, and past preformance of the effectiveness of skills, skill combos, individual builds, and group builds, develop a PvP team.

To make things interesting and to make sure they aren't the best guild overnight, start the bots off with a limitation of to being with all the builds they could use are current and past premade builds and only let them change one skill at a time (once they have 8 new skills they could use all old ones one match, all new ones the next, and a mix of the two the following match). Another limitation could be that all bots are limited to the first class combo they picked until the bot picks a build with 7 from one class and changes the extra one to a class it doesn't have.

Very very interesting Idea , but implementing such a feat would be a costly benture, I would love to see an outcome of such ingenuity but the Devs time are to busy for things like this IMO.

Way to complicated. Right now, AI doesn't have any way of going to a place that they can decide themselves. If you face'm in gvg, just split up.

Scripting the new AI would take several geniuses, and various gigabytes. No kidding.

This kind of AI will be first to exist in the military instead of gaming industry.

plus, Deep Blue is nothing intelligent, it is just a well-organized huge chess moves dictionary, pull out counters by the book.

^^^^Exactly. I'm going for a masters in Artificial Intelligence.

I know for sure, that A.Net cannot make their bots choose their own skills, without any input from a developer, for the simple fact, the skills don't have any meaning to the computer, yet. Coding the meaning will take....years.

They could give them 10-20 different builds (so you never know what you get), but their spike would be so well co-ordinated it'd be difficult to counter. Also, who gets favour when the bots win in HA?

I was thinking that it would be possilbe, but difficult. Take monks. Monks will usually be healing, protection, or smiting. So if the bot chooses a monk primary, it would have a weighted random chance to be any one of those things based on what the group build is... a build that is low on healing would be more likely to produce a healing monk. Then give every skill a simple descrption, healing, protection, smiting, healing+protection, etc and if the monk is a healer, healing skills would have a higher weight and thus more likely to be selected.

Of course if even that is too difficult like you guys are saying (very possible), maybe Anet could just give the bots some premade team builds, but I don't think that would quite be as fun.

i dont think anet will spend years of research , a lot of million of dollar for a thing witch will be a certain loss

seriusly ibm probably spended on deep blue project more money then anet or even ncsoft are worth