Frogboy Frogboy

Stupid AI tricks

Stupid AI tricks

image With Elemental: War of Magic v1.1 BETA now out, you can see the direction we’re heading with the game.  Global mana has made a huge positive impact on game play and having your population being put to work (literally) not only provides a great game mechanic but connects the player to the game more closely.

As you can imagine, such significant game play changes require a lot of new AI coding to make it effective.  With Kael joining us to relieve me as the Producer on Elemental, I have more time for working on the AI (I am still handling the role of senior lead developer on the game which is a position I’m looking to fill – hint hint). 

A 5 step process

The process I use to make good AI works roughly as follows:

Step #1: Make the AI multithreaded.

This is the most important thing because it allows the developers to focus on making the AI effective without worrying about time so much.  Games with single-threaded AI have painful late games where you can wait a long time between turns. 

This part was accomplished in the v1.1 beta. It still has to be debugged (90% of crashes are due to this) since you have to make sure the AI isn’t monkeying around with main game data.

Step #2: Play the game at a basic level

This is pretty obvious. Can the computer players do everything that the human can do? This is something that I’m working on this week.  This can be effectively described as playing the game without worrying about opposing forces.

Step #3: Play the game at a competitive level

This is where it plays the game knowing that there are others trying to thwart it or destroy it.

Step #4: Play the game at the meta-game level

This is where it gets good at handling diplomacy, its dynasty, trade, and maximizing resources.

Step #5: Eliminate stupid AI tricks

Stupid AI tricks are simply little holes or little exploits that really need to be explicitly looked for and eliminated.

So for those of you who have v1.1, can you list THREE (no more, no fewer) things that you see the AI do that are either stupid or result in its destruction and what you would do if you were them?

113,177 views 54 replies
Reply #51 Top

Don't know if this is the best place to put this, but...

 

AI EXPLOIT: My sovereign, equipped with a bow + blink ability, can SOLO the entire map. All I do is warp around the tactical field reigning hell, and the AI has no defense against this yet. (Eg. shoot enemies until they get close, then blink to the other side of the map before they can touch me. Loop till enemy pwned. Works every time.)

 

Another thing: I would like to see the AI's sovereign teleport in and snipe out mine when I dare to venture into their territory. As it stands, it doesn't feel like I'm risking anything by putting my sovereign on the front-line (even alone...) however; if they engaged me with their sovereign there would be an epic duel with advantage to the defender (since I have more to lose in their territory.) Playing sovereign on front-lines should feel overpowered yet risky IMO. (At least until I get the escape ability, which adds incentive to research it.)

Reply #53 Top

Did anyone catch the Nova special on Watson last week about Watson, the IBM Supercomputer that plays Jeopardy?

Quite fascinating AI development!

Most of the concepts don't apply to Elemental of course (parsing sentances and such)...

 

Imagine what Elemental would look/play like with a Watson-sized supercomputer crunching away!

Jeopardy! sit-in Watson, a supercomputer, will be using DeepQA technology to procure Jeopardy! answers. It is armed with a Linux system that is powered by 10 refrigerator-sized racks of IBM POWER 750, with 15 terabytes of RAM and a 2,880 processor core.