Hey, I've played Build 14 and found that, even on Death mode, it's too easy to win. The firefights ARE great and all, but they never defend their base. You can basically just come in and kill the brain, no problem. It's far too easy and I wish there was more challenge.

Data, will the AI start actually considering the safety of their base in the next build? They fight pretty well but never prioritize, such as getting gold (which isn't a problem for them but should be) and guarding their base/brain. Please can this be fixed!

AI is extremely hard to code, asking for teh AI to need gold is rediculous.

Sorry but I don't agree with you, Metroid--one, AI is EXTREMELY HARD TO CODE! As the above just stated. Even in something as simplistic as Dream Maker (hi Daman. :D) it's complex and requires far too much beta testing.

On top of that there are worse AI problems, like those damn skeletons STILL TRYING TO TWIST AND TURN THEIR WAY INTO MY BASE! I mean they can't DO it any more, but they try. (Last round I died to a green clone with an AK47 though. That wasn't so bad. But the round before that, I won--and those damned skeletons were STILL right on the concrete above my brain, twisting and turning.)

I wonder if the AI has awareness of it's environment or if it just plots x, y movement towards the brain.

Mabye the digging problem can be solved by setting it up so that after the AI repeats the same insructions a certain number of times then the AI does something else like walk in direction for awhile like our own patrol option.

Maybe setting a pathfinder once it reaches base

Like if he enters this entrance or section it will follow this corrosponding path to the brain room
Kay blue is an area which assuming all the computer will travel through
Once it enters that sector the computer will follow the path lay out by the mapmaker if possible( The green arrow lines ) to the brain room

It might be hard to code but heck i think its easier to recode the ai

Of course the there must be some color type for some specific order
Like

green - towards left
yellow - towards right
blue - jetpack up
red - land safetly down

The AI already has a program the tells it to find the brain. Data could just add a similar program that tells it to go to a place on the map (for instance in front of the base) and stand there.

Small ares on the map could be set as guard posts for each character, so that when you told a character to "guard the base", it would move to one of thes that wasn't occupied by another actor.

Whatever the case, I'm sure we can expect a competant AI in the future. Despite being fairly simple in its behaviors at the moment, it is a pretty solid base upon which a computerized terror could - and almost surely will - be built. Coding specifics into the AI is probably best left until the engine and the actual game irself are further in development.

One thing I noticed when playing the new invasion map is that when you put guys on patrol, they'll walk back and forth, but they'll totally ignore drops in the terrain.

I wouldn't think it'd be too hard to program them to detect drops like that, and make them turn around, so they don't lemming off the sides of cliffs.

lol yea you have to keep in mind that its still not even really a beta Data just released build 14 something he's been working on for a while now I think he needs a little bit of a break before working out new bugs and I'm pretty sure he knows about some of these things.

Despite AI being extremely hard to code, it has to be improved at some point. What I'm saying is that AI has to be improved at some point. As it is, there's no challenge in winning. The challenge is in holding your base, which is great that they can actually overwhelm your defenses if you're not careful. But if you want to actually go for the enemy brain, it's too easy and 9 out of 10 times you'll kill it.

As for the AI needing gold, it depends in that case. The thing is, without the need of gold they get an advantage. Sure right now you can win easily, but once the AI is improved (I say once because it has to be before this can be a challenging game) it will be unfair towards players. In RTS games the computers need money. In first person shooters they need to get ammo, power ups, etc. like we do. In competitive puzzle games they'll have the same advantages and disadvantages. When you compete in a skirmish with two human players, you both need gold. All the AI is is basically a computer which plays a team in skirmish, instead of a human player. That means that they should have the same advantages and disadvantages.

Of course, until AI is improved then they shouldn't need gold cause it would make it even easier to win.

I still think that Build 14 is a great improvement. I notice that it doesn't have compatibility with Vista's Aero but I haven't gotten any of the errors that other people are reporting (such as non-existent materials) and it's worked as good as previous builds. I'm just saying that I think AI should be a priority for improvement.

I agree with what you are saying. I think Data will have to wait until he has the engine 100 complete though, before he tries to complete the AI. He will probably even wait until he is working on the actual game before he really codes it because the AI will have to work for campaign mode (Which will most likely mean he'll need a few special AIs for bosses if they exist and for enemies to act in special ways on specific levels.)

I just don't think he'll be doing any real reworking or amazingly large changes to the AI anytime soon. I could still be wrong though. It'll happen eventually.

Lemming Command! Wheeeeee!
Sorry, just felt appropriate.

It is actually really simple to fix, something like "If X,Y = Z seconds, then move."

So if they are in one spot for so many seconds, they will go a different direction.

Ok, for the people who say AI is extremeley hard, I have to disagree, I'm somewhat of a 2D game programmer myself, and I've been working on some AI for my game, getting it to detect drops or obsticles and calclate the velocity and angle of a jump to get over them was the easiest of my challenges, and if Data added just that, it would make the game alot more challenging to play.

I usualy play on 'Prom Grounds' seeing as it and 'Grass Duel' are the only maps that work, and on 'Grass Duel' the enemy's base is halfway underground. My main strategy is to use my first rocket to block off the front enterance to my base, from here I place a sentry atop the hill which they can use to get in from the top, after that ther is no way I can lose, even on death,
because the enemies will midlessly run into my blocked entrance, as if there's no blockage there, eventualy the manage to make a hill of bodies and get over the rocket onto the lower ledge, but from here they make no progress, and I can sit there as long as I want.
The enemies coming from the hill are blasted by the sentry before they can get a shot off, and the enemies coming from the other side run mindlessly into yet another obsticle; Why give them a jetpack if they never use it?

-----

Onto my next point, enemy base defence. This is very simple, there's already a 'Sentry' AI mode for actors on the human team, apply the same thing to enemies, tell them to always keep guys in a few set potitions around their base, and presto: Harder to win.

-----

Lasty, I would just like to say that I'm not bugging Data to fix the AI, as I know that there's no point in doing it untill the final game, I'm just proving a point that it isn't as dificult as you make it out to be.

The problem with the set positions and pathfinder propositions, this would be somehow on a map, no?

Well, ah, then every custom map needs to do that as well.

On prom grounds/grass duel i simply use the starting rocket, by moving it right bit by bit, to block off the route to the brain. The AI is too stupid to dig through it to get to the brain.

Then i just buy rocket and fly them really fast for cool crashes