Ok, I'm starting a new ai project. The idea is to have a creature that interacts with it's enviroment and learns. The last one I did like this ran into some stumbling blocks(It ened up just flottering around and eating berries while pooping every where.) because I had trouble deciding what inputs to use from it's enviroments and what the possible outputs should be. In other words what makes something life like. I'm not asking it as a profound philosophical question. (thats why i posted it here and not the philosphey area) i mean what should my creature look at in his enviroment to determine what action he takes. I don't mean the actual behavior that should be emergent(i think) I mean what behaviors should he be able to choose from and what enviromental input should he have. Any suggestion welcome. This is my first real attempt at "embodied" intelligence out side of simple BEAM bots. Any ideas would be helpful. Anyway it's 12 amand i'm rambling so please reply.

Note:The creatures world is a simple 2d enviroment with various objects in it that will be determined when i write the program.

Well, you should specify some more details about the environment and what you want out of your creature.

Inputs:
Vision? Get nearby squares in direction the agent faces
Hearing? Check hearing radius for "sounds"
Internal states of the agent? Energy levels (Is there a cost of living? A cost for movement?)

Outputs:
Desired movement

The inputs you choose rely heavily on what you want out of your creature. The outputs should be minimal and checked by the world's rule system. I say all this coming from my mainly ALife background.