Challenge for Thinkers, Designers and creative individuals.

[alexander]

or we can come up with something that hasn't been done yet and go with that?

[IDMclean]

oh no, not at all like tic-tac-toe, i was thinking more like Dwarf Fortress Wiki

 

I meant Tic-Tac-Toe as a first gen prototype of the properties of the AI that we are to develop for something like Dwarf Fortress or the Lifesim. Tic-Tac-Toe is a tightly bound reasonably understood AI problem for us to cut our teeth on. Say for instance, we use genetic algorithms to evolve an AI towards being the best Tic-Tac-Toe player. We can later use what we learned from that prototype to develop a similar AI for a more complex game such as Checkers or Chess. What we learned from developing that AI might then be applied to a real time, open world sim like Zelda. Successive iterations of the AI prototype would be applied to successive generations of ever expanding problems.

 

Out of shear curiosity, did you have an idea of how to jump straight into a more realistic problem set right away?

[alexander]

yes actually, we do it like every other team of people set to develop a game :phones:

 

create a concept around the game

define characters, local environment, the world

define the types of interactions, the concepts behind strategies, the moves, the goals, and ways of "winning" i.e. how do we define a strategy that is a more or less winning

figure out the reflexes

figure out how to store game strategy

figure out how to learn (this comes out of some of the above) from the experiences and change the strategy

we need to create a genetic algorithm for creating strategy, and ways to "rate" those strategies

 

all i got for now

 

good thing i am getting people that are familiar with game development into this. I am still waiting on Graig and Buffy they should have some neat ideas...

 

also working on collaboration space :P

[phillip1882]

It sounds a lot like the Settlers series.

yeah, i've never played it, but i posted this idea on another forum and some else pointed me to this game as well, so i'll probably buy it. i can't say what i'd do differently until i play it :phones:.

[IDMclean]

yes actually, we do it like every other team of people set to develop a game :)

 

create a concept around the game

define characters, local environment, the world

define the types of interactions, the concepts behind strategies, the moves, the goals, and ways of "winning" i.e. how do we define a strategy that is a more or less winning

figure out the reflexes

figure out how to store game strategy

figure out how to learn (this comes out of some of the above) from the experiences and change the strategy

we need to create a genetic algorithm for creating strategy, and ways to "rate" those strategies

 

all i got for now

 

good thing I am getting people that are familiar with game development into this. I am still waiting on Graig and Buffy they should have some neat ideas...

 

also working on collaboration space :)

 

While I am aware that the above is the big picture process of developing a whole complete game, it is a conceptual simplification which omits the details of actually developing from concept to product. It is often the case in video game development to chop the problem up into a number of domains of specialty and solve them asynchronously before bringing them together as greater and greater wholes approaching the final product. The AI of the game will have certain structures which can be readily tested by far simpler cases than the whole final product.

 

On that note, it would seem prudent for us to identify the kinds of AI problems we want to solve, so we may develop experiments to test potential solutions against. As I see it, the main problems to nail right away are navigation, situational awareness, and strategic awareness. Navigation and situational awareness are highly context sensitive; thus, I feel we best differ solving them until we have picked an appropriate platform from which to develop the rest of the game. We would be using Tic-Tac-Toe or a similarly tight bound test to analyze and develop the learning algorithms and rudimentary awareness.

 

Develop artist studies (prototypes) as proof of concept and use those studies to inform the development of the final program. This is simply how I conceptualize solving a problem as complex as a simulation. If you have an alternative which leads to a plan of action, I will differ to your expertise. If you already know what you're going to do, I await instruction, so I may aid you in your process.

[alexander]

in no way am i against conceptualizing and testing, lol :)