[Freeciv-Dev] Re: Preparations for other AIs and humans

[Raimar Falke <hawk@xxxxxxxxxxxxxxxxxxxxxxx>]

On Mon, Jan 15, 2001 at 02:28:36AM -0600, Jesdyn Flamestrike wrote:
> On Mon, Jan 15, 2001 at 08:52:04AM +0100, Raimar Falke wrote:
> 
> > But you need some kind of history. If for example three settlers got
> > killed while working at an important square the ai shoudl detect this
> > and should not send the 4th settler to this square.
> 
> You're trying to make it learn, but that sounds either unworkably complex or
> too easy to manipulate. I mean, if I know what
> TRY_TRY_AGAIN_BUT_ONLY_UNTIL_THIS_MANY_TIMES is set to...
> 
> The AI should stick by its guns -- if it thinks somewhere is a good place to
> settle, then by-God settle there. Of course, it should re-evaluate the
> situation each time... if there's a large army there, it should decide it
> doesn't want to settle. Fortify, maybe...
> 
> A discrete plan may be too much to hope for. Instead, simply design so that
> the AI's next step flows from what it's done. The AI should, with every new
> settler, figure out what it wants to do with it. This will at worst make it
> mechanistic; at best, tenacious and determined.

Agreed. However on the charter for a new (client side) ai also stands
"should have long term goals". I don't know how hard it is to code it.

> [This does keep the AI from using Insta-Wonders. The individual advisor
> would keep track of that... and considering the vastly different needs of
> each advisor, maybe you're better of HAVING five or six different state
> files, if you really want to save state at all.]

This infrastructure is on the way.

        Raimar

-- 
 email: rf13@xxxxxxxxxxxxxxxxx
 "We just typed make..."
    -- Stephen Lambrigh, Director of Server Product Marketing at Informix,
                         about porting their Database to Linux