[MUD-Dev] AI not worth doing in our games?

[Sasha Hart]

[Ted L. Chen]

> Mind you, I use scripting very generically, so much so that I even
> consider airplane autopilots as script-based AI's.  If you can
> make assumptions about the operation environment and how it
> changes, then scripting is a very powerful tool.

This gets at an interesting tension. Hume noted that we need some
kind of way of getting from the observation that the sun rose in the
past to get to the prediction that it will rise in the future. This
way itself might be learned by the past futures being like their
pasts, but that itself needs an assumption to get any conclusions...

You need to make assumptions about the environment & how it changes
even to learn at all. These assumptions themselves either come from
somewhere (in which case other assumptions were necessary to pump
them out of the environment) or they don't, in which case there are
even poorer guarantees that they'll be useful.

With regard to scripts, note that the abstraction effectively gives
a superset of possible scripts. There is no reason why you
necessarily have to relinquish top-down control of every kind. Doing
so reduces the parameter setting you have to do, but that's just
what top-down control is anyway. Worst case scenario, you turn off
the AI and turn on the DFA, or whatever. (Actually, worst case
scenario you get confused and have to throw away a lot of
work. Obviously I'm not advocating that anyone work on areas that
they know they will get frustrated and bored with, unless they
really want to).

But if you don't need the range of possibilities and you just need
the one script, I agree entirely. If I want to write
crossword-puzzles, I don't need good sort algorithms either. But
sort algorithms are still very useful and I'm glad that people have
worked on them.


_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev