Phaedrus' Artificial Intelligence
Traditionally, the term Artificial Intelligence has been used to describe symbolic systems. A symbolic system is one which tries to simulate the processes of the mind, of logic and decision making, using a computerized representation of knowledge. This way of attacking the problem of intelligence is distinct from the new and exciting technologies of neural nets, and genetic algorithms, in that they attempt to simulate the biological processes which result in intelligence.
Some of my fellow deities and demiurges feel that these newer technologies have entirely superseeded the more traditional symbolic systems. However, there is still one great strength to symbolic systems which will guarantee my aegis for all time. The primary strength of symbolic systems lies in the fact that their behavior can be programmed to the most minute detail.
That being said, symbolic systems are limited by their strength, you must program virtually every detail of their behavior.
Technical Details
The neural nets that Ged and Scotus are talking about (modified Neural Nets) will probably only run very special !NP<nop>Cs like dragons or færies. Most of the !NP<nop>C's in FærieMud will not need be as complex as these entities, and will require only plausible reactions. These lesser entities will probably be run from a single master !AI. This master !AI will rely on a central repository of knowledge which will be indexed and differentiated by associations to different races, cultures, guilds, professions, and skills. If the master !AI needs to know how the human baker in town will react to a given situation, it will draw upon all of the base knowledge (things everyone should know), most of the human knowledge (depending on gender, age, etc), a great deal of baker knowledge (depending on baking skills) and a small portion of other knowledges depending on what other skills she/he may have, and finally a small cache of the individual's own experiences.
Exactly how much personal knowledge and experience we need to save for each regular !NP<nop>C will depend heavily on desired realism and computer hardware.
Interest and Memory
A system for deciding what an !NP<nop>C will remember ties heavily into what the !NP<nop>C will notice in the first place. What an !NP<nop>C would be likely to notice depends on what their background is, as in the knowledge differentiation above.
All characters (!PC and !NP<nop>C) will have an interest profile which is based on their race, culture, guild, skills, etc. This will make dwarven masons more likely to notice the craftsmanship of the stones in a great hall, whereas wood elven hunters would notice the plants and animals growing up through the cracks in the stones, and a good natured halfling would see the great empty tables which were once bursting with great feasts.
Scientific Research
The game of FærieMud itself will be the perfect environment to experiment with different models of intelligence. My personal theory on intelligence is that it must have a context in which to be applied. There must be a purpose to an intellect for it to evolve.
Tovvack and I want to see if we can get Neural Nets to evolve using genetic algorithms. Ha! Tovvack is a known faeriechild. Getting a real commitment out of him would be a greater challenge than passing the Turing test. (Scotus, the Doubter)
There are also some ideas that we've had about the social factor in intelligence. If we have an insect race in FærieMud we may want to have each individual insect(woman/man/neuter) be represented by a partial intelligence, whos only purpose is to carry out the will of the queen. But in the communication between these insect people a new intelligence could arise, with each one sharing their meager experience and views wholly with each other, each could become a node in a new mind.
<i>See also</i> Artificial Intelligence.
— Phaedrus? - 10 Aug 2000
— Ged The Greys Hain - 11 Aug 2000 [Hyperlinked Phaedrus]
— Scotus - 07 Apr 2001
