by Harvey » Thu May 12, 2022 9:13 pm
^ Interesting.
I was thinking along similar lines in a more basic fashion about fifteen years ago. I tried to design a workable but realistic simulated 'AI' system, only on paper - no coding or math - for a game I was working on with some friends. The problem as I saw it, creating complex simulated behaviours in the game world, could be cracked by creating a stacked hierarchy of priorities/obligations deriving from needs/desires which replace each other in the hierarchy according to and arising out of the game context - with specific values weighted (using a slider for each variable) to create a range of different characteristics, behaviours and complexity and do it all procedurally.
A basic example, someone fires a gun in the game world. Any desire to find a cold drink, hot food, some shade or if male, to ogle the beautiful girl walking by would immediately be replaced by evaluation of and reaction to the new context.
The reaction might be: Duck > Was the shot fired at or near me > Yes - Find safety/cover > Is there continuing danger > Is it easy/safe/desirable to remove the source of the danger > Are others better placed/equipped/prepared > Does something else outweigh my need for safety > Am I selfish/bold/rash/cowardly/tired of life - and so on, depending upon other variables in the hierarchy.
The idea was to create characters capable of acting in complex ways independently, everything from affection to quarrel and requiring characters to perform basic functions like eat, drink, sleep, defecate, enjoy recreation or rest. If some or all of these things are denied by a particular context, this creates other weights upon the hierarchy, pushing different potential behaviours closer to the top of the stack. Once all the necessary motivations/actions are decided upon then animations and dialogues for each would be created, with considerable potential variability among those too. For another layer of complexity add in some random shifts in the hierarchy, operating within parameters set by procedurally established characteristics like race, culture, attractiveness, metabolism, education, body weight, body type, tolerances, interests. And so and so on. The system was much more elaborate, initially difficult and labour intensive to build but simple to use and endlessly scalable, I can't remember now. I've no idea whether it would have worked though it went down quite well with my colleagues.
Anway, it sounds intriguingly similar to the philosophy behind Solms inital inquiry.
Sorry for the derailment!
And while we spoke of many things, fools and kings
This he said to me
"The greatest thing
You'll ever learn
Is just to love
And be loved
In return"
Eden Ahbez