12/24 11/24 > > we were dissociatively on fence between two hard algorithm touching tasks, > and one appears more inhibited so trying to write a little > > we were thinking of making a smarty maze solver, like take general ai goal > and then put near simple algorithm goal. there was a space where it worked > out. > > maze solvers are classically an agent that tries moving in all directions, > tracking where it has been so as to not repeat ground. one approach is to > draw maze on a buffer like the screen in a handfull of colors, e.g. EMPTY > and WALL and fill in EXPLORING along EMPTY until hit dead end of WALL then > backtrack replacing EXPLORING with EXPLORED until find more EMPTY. very > simple algorithm, finds exit by brute force. > > but is more interesting to make a combine-parts-to-accomplish-task doer, > because of executive functioning and goal pursuing inhibitions. > > so maybe we could consider holding a GOAL of the AGENT being at the EXIT, > but the agent starts at the ENTRANCE > ideally we might need RULES of the maze system so that a path or sequence > of STEPS can be derived from then GOAL > these RULES could be based on discrete or rational logic to form knowledge > or probabilities or metrics > > it’s reasonable to imagine spatialness, for example, many rules, or > storing a map or list of options, but it can also make it hard to stabilize > (((sadly we realized the basic maze algorithm is itself too complex when > written :( :( it felt much simpler when imagined. (((maybe a small p—? >
but we always like the idea of iterating combinations of available steps, judging them with metrics and logic, and performing them to meet goals successfully because this is _so_ inhibited in us that we would _love_ to have an external system demonstrating that basic logical goal-meeting can succeed reliably !!! >
