Thinking about AI’s trouble with simple binary logic. It shouldn’t happen in a computer, since computers ARE simple binary logic all the way through.
But real life is mostly analog, mostly positive or negative CHANGES from a center. There have been attempts to capture this in mechanical logic, but most don’t really work.
In college I did a senior paper on Lukasiewicz’s three-value logic. True vs False vs Indeterminate or N/A. This setup works pretty well for most business or legal or scientific decisions. Holdem, foldem, walk away. Guilty, innocent, mistrial.
At the same time I was taking a required art course in order to graduate fast after too many years of part-time college. I painted a colorful truth table of the Lukasiewicz system. The two curves are p and q for the usual names of two propositions. The four intersections, not visible after 45 years of fading, are labeled C A K E for the four main symbolic functions. The colors in each intersection represent the truth values of the result.
Most ordinary decisions are analog and vector, modeled better by Newton than by Boole or Lukasiewicz. Turn left, turn right, go straight. Speed up, slow down, steady speed. Heat up, cool down, leave same. Eat more, eat less, steady diet.
The steady state takes less mental and physical energy. Every kind of change or acceleration, whether left or right or up or down, takes more energy.
Our internal map is polar, not rectangular. The internal panorama on the hippocampus is purely vector, direction and distance from here. There’s no absolute baseline, no permanent (0,0) origin.
Clamping decisions into two-state all or nothing, and clamping maps into rectangular form, are parts of the Endarkenment’s distortion of reality. Before then, we were happy with analog decisions, constantly changing to reflect changing circumstances, measuring from right here and now instead of a predetermined constant zero point. Land descriptions were also vector, using the metes and bounds system.
Polistra's Mill 