Artificial Understanding
This thesis explores the processes an artificial agent needs to understand its environment. It extends on the Apperception Engine and intensively applies insights from Kant’s Critique of Pure Reason. Techniques from logic programming, topology, graph theory and several other disciplines are harnessed to bring these two frameworks into further correspondence with one another. The result consists of two computational systems. The first is a direct extension of the Apperception Engine that uses geometric logic to express Kant’s functions of judgement. The second is a Figurative Apperception Engine that implements Kant’s spatio-temporal or figurative synthesis: input is taken up and combined in a unifying process that builds both space and time as qualitative structures. By applying program synthesis within a Kantian architecture, a step is made towards the development of artificial agents that are both explainable and generally competent.
by Arie Soeteman
https://msclogic.illc.uva.nl/theses/archive/publication/5319/Artificial-Understanding