The following is meant to be both a description of what an intelligence is in general, as well as how one might go about programming a robot so that it may become intelligent. This description involves two principles and their elaborations.
1. AI must be able to reproduce another AI which can repeat the same cycle as itself. In order to do so, the AI will have an algorithm which will constitute its "nature." Part of what the AI will do with this algorithm is to duplicate this algorithm by making use of things other than itself. The AI will, firstly, duplicate itself, and then, secondly, multiply. More specifically, the AI will have to identify the parts of itself which has become dysfunctional, remove those parts, replace them, and incorporate them as part of itself. Or, it must learn to assemble other parts in such a way as to let the parts internally relate to each other, thereby starting to move on its own in the way which the original AI did. The implication of this is that when the AI fails to find a part of itself, it will stop functioning and die, but then another AI can come along and assemble the remaining parts with other parts taken from other "dead" AIs, and in this way reproduce another AI.
2. When confronted with something which contradicts its cycle, AI must be able to fit this something into its cycle. The way it will do this is as follows. First, the AI will somehow recognize this something as a particular, abstract, bounded object. This recognition must not happen as a purely passive scanning of the properties of the object. Rather, it must be an act which is clearly distinguished from the reproductive cycle of the AI. In this way, the something is explicitly recognized by the AI as something which is other than itself. Based on this recognition, the AI will choose to store some of the object's properties into its own cycle, while aiming to destroy the others which cannot be so stored. Thirdly, the AI must be able to re-start a new, updated cycle of reproduction, as if it were the same reproduction as before. For example, it may continue to express the same intention, only now in different words. However, the final step for the AI to be fully intelligent is this: to recognize the contradiction between what it intends to say and what it actually says, and then to discard or delete its intention-data and to instead take up the newly acquired words as themselves the expression of its intention. In this way, the AI gets to learn something concretely.
In a nutshell, I think that the above is all that human intelligence really amounts to. Replace the word "AI" with "Concept," and this short entry becomes a summary of what Hegel means by the "Concept."
