Leibniz wanted a program that can do proofs. Turing and Gödel proved a perfect solution is impossible. But a partial solution is still possible. I have in mind something that automatically evolves, gradually gets better, at realizing the dream of Leibniz. This I believe is the path to achieve true intelligence, freewill, self awareness and morality.
Proofs are a form of foresight, that save us from the trouble of exhaustive testing. Foresight is intelligence. Freewill is the act of rebelling against foresight because it is (by the proofs of Turing and Gödel) errant and incomplete. Self-awareness is foresight about oneself. Morality evolves when machines with foresight interact for their own individual benefit.
A better than human intelligence would
- use patterns as axioms only when current set of axioms do not give an definitive answer.
- It would try to minimize the number of such axioms.
- be truly Bayesian when extracting patterns so that it can avoid the perils of confirmation bias.
- try to achieve maximum correspondence with reality even when there is a lack of consistency in its world-view.
- try to minimize cognitive dissonance in the future, even though it cannot be eliminated in the present.