A path to AI (1712.03080v2)

Abstract: To build a safe system that would replicate and perhaps transcend human-level intelligence, three basic modules: objective, agent, and perception are proposed for development. The objective module would ensure that the system acts in humanity's interest, not against it. It would have two components: a network of machine learning agents to address the problem of value alignment and a distributed ledger to propose a mechanism to mitigate the existential threat. The agent module would further develop the Dyna concept and benefit from a treatise in sociology to build the missing link of artificial general intelligence - a world simulator. The perception module would estimate the state of the world and benefit from existing machine learning algorithms enhanced by a new paradigm in hardware design - a quantum computer. This paper describes a way in which such a system could be built, analyzing the current state of the art and providing alternative directions for research rather than concrete, industry-ready solutions.