A path to AI

Abstract

To build a safe system that would replicate and perhaps transcend human-level intelligence, a system composed of three basic modules: objective, agent, and perception is proposed for development. An objective module would make sure the system acts in humanity's interest, not against them and would have two components: a network of machine learning agents and a cryptographic ledger. The network of agents would benefit from a form of semiotics, known as deconstruction to address the problem of value alignment. A decentralized cryptographic ledger to propose a mechanism to mitigate the existential threat such a system would pose to humanity. To build the agent module that would act intelligently on the world, the concept of Dyna would benefit from a treatise in sociology to build up the theoretical framework needed for the missing link of artificial general intelligence - the world simulator. A perception module that would estimate the state of the world, would benefit from existing deep learning algorithms enhanced by a new paradigm in hardware design - a quantum computer. This paper describes a way of how such a system could be built, analyzing the current state of the art and providing rather alternative directions for research than concrete, industry-ready solutions.