Abstract
There has never been a time in history where technology has been so pervasive.Today, Artificial Intelligence (AI) is transforming the industry. As a long-term researchcommunity goal, achieving general human-like intelligence is one of the highest aspirationsfor AI but is still out of reach. Here, we explore AI augmentation of the knowledgeworkforce as our vision, and type theory as a foundational means to make it a reality.In our first work, Group-Assign: Type Theoretic Framework for Human AI Orchestration,we develop a framework and an associated implementation, built upon afoundational set of type theoretic framework axioms, and Group and Assign as basemethodologies for data handling and assignment of intents to associated implementations.With the correspondence of intents as types, the framework has type theoreticexpressiveness, which enables composition and conglomeration of intents to constructwork plans. The resulting work plan is imbued with type theoretic properties suchas computability, constructivism and explanability. With this, work plans are coherentand transparent while allowing for distributed contribution of intents from diverseusers, under a unified view of human and AI operations grounded in type theory.In our second work, Exemplar-based Grouping for Information Processing, we definean interactive framework between the human and AI such that, based on the human’sintuition, the AI will group contextually relevant information even of heterogeneousnature, and thereby lets the user act on related groups of information in abatched manner. At the human-individual level, this helps reduce task-switching overheadsand bridges the gap from the conundrum of multi-tasking as a means to do more,only to be counterproductive. This proposed work is also technically synergistic withGroup-Assign: Type Theoretic Framework for Human AI Orchestration, as a possibleimplementation of the Group method.In our third work, Type theoretic-Resource Description Framework (TT-RDF), wedevelop a set of TT-RDF vocabulary and a parser that translates between type theoreticexpressions and RDF-friendly format. TT-RDF bridges linked data and type theory toenable the sharing and computation of type theoretic objects across a network. WhileGroup-Assign: Type Theoretic Framework for Human AI Orchestration provides themeans to compose and conglomerate intents into type theoretic work plans, TT-RDFcontributes the synergistic ability to encapsulate these work plans into well understoodlinked data objects, that can be communicated across vast networks.Altogether, our proposed works aim to increase human AI synergy to navigate anincreasingly connected world that is rich and overwhelming with information.