Experimental Design for Multi-task Deep Learning toward Intelligence Augmented Visual AI

Gartner's top ten strategic technologies for 2020 and 2021 include Democratization and Intelligent composable business. It predicts the democratization of AI technology and the acceleration of DT in the analog business aspect. In this study, we propose a task-based deep learning architecture for computer vision as the first step towards AGI (Artificial General Intelligence) from the perspective of narrow AI. Multi-task deep learning is composed of Guide-task model, Sub-task model, Glue code, and Integration module, and the functions of the components are defined.