A Data-Driven Learning System Based on Natural Intelligence for an IoT Virtual Assistant
Nicholas Dmytryk, Aris Leivadeas · 2020
The ability and functionality of today's learning systems are dependent on configuration and designed for specific use. These systems have strong dependencies on training data and remain static in capability. This translates to commercial IoT sensor management systems that are limited to a set of predefined functions. Although often labeled as or associated with artificial intelligence (AI), these systems lack the behavioral qualities associated with intelligence. Our research details a novel learning system autonomously motivated by its environment, analogously to humans. The algorithmic processes of the system produce behavioral byproducts of intelligence, resulting in an entity capable of tackling general problems of its own interest, contrary to constrained solutions. Because of the system's generic nature, the intelligence produced is limited only by its ability to sense and actuate. The research contributes a system that learns through different interfaces with the same generic algorithms-where language and image processing ability would traditionally be learned in separate modules, our system uses the same algorithms to learn ability with data from both modalities. The proposed framework is showcased through the embodiment of a next-generation intelligent IoT virtual assistant application.