I-Corps: Commercializing the Integration of Human and Artificial Intelligence for Large Scale Multimedia Analysis

International Computer Science Institute, Berkeley CA

Investigators

Abstract

Currently, crowdsourcing platforms are only being used to perform tasks that are easy for humans. Researchers have developed methods for using these systems to do tasks that are difficult for humans. They have developed methods to accomplish difficult tasks and with these techniques can produce solutions that are more accurate and efficient than any currently present in the marketplace. This method is a hybrid approach that combines artificial intelligence systems with low-cost crowdsourced labor enabling a flexible approach to the end user that enables them to analyze a wide variety of events with high accuracy while still achieving the time and cost savings associated with artificial intelligence-based systems. Multimedia content is a major part of people's ever day lives, and the ability to understand the data that is being acquired at a rapid pace will have a huge impact on society. Researchers aim to aid this by developing new techniques for the processing of large scale multimedia databases, creating methods for using existing crowdsourcing tools in a significantly more efficient and accurate way to produce high quality results, based on a n hybrid system of machine learning and human intelligence. This hybrid system could have an great public impact, leveraging the advantages of machine learning and human annotation simultaneously (having the machine learn from human crowdsourcers) providing a high accuracy solution at a lower cost than is currently possible.