A Cloud-Based Computing Framework for Artificial Intelligence Innovation in Support of Multidomain Operations

Abstract

The DoD’s artificial intelligence (AI) strategy requires the delivery of transformative and disruptive capabilities that impact the “character of the future battlefield and the pace of threats” that US forces must be prepared to handle. Candidate frameworks must also address key mission areas while enabling partnerships with the private sector, academia, and global allies. To meet these challenges, a flexible, cost-effective, and scalable computing infrastructure that incorporates cutting edge technologies and complies with stringent information assurance requirements is necessary. The DoD AI strategy mandates the agile employment of innovative AI capabilities that “rapidly and iteratively” execute experimentation with new operating concepts, and leverage lessons learned in subsequent experiments. Using cloud computing, we present a flexible approach to solve complex systems problems. Promoting “rapid experimentation” and collaboration on problems such as recursive algorithm implementation, deep learning, and inference in neural networks has enabled inherent advantages over existing computing frameworks. Leveraging the cloud to implement shared responsibility security models, serverless architectures, and high-performance virtual machines, aspects of the AI lifecycle including build, deploy, and monitor have resulted in an adaptable and scalable computing framework that is not only disruptive to the current computing paradigm but also promotes enhanced and productive collaboration.