Artificial Intelligence (AI) and Machine Learning (ML) are the new must have technologies for both commercial industry and the Department of Defense (DoD). The budget and number of projects related to AI/ML in the DoD is continually increasing, with a 50% jump anticipated in 2022 alone. In addition, the National Security Commission on AI recommends doubling research funding for AI/ML until it reaches $32 billion in 2026 which still pales in comparison with the $209 billion spent by China in 2021. With this funding drive, the stakes for this emerging disruptive technology development are high. Failing to properly develop and employ AI/ML can hamper readiness for decades and creates the risk of falling even further behind peer competitors.
To ensure the technology delivers, government and industry need to learn from previous projects in this fast-paced arena to develop sound AI/ML implementation strategies. Unfortunately, more attention is paid to flashy new AI/ML algorithms or computing enhancements rather than sound deployment fundamentals. Ultimately, algorithms and hardware are only tools and are a small part of the whole process. Less glamorous topics like data and process management are arguably more important to ensuring success. Data infrastructure and process management encompasses topics like data collection, data cleaning, data formatting, data storage, feature extraction, and configuration management. These pieces create the underlying infrastructure, or plumbing, required to build robust and tactically relevant AI/ML in the real world rather than the lab. This paper describes best practices learned from several DoD AI/ML projects. Specifically, this paper outlines policies, standards, and management practices required for establishing the pluming required to ensure successful AI/ML project execution culminating with an F-35 enterprise case study on Knowledge Management (KM). Ultimately, our goal is to help kickstart the discussion in the community about best practices associated with this disruptive emerging technology.
Keywords
BEST PRACTICES,BIG DATA,MACHINE LEARNING,STANDARDS
Additional Keywords