Kristen Smith
Scaling artificial intelligence involves developing a framework that prioritizes solving the main problem before addressing minor challenges, said Jay Meil, chief data scientist and vice president of artificial intelligence strategy and solutions at Science Applications International Corp., during an interview with Air & Space Forces Magazine on Wednesday.
The framework should be built in an extensible and scalable manner by ensuring it can accommodate more computing and storage capacity, increased functionality and expanded data sets, Meil added, noting that the architecture must be designed with anticipation that potential future applications may need additional data sets. “You want to have very robust processing pipelines and compute pipelines in order to be able to scale it organically over time,” he said.
Military leaders are exploring AI for various use cases, including autonomous aircraft, logistics and cybersecurity. However, the military faces challenges in scaling from pilot to operational programs.
According to the chief data scientist, a lack of historical data is sometimes the hardest part of scaling AI, particularly in military and defense applications, but he said generative AI can fill the gaps in sparse data sets by providing synthetically generated data.
As part of the framework, intelligence data and command and control data must be combined but with appropriate tagging to ensure users can only view the information they are permitted to access and make an AI application readily scalable, Meil explained.
