Data science methods provide a means to establish analytic tradecraft, capable of managing a large amount of data, allowing for full characterization of actor behaviors, and providing valuable insights.
As data volume increases, AI/ ML plays a significant role in this "high data entropy" space, providing users with the means to combine multi-sourced datasets, with the goal of learning and identifying patterns, and develop actionable insights while assuring they follow the organization’s law and policy boundaries.
Efrain Rodriquez, data manager, U.S. Department of Defense (DoD), presented a case study on how the intelligence community (IC) is addressing these challenges by establishing innovative AI/ML governance and data management methodologies, during his Data Summit 2023 session, “Using Data Management Methodologies to Foster Development of Transformational AI/ML Tradecraft.”
The annual Data Summit conference returned to Boston, May 10-11, 2023, with pre-conference workshops on May 9.
“AI is ubiquitous, you can see it everywhere,” Rodriquez said. “We recognize there’s a culture change; we recognize there are things we can do to prepare workers for what is coming.”
The AI/ML ecosystem itself is misunderstood, he explained. Many people see the output but, not what’s behind it. It’s not rigid, the ecosystem is fluid. Governance comes into play to control the access roles, policy, and activity for AI/ML models.
“We need to minimize risk while maximizing ROI,” Rodriquez said.
The three most valuable parts in the AI universe include the data sourcing, model sharing, and the model output.
“We have to make sure the AI is unbiased as possible and at the same time provide context,” Rodriquez said.
Data management methodologies promote the advancement of AI/ML tradecraft, enabling the combination of data from different sources. These actions lead to discovery and augment insights. They establish a common frame of reference to foster development of policy-compliant AI/ML models.
He presented a case study involving the DoD establishing AI/ML governance and data management. The approach for this was to uncover the main value drivers for implementing AI/ML and learn what AI can and cannot do.
“Oftentimes you need AI to solve your problems,” Rodriquez said. “Think about what you want to do and what AI can do for you.”
As humans, there are certain things we cannot do. AI can allow you to find particular information to allow you to grow, he explained. This can lead to the creation of an interactive environment that will help navigate chains of cause and effect that are longer than we are able to learn on our own.
To mitigate risk, the intelligence community created tagged labels for those looking at, sharing, and using the data. From there, permissions were created for certain users to make sure information is secure throughout the DoD and other departments.
“We recognize that our journey is ongoing but we know we are going in the right direction,” Rodriquez said.
Many Data Summit 2023 presentations are available for review at https://www.dbta.com/DataSummit/2023/Presentations.aspx.