Artificial Intelligence

DoD wants to ‘popularize’ data patterns to leverage AI

Before DoD can leverage AI at scale, it needs access to data. But challenges ranging from policy constraints to technical issues to mistrust continue to hinder ...

Anastasia Obis

February 13, 2024 3:44 pm

3 min read

Before the Defense Department can even start leveraging artificial intelligence at scale, it needs access to data.

But barriers ranging from policy constraints to organizational silos to technical domain challenges to acquisition pathways to lack of understanding and trust continue to hinder effective implementation of the nascent technology across the department.

The DoD’s Chief Digital and AI Office wants to “popularize” data patterns and establish common practices for organizing and structuring data.

“I hesitate saying standards because those sound a little bit draconian or a little bit top-down. But we want to use good patterns so people know what it means to share data,” Bill Streilein, CDAO’s chief technology officer, said at the DON IT West conference Monday.

To address the lack of trust in artificial intelligence, the CDAO published its Responsible Artificial Intelligence (RAI) Toolkit to ensure that AI applications align with ethical standards. The CDAO’s RAI team plans to add more features to the toolkit, including an acquisition toolkit with standardized contract language and RAI project management tools.

The CDAO is also investing in educational programs to enhance the understanding of AI across all levels within the department.

Streilein said the department had just launched executive-level courses at the Johns Hopkins University, the Naval Postgraduate School and the Massachusetts Institute of Technology. These three to four-day courses equip professionals at the executive level with a better understanding of the implications of incorporating data analytics and artificial intelligence into daily operations. The CDAO also plans to develop courses at other levels.

DoD is also offering 11 new positions that “recognize the value and power” of people with data analytics and AI expertise.

“People can maintain this skill set through a career in the department, perhaps going in and out of industry to carry the lessons learned within the department and from industry back in to have an effect,” Streilein said.

DoD recently released its data, analytics and AI adoption strategy. It builds on DoD’s previous data and AI strategies and addresses recent industry advances in federated environments, decentralized data management and generative AI.

Streilein said experimentation with AI is at the core of the strategy, promoting an approach of exploring and testing out AI technologies rather than expecting them to be flawless and immediately deployable.

“One of the core words in our strategy is experiment with AI. We shouldn’t expect this technology to be right off the shelf ready to use like a toaster. It’s something we have to bring in, experiment with, try out and learn,” Streilein said.

“The main idea is to continuously experiment and deploy, but it’s being done with user feedback. It’s not that we’re going to throw capabilities over the wall, and the users just have to figure out how to use it. We want that feedback because that’ll help us understand what needs to change the next time.”

The CDAO wants DoD to adopt a product-oriented approach. It wants the department to treat data as a product rather than a strategic asset.

“Product orientation changes our mindset from thinking about things like data as strategic assets, which tend to be hoarded if you think about what an asset is, to a product, which is something that somebody uses,” Streilein said.

“So if somebody has a data product, the value is realized because there’s a customer. Our goal is to enable each CTO and CIO across the department with the skills and the tools to be a product manager so that they can provide their customer with what they promised.”