The Chief Digital & Artificial Intelligence Office (CDAO) reached one year of operations this month with several achievements in creating, implementing, and steering digital transformation and artificial intelligence (AI) in the Department of Defense (DoD).
Announced in December 2021, the CDAO was established on June 1, 2022. The CDAO made significant strides in setting policies, guidelines, and the structure to guide AI development in the DoD with the goal of enabling an enduring advantage for America’s armed forces.
Within the first year, the CDAO achieved significant accomplishments, most importantly, introducing the foundation of defense AI systems, including an AI hierarchy of needs and the establishment of five strategic initiatives that outline the DoD’s path to AI-enablement in FY2023: digital talent management; improved data quality; AI/ML scaffolding; business performance metrics; and the Combined Joint All-Domain Command & Control (CJADC2) system.
Digital Talent Management – The CDAO was designated as the DoD’s Functional Workforce Manager for data, analytics, and AI personnel, allowing the CDAO to set standards for the digital workforce. It established an executive training program to increase our senior leaders’ digital literacy. The CDAO also launched a pilot program for 1,500 employees to expand access to MIT Horizon’s library of resources to improve digital literacy across the DoD.
Improved Data Quality – The Maritime Object Collection project collected 17TB of Coast Guard Air Stations, annotated and labelled over 12,800 images and launched Phase 3 of its plan to develop AI models for use in manned and unmanned aircraft to enable the rapid detection and classification of specific maritime targets. Project Harbinger achieved a significant milestone in the past year by partnering with industry to develop advanced machine learning algorithms for detecting acoustic energies. And finally, the Publicly Available Information (PAI) project has explored overseas open-source information. These models review over 15 million posts daily and filter them down to 100-200 ones of interest, dramatically increasing the efficiency of intelligence analysts.
Business Performance Metrics – The CDAO increased use of the Advana business performance metrics platform from 40,000 registered users to over 72,000 in its first year, and has provided the Secretary and Deputy Secretary of Defense an analytics dashboard called PULSE that enables them to evaluate outcomes across the DoD significantly faster than before.
CJADC2 – The CDAO continues to provide key support to CJADC2, including the continued development of the data integration layer and developing mature software tools to digitize battle management. Also, the CDAO has continued the Global Information Dominance Experiments (GIDE) to test and advance the digital infrastructure of AI systems. GIDE V was completed in January and GIDE VI is currently in progress.
"Our goal was to set the conditions and environment for the Department to tackle the monumental task of transforming its data to make AI systems possible," said Dr. Craig Martell, Chief Digital & Artificial Intelligence Officer. "We’re excited about the momentum we have built and look forward to getting closer to our goal of AI enablement in our second year."
The CDAO’s partnership with industry provides the DoD an advantage in AI development. The CDAO continues the Tradewind acquisition program, designed with small business innovation in mind, that has provided nearly 300 solutions to its data and data integration problems, 96% of which came from small businesses. The CDAO also introduced the “TryAI” program, which enables the DoD to test industry solutions quickly before initiating a purchasing action.
Guiding the CDAO’s strategic initiatives has been its commitment to responsible and trusted AI development. The office is instrumental in assisting the Deputy Secretary of Defense to publish the DoD Responsible AI Strategy & Implementation Guidelines and established tenets that will ensure that AI systems are continuously checked and verified, and decisions can be traced to identify and eliminate biases in the data.