“NATIONAL DEFENSE MAGAZINE” BY Josh Luckenbaugh
“A project to explore potential applications of generative AI for acquisition activities as the DOD as a whole is making a concerted effort to put advanced capabilities not only into the hands of warfighters but also those buying their equipment.”
______________________________________________________________________________
“For all that has been made of how artificial intelligence can increase the military’s warfighting capabilities, such as improving targeting and intelligence analysis, an important group at the Pentagon has largely been left on the sidelines.
“For the trillions of dollars that we put” toward creating “exquisite capabilities for the warfighter — and rightfully so — we haven’t given many capabilities to the people who manage those activities,” Tara Murphy Dougherty, CEO of software company Govini, said during a recent panel discussion at the Association of the United States Army’s annual meeting. When the Defense Department or a service awards a contract for “some unbelievable autonomous system, that program is going to get managed on the government side in a spreadsheet.”
Two days after Murphy Dougherty’s comments, the Army announced a new pilot project to explore potential applications of generative AI for the service’s acquisition activities. At the same time, the Defense Department as a whole is making a concerted effort to put advanced capabilities not only into the hands of warfighters but also those buying their equipment, officials and experts said.
The pilot, called #CalibrateAI and led by the Deputy Assistant Secretary of the Army for Data, Engineering and Software under the Assistant Secretary of the Army for Acquisition, Logistics and Technology, will utilize LIGER — a generative AI solution developed by management consulting firm LMI — that is “designed to deliver tailored responses that are highly relevant to specific topics, improving the efficiency and effectiveness of information retrieval and analysis,” a service release stated.
An LMI release said the service will use LIGER to streamline complex tasks; locate and synthesize information from Army policies, market research and acquisition documents; and develop high-quality products more quickly and efficiently.
An Army spokesperson said in an email that one example of a task the service intends to pilot through #CalibrateAI is aggregating and summarizing responses to requests for information.
“This allows for acquisition professionals to be even more effective with their time and ensure we are able to take into account the responses that we receive, quickly identifying differences and trends,” the spokesperson said.
The pilot will run through the second quarter of fiscal year 2025, and the service hopes it will inform its efforts to scope out a cost-effective path for the Army’s broader adoption of AI capabilities, the spokesperson said.
“By using off-the-shelf AI tools … #CalibrateAI will explore how we increase productivity while enhancing the accuracy of information,” Jennifer Swanson, deputy assistant secretary of the Army for data, engineering and software, said in the service release. “The ability to query curated document sets for generating new content, along with providing citations, will ensure that our outputs are not only accurate but also easily fact checked.”
LIGER can handle controlled unclassified information data, “ensuring that sensitive information is managed with the highest level of security,” and also includes customizable user-access controls to protect “need to know” information, the release stated.
One potential risk of using generative AI is the possibility of “hallucinations,” or responses that contain false or misleading information.
The service spokesperson said #CalibrateAI will help the Army “develop the right approach to identifying hallucinations. The tool itself will include citing of all sources to ensure that our humans in the loop can qualify and validate information as needed and will indicate when an answer may be a hallucination.”
The Defense Department’s Chief Digital and Artificial Intelligence Office, or CDAO, is also investing in AI tools for acquisition, said Dr. Radha Plumb, the head of the office.
One such tool is AcqBot, a large language model that can help contracting officers write contracts more quickly, Plumb said in an interview.
AcqBot ingests contract language, and “what it can do is take the contract language and the contract workflow and do some initial creation and generation parts of the process for contracting officers,” increasing the speed of contract writing by about 400 percent, she said.
“There’s still dedicated human oversight and validation of that textual content at each stage of the contract writing process — we’re not outsourcing contracting writing to machines — but the contract writers don’t have to look through thousands of pages of documents to generate those contracts,” she said. “They can get potential options from the actual AcqBot tool.”
Plumb’s office is also in charge of the Defense Department’s Task Force Lima, which is exploring potential use cases for generative AI in the department.
CDAO is in the final stages of reviewing the task force’s report, which identified over 200 specific use cases for generative AI, she said.
One area the task force identified where generative AI could be useful is program management, “where algorithms can look at historical performance information, cost data, information in bills of sale,” as well as the supply chain, “and put that all together in a single place for program managers to really have greater oversight on their environment.”
Additionally, AI is helping the department make significant gains in predictive maintenance, the practice of continually monitoring the condition of platforms “to get ahead of maintenance issues rather than wait for them to have a maintenance issue,” Plumb said.
AI can help operators evaluate “much more disparate sources of information” such as weather effects and a system’s operational tempo to “really understand where they can or should take risks on maintenance based on current operational conditions and current risk to the particular platform,” she said.
Murphy Dougherty said in an email that Govini has developed a suite of AI tools called Ark that one program office used to proactively monitor more than seven million parts under its management and reduce their time to find alternative parts by 98 percent.
“There are more of these wins that we can turn into a reality today, but success ultimately hinges on DoD’s willingness to translate AI’s potential into real-world, operational value,” Murphy Dougherty said.
Ryan Novak, AI and acquisition outcome lead at MITRE, said a major barrier to more widespread adoption of AI in the acquisition community is cultural resistance to change.
“I compare it to when acquisition used to do things on paper, and then we moved to computers, … and then we moved from computers to the internet, and … there was pushback,” Novak said in an interview. Incorporating AI into acquisition processes is a similarly significant change, he added.
Plumb said the Defense Department currently has “really great acquisition professionals” who “know how to navigate the system,” but “they are not always trained in the most cutting-edge digital tools, and we in CDAO have the responsibility of helping our buyers be the smartest buyers they can be.”
The office is working with the Defense Acquisition University and the newly established Defense Civilian Training Corps to provide AI-centric training and educational events focused on the use of AI in acquisition, she said. Through these offerings, CDAO can show both experienced acquisition professionals and new recruits “the art of the possible” and help to get “teams or organizations inside DoD interested in how we can leverage [AI] in their specific workflow.”
As the adoption of AI in the acquisition community increases, the workforce itself will evolve, Novak said.
“I believe that the way we do business from an acquisition perspective is going to fundamentally change in the next three to five years,” he said. “The people we hire, the skill sets we hire for, the processes we use, regulations, procedures, all that stuff, is going to fundamentally change.”
Plumb said software engineering know-how has not historically overlapped with the acquisition workforce, but that absolutely must change going forward.
“The vast majority of the digital solutions in the Department of Defense are going to be procured from the cutting-edge innovation that happens in the private sector, and for that procurement and then the program management afterwards to be effective, we’re going to need to have the acquisition professionals with the training and expertise to both buy the right things and then manage those things over time,” she said.
The skill set for managing digital solutions is very different from the one needed for a major hardware platform, she added.
“It’s not enough to have … a small number of folks who know how to code or know how to tweak” a user interface, she said. The Defense Department will need “some computer scientists or data scientists who are acquisition professionals instead of aerospace engineers … so we need to start thinking about the underlying expertise and making sure we have the pipeline there.”
As the Defense Department thinks about “what the future force structure looks like for acquisition professionals,” the Chief Digital and Artificial Intelligence Office is focused on providing the current acquisition workforce with training courses “to be able to execute this transformation inside the department,” she said.
“We’re going to have to upskill the workforce we have, [and] I think there’s a ton of appetite for that,” she said.
Murphy Dougherty expressed optimism in the Defense Department’s adoption of AI for acquisition, but said there is little time to waste.
“Our adversaries and competitors around the world are outpacing us in the defense innovation race. Not because they are better or smarter or more patriotic, but because they are producing faster,” Murphy Dougherty said. “We have to make up time. The only way to make up time is to work faster. We cannot do that without effectively adopting AI-enabled capabilities for acquisition.”
Kommentare