Love it or hate it, artificial intelligence (AI) is flourishing in workplaces across the globe. However, this growth does not extend to offices filled with employees of the U.S. federal government, as highlighted in a recent report from Fedscoop. The findings raise concerns about Washington’s ability to keep pace with technological advancements. According to this analysis of various federal agencies, there is a severe lack of funding and talent required to adequately train personnel to harness the advantages of AI.
Last year, President Joe Biden experienced a wake-up call regarding the implications of AI after watching Mission: Impossible—Dead Reckoning Part One. Alarmed by the possibilities and potential risks associated with AI, he signed an executive order urging technology companies to develop AI responsibly. Additionally, the White House instructed federal agencies to compile reports outlining their AI usage plans, strategies to mitigate risks to humanity, and challenges that hinder the mass adoption of AI technologies. These reports were to be made public by September. Fedscoop has gathered these submissions, revealing a common thread among them.
Among the twenty-nine agencies that submitted reports, twelve mentioned the challenges related to data access, six pointed to a deficiency in AI-trained personnel, and six highlighted funding limitations that hinder their AI initiatives. Notably, the Department of Energy, which oversees the nation’s nuclear arsenal, expressed frustrations concerning security concerns around cloud services, as well as a shortage of graphics processing units (GPUs). A report from the DOE emphasized, “The IT infrastructure barrier extends beyond the serverless CSP services to the availability and timeliness of securing virtual machines with the requisite Graphics Processing Unit (GPU) hardware to develop, train, manage, and deploy advanced AI models.” The report further explained, “This challenge is industry-wide; however, it will impact the rollout and adoption of more advanced customized use cases that require dedicated GPU hardware.”
The ongoing data dilemma is particularly significant. Many federal agencies have existed for decades, resulting in an ebb and flow of personnel, with some staying for years while others cycle in and out with each new presidential administration. Consequently, many technological systems within these agencies are developed in an ad-hoc manner. Equipment is typically replaced only when absolutely necessary, a practice exemplified by America’s nuclear weapons systems, which utilized massive eight-inch floppy disks to run command and control software until 2019. When Colin Powell took over as Secretary of State in 2001, he found himself surrounded by pre-internet era computers from a company that had gone bankrupt nearly a decade earlier.
The mass adoption of AI presents Washington, D.C., with technology hurdles that need addressing simultaneously across all departments. The repeated mention of data issues across multiple reports reflects this outdated infrastructure. To train internal large language models (LLMs) for governmental purposes, data must be centralized and secured. In many agencies, however, data is scattered across hundreds of different locations, and off-the-shelf solutions often lack the necessary security for government operations. Another prevalent theme was a lack of AI understanding among the workforce, accompanied by a considerable fear of embracing the technology. The Nuclear Regulatory Commission, which oversees nuclear power plants and radioactive materials, reported that while its workforce showed interest, there was also apprehension fueled by a general lack of knowledge regarding AI’s capabilities. To address this, the agency recognizes the need for continuous effective change management, allowing employees to fully leverage AI’s advantages as they are introduced.
Funding issues were also highlighted among many agencies. The NRC indicated that it could only assess, test, implement, and maintain new capabilities when resources were accessible. The Export-Import Bank of the United States reported, “AI use cases compete for funding and staffing with other important priorities at the Bank including non-IT investments in core EXIM capabilities, cybersecurity, and other use cases in our modernization agenda.”
Catching up with the technological advancements seen elsewhere in the country will be both costly and time-consuming for these federal agencies. Reflecting on similar challenges in the past, Powell recounted in 2001 having to procure computers for everyone in his office, amounting to 44,000 machines. “Don’t ask me how I got the money, because I won’t tell you,” he shared during an IT symposium in 2019.
Discover more from Marki Mugan
Subscribe to get the latest posts sent to your email.
