DoD can’t define OR track it – NOR cultivate it
Danielle C. Tarraf 19, William Shelton, Edward Parker, Brien Alkire, Diana Gehlhaus, Justin Grana, Alexis Levedahl, Jasmin Léveillé, Jared Mondschein, James Ryseff, Ali Wyne, Daniel Elinoff, Edward Geist, Benjamin N. Harris, Eric Hui, Cedric Kenney, Sydne Newberry, Chandler Sachs, Peter Schirmer, Danielle Schlang, Victoria M. Smith, Abbie Tingstad, Padmaja Vedula, and Kristin Warren, The Department of Defense Posture for Artificial Intelligence: Assessment and Recommendations, RAND Corporation, 2019. https://www.rand.org/pubs/research_reports/RR4229.html, nihara
DoD lacks clear mechanisms for defining and tracking AI talent. This situation is because of the lack of an AI workforce classification and the lack of consensus on the definition and requisite skills of an AI worker. Currently, the technical AI talent that exists is being placed inconsistently in military and civilian occupational specialties.
For example, our interviewees across the services noted that civilian AI talent is found across occupations, including operations researchers, program analysts, scientists, software developers, and engineers. On the uniformed side, the Army and Navy are actively considering a new occupational specialty for servicemembers just for AI, but the Air Force and Marines are not. Adding to this complicated picture, some existing uniformed and civilian occupations that might have a role to play in developing AI technologies, such as software engineers, are classified as cyber talent (see section “DoD Posture for Cyber” in Appendix D). On this note, although it is not prevalent, we observed an inclination among some of our interviewees to view AI, software, cyber, and data science as somewhat interchangeable, particularly with regard to technical talent.37
DoD struggles to grow and cultivate AI talent. Our interviews suggest a mixed appreciation for what technical AI talent consists of and which AI talent is needed. Several entities we interviewed, such as the service labs, had a clear sense of AI talent needs, but the majority were still in the beginning stages of such considerations and were more likely to emphasize contracting out for technical talent. Moreover, for those that were clear on AI talent needs, it was a challenge to define the exact knowledge, skills, and abilities they perceived that were required. Ultimately, the AI talent needs of DoD (type,38 quantity, and mix) will depend on the broader strategy pursued for scaling AI, and the extent to which scaling AI will rely on the development of products in-house as opposed to through contracting and outsourcing. The skill sets needed for development of products in-house are significantly different from those needed for contracting and outsourcing, though all AI talent (technical or managerial) is difficult to access in the present market. Nonetheless, the consensus is that DoD faces stiff competition for AI skills and expertise, as evidenced by our interviews across academia, industry, and DoD.39 Many of our DoD interviewees discussed the challenges related to attracting and recruiting technical talent more generally, and expressed the belief that AI talent would be no different. In that spirit, we point to a recent RAND study on career paths for data scientists within the Defense Intelligence Agency.40 Interviews across DoD cited intense competition with the private sector, the limited ability to compete on salary, and long hiring processes. At the same time, the majority of our interviewees were optimistic that DoD could compete well on mission and the opportunity to work on important and interesting problems (see section “Talent” in Appendix B, and section “Thoughts Across Industry: On DoD Competing for AI Talent” in Appendix C).41 We note that similar themes to the ones we highlight here appear in the DIB’s SWAP study, even as we reemphasize that AI is not software.42
Overall, the services generally acknowledged the need for greater permeability of technical talent between civilians and uniformed services.43 Several interviewees expressed the belief that they need to exploit existing authorities for civilians to do a tour of duty for several years at DoD, with one noting they would be more than happy to take technical talent that was “burned out” in industry or academic jobs and wanted a change. This idea complements what we heard from industry and academia, where a variety of mechanisms are used to support external activity of employees in the interests of retaining them (see section “Academia: Talent” in Appendix C, and section “Industry: Talent” in Appendix C). Interviewees also described retention efforts for current service members with technical skills; these efforts leverage existing programs that enable some degree of permeability. This process includes formal training and rotation opportunities for officers in technical disciplines through established partnerships with industry and academia.44 A recent congressionally mandated RAND study catalogues more examples of such rotation programs.45 Regardless of the approach taken, if DoD is to be successful in scaling AI, it needs to ensure it has access to some level of AI talent, both technical (R&D) and managerial (acquisition), and that the talent that exists maintains its knowledge and skills in a fast-changing technical environment. DoD also needs to cultivate respect of and promotions for military personnel involved in AI activities and to compete in an AI talent market in which individuals expect change (in jobs and employers) every few years.