News

New AI initiative at CMU funded by the Office of the Director of National Intelligence

The Software Engineering Institute (SEI) at Carnegie Mellon is now at the helm of a new national AI initiative funded by the Office of the Director of National Intelligence (ODNI) to “advance the discipline of artificial intelligence (AI) engineering for defense and national security,” as written in the Oct. 14 announcement. The initiative seeks to develop an AI engineering framework that can be applied to anyone working toward the goals of the Department of Defense.

The SEI is a federally funded research and development center (FFRDC), and it is one of three research and development FFRDC laboratories for the Department of Defense. This new ODNI initiative centers the SEI as a hub for AI engineering nationally.

Matt Gaston, director of the SEI’s Emerging Technology Center, told The Tartan in an interview that there is a commonly held belief that the development of new AI technologies is a “craft.” He said, “there are amazing experts that are building these [new technologies] all over the world. But the repeatability of that process, or the scalability of that process, is difficult, and the best practices are not captured or not shared. Moving the discipline from a craft to an actual engineering practice stands to benefit everyone.” This task of moving AI from a craft to a structured engineering practice is the one main goal of the initiative.

The difference between this initiative and others, Gaston explained, is that “most AI projects are focused on AI capabilities specifically,” like the actual implementation of a particular AI technology, or what some call “mission capability.” This new initiative is more about fleshing out the processes by which AI systems are developed, “so that they can be reliable, operate the way that they're expected to operate, and be maintained and evolved over time and in a responsible and methodical way,” Gaston said.

When asked about how this initiative compares to something like the Carnegie Mellon-housed Army AI Task Force, Gaston said that though the “scope, and grand vision” of this initiative may be larger, “the initial funding is, as far as I know, definitely not larger” than the task force.

No additional information on the initial ODNI investment, or expected investment over time, has been released at the present. Richard Lynch, spokesperson for the SEI, told The Tartan, “we don't ever release details like that, unless the sponsor of the work has approved that information to be released. And in doing this release with the ODNI, they didn't choose to share that information.”

Gaston said that the ultimate goals of the initiative will evolve over time, given the fast rate of evolution in AI technology. “For some period of time, we'll see incremental progress into getting to better practices, more repeatable practices, dealing with topics like robustness, or how to implement ethical principles,” Gaston explained, noting that “the end state is hard to envision, right? Who knows if it's textbooks or courses or tools, it's probably all of that, and those will continue to evolve with technology.”

As for what's happening right now, Gaston said of this new initiative, “it's very early days,” adding that the organizational work is just getting started. Currently, the SEI is seeking partnerships with other organizations to begin scaling this work up.

Gaston says that the SEI is viewing this project as a “community initiative,” hoping to gain partnerships across the intelligence and academic communities. “We really view it as a movement, more than anything else,” Gaston stated. “Our expectation, and experience over the years, is that [partners] will bring their own resources to this, because they see the value of advancing this AI engineering discipline.”

In the past, students and faculty have raised concerns regarding other AI programs like the Army AI Task Force, and they have protested the university’s hosting of Palantir, a company that sells technology to ICE. Gaston declined to comment on the potential for community pushback against initiatives like this one, saying instead, “AI is a collection of technologies that, I think, can have a very positive impact on lots of different parts of our economy, and our world. [The] desire to create an engineering discipline around how to do that as well as it can be done [has the potential] to benefit a large portion of society.”