That’s what the outgoing HASC ranking member says, and he’s not alone.
Mac Thornberry knows what it’s like to have good technical ideas thwarted by the politics of public perception. In 2011, the Republican congressman and some colleagues came up with a list of cybersecurity recommendations at the behest of then-House Speaker John Boehner, R-Ohio.
“Just as we were releasing our recommendations, we had the Snowden leaks. We had Wikileaks,” said Thornberry, who is departing Congress as the minority leader of the House Armed Services Committee. “Everybody decided that the government was reading your emails to Grandma” and it became “politically impossible to consider any sort of cyber-related legislation in Congress for several years.”
Now Thornberry fears that a similar failure to anticipate and defuse public backlash will undermine Pentagon efforts to develop various emerging defense technologies.
“Our adversaries do not have ethical concerns, but we can paralyze ourselves by misinformation or lack of understanding when it comes to artificial intelligence, robotics, human performance enhancement, all sorts of issues,” he said. “I believe it's important to have a little inoculation with hearings, think tank seminars, papers, about these technologies and what they mean or don’t mean to help prevent this sort of paralysis in the future.”
The most dramatic recent example is Project Maven, launched in 2017 to create a machine learning program that could help DOD analysts sift large amounts of intelligence, surveillance and reconnaissance video. When Gizmodo revealed Google’s participation in the project, many employees objected and some even resigned in protest. The company declined to renew the contract, though it has since indicated a willingness to work with the Defense Department on other projects.
The Pentagon’s challenge in adopting emerging technologies may have less to do with a reluctance to talk about them in public and more with the fact that it can’t seem to settle on a consistent approach. Technology priorities can change drastically from year to year as officials chase after buzzwords and shiny concepts, CNAS analysts conclude in a new report.
“In 2014, then-Secretary of Defense Chuck Hagel launched the Defense Innovation Initiative to develop a ‘game-changing third “offset” strategy.’ Then-Deputy Secretary of Defense Robert Work subsequently named AI and autonomy the ‘technological sauce’ to empower this Third Offset Strategy,” authors Paul Scharre and Ainikki Riikonen write.
The arrival of Michael Griffin in February 2018, the Department’s first defense undersecretary for research and engineering, was supposed to help establish clearer technology priorities, and Griffin did set priorities in April 2018.
But before long, changes crept in. “These priorities fluctuated in number and order; they ranged from 10 to 13 priorities with AI, hypersonics, nuclear modernization, and other technologies rotating up and down in importance,” note the CNAS authors. Currently, the undersecretary’s office lists them as: artificial intelligence, hypersonics, space, quantum, science, joint, all-domain command and control, microelectronics, autonomy (as distinct from AI), cyber, and biotechnology.
When Mark Esper went before the Senate for his confirmation hearing in July 2019, the defense-secretary-to-be added a new wrinkle: some of the priorities were of higher priority than others. “Different people put different things number one. For me, it is artificial intelligence,” Esper said. “I believe whoever masters it first will dominate the battlefield for many, many, many years. It gets to how we can think more quickly, how we can work and semi-autonomously. I just think it is a game changer.”
Last January, the FY 2021 Defense Wide Review added nuclear modernization to the mix, perhaps to shield one of the more politically vulnerable areas of defense investment from potential congressional critics. In April, the top priority shifted to microelectronics. In August, after Griffin left the Pentagon, a new acting undersecretary, the relatively inexperienced Michael Kratsios, highlighted microelectronics, 5G, hypersonics, and AI and made a passing reference to quantum computing and “other industries of the future quickly extend across borders.”
Those moving priorities may be, in part, because the Pentagon doesn’t have a good way to track exactly how much contractors are spending on independent research and development in those areas, according to a September GAO report.
Tara Murphy Dougherty, CEO of Govini, an artificial intelligence-driven analysis firm, says that the Department “has put a lot of money into developing advanced capabilities, such as AI, data, hypersonics, cyber, and more.” Her company’s analysis from June shows that the Pentagon’s research, development, testing and engineering budget rose to $80.6 billion in 2020, an increase of more than $20 billion over 2016.
Funding for prototypes in particular has nearly doubled from $14.3 billion in 2016 to $28.5 billion in 2021.
“The most important aspect of this, however, is whether DoD knows what it is getting for these investments,” said Dougherty. As more and more funding is put toward prototyping, for example, those are dollars that are not directed toward systems development or operations and maintenance that can drive readiness. Without decision science to evaluate these tradeoffs, DoD will struggle to evaluate the [return on investment] of its [research, development, testing and engineering] investments--a trend that will only increase as the need for advanced capabilities intensifies but the overall defense budget topline likely stays roughly even.”
Say Scharre and Riikonen, “Too often, investments appear to be driven by the whims of department senior leaders. This is an unsatisfactory approach. At best, the department is whipsawed from one priority to the next, without the sustained investments needed to mature any one given area.” They suggest the Department adopt a “transparent framework for identifying technology priorities that will provide clarity and stability in the department’s priorities.”
In particular, they say to focus on those areas of information technology that are maturing the fastest. “The highest priority technology area for the DoD should be digital technologies that are riding exponential curves. These technologies are maturing rapidly no matter what the DoD does. There is twice as much money spent annually on information technology as all military spending from every country combined,” they write.
That means better engagement with consumer-facing technology companies and academia. Historically, that has not been a strong area for the military but it’s an area that virtually all defense leaders stress as a key priority for them.
Lawmakers can help, said Thornberry, if they can spend less time worrying about requirements for specific programs and provide funding flexibility.
“Congress, especially the appropriators, have to get more comfortable with a pot of money being available for a particular purpose and then full transparency on how those funds are used. Flexibility is the key to attract more suppliers to do business with the Department of Defense to overcome that infamous valley of death to enable more experimentation and prototyping and to get technology into the field faster,” he said.
“We’ve made progress--and hopefully we’ll make a little more progress in this year’s NDAA that we’re now negotiating--in encouraging non-traditional suppliers to do business with DOD, to help small and mid-sized companies bring their innovation to the table.”