The Pentagon must act quickly lest it lose its AI advantage to China’s well-funded advances, according to experts testifying on Capitol Hill and a new report from a government-data company.
China is directing more of its AI-related research into defense applications than the United States, whose tech sector is more focused on consumer AI services such as ChatGPT.
“We need to consider what the overall investment into military implementations looks like. And that’s where there’s a large disparity,” Scale AI founder Alexander Wang told lawmakers. “If you compare as a percentage of their overall military investment, the PLA is spending somewhere between one to two percent of their overall budget into artificial intelligence whereas the DoD is spending somewhere between 0.1 and 0.2 of our budget on AI.”
China’s larger investment in key research and development was reinforced by a report released Monday by Govini, a data analytics and decision sciences company. Echoing its findings in 2018, the company’s 2023 National Security Scorecard shows a big gap between the top 100 U.S. defense companies and the top U.S. companies working in AI.
“What this chart shows is that we’ve made very little progress. And while we might be attracting some of these non-traditional entrants into the defense procurement system, to contribute to AI on national security problems, we certainly aren’t scaling them, at least not based on these numbers,” Govini CEO Tara Murphy Dougherty told reporters on Monday.
This year, “China is outpacing the United States by innovation measures,” Dougherty said.
The Pentagon is increasing its research-and-development spending, but not enough to close the gap, according to Bob Work, who as deputy defense secretary was credited with pioneering a sea change in how his department organized itself around new technology.
“We’re in this situation where we have the largest R&D budget in the Department of Defense’s history. We’re patting ourselves on the back and saying, ‘Man, aren’t we doing good?’ But when you blow out and look at global data, as Govini has done, you say, ‘Hey, it’s not that good,’” Work said during the Govini call. “So, in kind of marginal terms, we’re flat and the Chinese are outspending us. This is why China is so different. In the past, we’ve always been able to outspend our competitors. This time, China will be able to outspend us if they choose to do so.”
Yet another study released on Monday by GDITfound that U.S. military and federal workers are eager to embrace new technologies like AI but are held back largely by leaders’ uncertainty about how much it will cost to hire and train AI experts.
“I think it’s interesting that they want to save money through emerging technology, but it’s also very interesting that they want to look at this as a challenge, as a concern, right? How do they buy? How do they acquire this? And the follow-on resources and tail that comes with implementing new technologies,” said Mike Cole, vice president and chief technology officer for GDIT’s federal civilian division.
Klon Kitchen, a nonresident senior fellow at the American Enterprise Institute, told lawmakers on Tuesday that the United States has advantages in terms of data, particularly military data. While China is in a great position to harness data on its citizens and even purchase data on U.S. citizens, that data doesn’t yield much military advantage. And further, the way that China surveils its own population could undermine the integrity of the data they want to harvest.
“It is my hope, for example, that the Chinese government’s political fragility, strict content controls and general oppression of its own people will compromise or bias much of the data that it collects, diluting its utility and ultimately limiting the development of Chinese AI,” Kitchen said.
Wang said that perhaps the most important step that the U.S. military could take right now is to simply take stock of its data and move to organize it so it can be used to train new AI models. Currently, the military collects 22 terabytes of operational data every day, not just concerning the battlefield but also manufacturing, transportation, and logistics.
“I would say first, establishing the central data repository and then creating a plan by which as much of the 22 terabytes of data generated a day goes into that central data repository, and then creating a plan by which as much of that data is processed and labeled and annotated to be as AI ready as possible,” Wang said, and cautioned: “These are all multi-year efforts that are not going to be solved tomorrow at the snap of a finger.”
Comments are closed.