The US Government Will Spend $1B on AI Next Year — Not Counting the Pentagon

The federal government plans to spend almost $1 billion in nondefense artificial intelligence research and development in fiscal 2020, according to a supplemental report to the president’s budget request.

“The U.S. has pushed the boundaries for computational power, we have given our innovators the freedom to thrive, and today we can proudly say America continues to be the leader in artificial intelligence,” federal Chief Technology Officer Michael Kratsios said Tuesday at a Center for Data Innovation event. “This new supplement report demonstrates just how diverse and extensive our efforts are.”

The figure indicates a weighty increase from 2016—when all agencies, including defense, collectively spent about $1 billion on AI. However, immediately after Kratsios’ announcement, industry experts and former federal officials weighed in on what more needs to be done to secure and sustain American leadership in AI across the global technological landscape. 

Related: White House Updates National Artificial Intelligence Strategy

Related: NIST Has a Roadmap to Develop Artificial Intelligence Standards

Related: 42 Countries Agree to International Principles for AI

Created in support of the president’s February executive order to promote American advancements in AI, the supplemental report constitutes the “first-ever reporting of agency-by-agency federal investments” in nondefense AI research and development. The document describes research and development coordination activities that certain federal agencies are planning for 2020, as well as information on actual agency investments for fiscal 2018, estimated investments for fiscal 2019, and the requested funding levels for fiscal 2020. 

The report also lays out key programs and the administration’s strategic priorities around AI, including coordinating long-term federal investments in research, promoting safe and effective methods for human-AI collaboration and improving the measurement and evaluation of AI technologies through new benchmarks and standards. The document also provides an important baseline that supports federal insiders in tracking U.S. AI spending going forward, Kratsios said. 

“By breaking down exactly how we are spending our nondefense AI [research and development] dollars, we can better identify opportunities for future investment, conduct long-term strategic planning across the government and find new opportunities for collaboration between the federal government, industry and academia,” he said. “And it demonstrates the increased emphasis on AI [research and development] under this administration.”

On a panel following Kratsios’ keynote, industry leaders agreed that the billion in investments for AI spending next year is a step in the right direction, but they also said the government should be doing much more to ensure the nation maintains leadership in the space. 

Fiona Alexander, distinguished policy strategist at the American University School of International Service who also served in the federal government across four administrations, said it’s critical that the administration and Congress offer up clarity on America’s privacy regime in the context of AI. She said the Obama administration pushed to create a comprehensive approach to data privacy, as well as a proposal for a privacy blueprint, but the efforts didn’t gain traction from the private sector.

“And the result was that the privacy playing field got taken over by Europe and the [General Data Protection Regulation is] setting the standards, whether you like it or not, and the question is how do you deal with that?” Alexander said. “What’s lacking really is this clarity on policy—and privacy is not part of this AI strategy. ... Maybe someone needs to take a look at that if you are looking at American leadership going forward.”

Panelists also weighed in on the need to build and boost the federal workforce’s artificial intelligence skills. Michael McLaughlin, a research analyst for the Information Technology and Innovation Foundation who recently produced a report on global AI leadership, warned that China plans to train 5,000 students in AI in the next five years while the U.S. lacks a unified technological education strategy. The Center for Data Innovation Director Daniel Castro said he searched USAJobs.gov for all jobs in the related categories and found 12 federal job listing mentioning artificial intelligence and 14 related to machine learning. 

“Where are the jobs?” he asked. 

The administration also held an AI summit at the White House Monday, hosting 200 leaders from the government, industry and academia to address priorities across the burgeoning landscape. Intel’s Director and Managing Attorney of U.S. Artificial Intelligence and Healthcare Policy, Jackie Medecki, attended the event. 

“What I was surprised about yesterday is, as I was listening to them give examples of the different ways that they’ve been using AI with the different agencies, was just how basic the AI that they were implementing really was,” she said. 

While Medecki noted some fascinating pilots were highlighted across certain agencies, she said it became clear that new insights and developments are not being shared across agencies, that many systems are not interoperable, and that there’s a variety of “low hanging fruit” to be addressed. 

“I think there's a ton of really basic work that needs to be done to set the stage for successful use of AI within the government,” she said. 

And while the supplemental report identifies future investments in AI development and deployment, the administration did not request additional funding for future initiatives. Panelists agreed that increased funding is necessary to preserve American dominance. Some noted that compared to the government’s spending on other specific technological use cases, $1 billion does not measure up.

“I would put a lot more money on the AI thing,” Anthony Robbins, vice president of the North America public sector at NVIDIA, said. “So you know, Michael Kratsios today referenced putting a billion dollars on that and that’s interesting, but we will spend billions of dollars trying to build the fastest supercomputer in the world, which is down at Oak Ridge.”

Robbins said when the administration initially released its executive order on AI, many questions were raised as to why there was no funding attached to support agencies in their efforts to prioritize the new technology. 

“That was a big question that the industry had, ‘OK, where was the funding?’” he said. “So a billion is certainly a great thing, and it’s certainly interesting, but it’s not nearly enough.”