Hundreds of technology experts from the public and private sectors, as well as academia, came together earlier this month for NVIDIA's GPU Technology Conference to discuss U.S. federal agency adoption of AI and how industry can help.Leaders from dozens of organizations, including the U.S. Department of Defense, the Federal Communication Commission, Booz Allen Hamilton, Lockheed Martin, NASA, RAND Corporation, Carnegie Mellon and Stanford Universities, participated in approximately 100 sessions that were part of GTC's Public Sector Summit.
They talked about the need to accelerate efforts in a number of areas, including education, access to data and computing resources, funding and research. Many encouraged government executives and federal agencies to act with a greater sense of urgency.
Artificial intelligence is inspiring the greatest technological transformation of our time, Anthony Robbins, vice president of federal at NVIDIA, said in a panel with former Federal CIO Suzette Kent and retired Lt. Gen. Jack Shanahan during one of the talks focused on Building an AI Nation. The train has left the station, Robbins said. In fact, it's already roaring down the tracks.
We're in a critical period with the United States government, Shanahan said during the panel. We have to get it right. This is a really important conversation.
Just Get Started These and other speakers cited a common theme: agencies need to get started now. But this requires a cultural shift, which Kent spoke of as one of the most significant challenges she experienced as federal CIO.
In any kind of transformation the tech is often the easy part, she said, noting that the only way to get people on board across the U.S. government - one of the largest and most complex institutions in the world - is to focus on return on investment for agency missions.
In a session titled Why Leaders in Both the Public and Private Sectors Should Embrace Exponential Changes in Data, AI, and Work, David Bray, former Senior National Intelligence Service Executive, FCC CIO, and current inaugural director and founder of the GeoTech Center at the Atlantic Council, tackled the same topic, saying that worker buy-in was important not just for AI adoption but also for its sustainability.
If you only treat this as a tech endeavor, you might get it right, but it won't stick, Bray said. What you're doing isn't an add-on to agencies - this is transforming how the government does business.
Make Data a Priority Data strategy came up repeatedly as an important component to the future of federal AI.
Less than an hour before a GTC virtual fireside chat with Robbins and DoD Chief Data Officer David Spirk, the Pentagon released its first enterprise data strategy.
The document positions the DoD to become a data-centric organization, but implementing the strategy won't be easy, Spirk said. It will require an incredible amount of orchestration among the numerous data pipelines flowing in and out of the Pentagon and its service branches.
Data is a strategic asset, he said. It's a high-interest commodity that has to be leveraged for both immediate and lasting advantage.
Kent and Shanahan agreed that data is critical. Kent said agency chief data officers need to think of the federal government as one large enterprise with a huge repository of data rather than silos of information, considering how the government at large can leverage an agency's data.
Invest in Exponential Change The next few years will be crucial for the government's adoption of AI, and experts say more investment will be needed.
To start, the government will have to address the AI talent gap. The exact extent of the talent shortage is difficult to measure, but job website statistics show that demand for workers far exceeds supply, according to a study by Georgetown University's Center for Security and Emerging Technology.
One way to do that is for the federal government to set aside money to help small and mid-sized universities develop AI programs.
Another is to provide colleges and universities with access to more computing resources and federal datasets, according to John Etchemendy, co-director of the Human Centered Artificial Intelligence at Stanford University, who spoke during a session with panelists from academia and think tanks. That would accelerate R&D and help students become more proficient at data science.
Government investment in AI research will also be key in helping agencies move forward. Without a significant increase, the United States will fall behind, Martijn Rasser, senior fellow at the Center for New American Security, said during the panel discussion. CNAS recently released a report calling for $25 billion per year in federal AI investment by 2025.
The RAND Corp. released a congressionally mandated assessment of the DoD's AI posture last year that recommended defense agencies need to create mechanisms for connecting AI researchers, technology developers and operators. By allowing operators to be part of the process at every stage, they'll be more confident and trusting of the new technology, Danielle Tarraf, senior information scientist at RAND, told the panel. Tarraf highlighted that many of these recommendations were applicable government-wide.
Michael McQuade, vice president of research at Carnegie Mellon University and a member of the Defense Innovation Board, argued that it's crucial that we start delivering solutions now. Building confidence is key to continue to justify the increasing support from authorizers and appropriators for the crucial national investments in Al.
By framing AI in the context of both broad AI innovations and individual use cases, government can elucidate why it's so important to knock down barriers and get the money in the right place, said
