As Federal agencies look to further adoption of artificial intelligence (AI) technologies, experts both in government and industry are stressing the importance of making sure AI is mission-integrated, and that agencies are moving towards a more distributed AI approach.
Pamela Isom, director of the AI and Technology Office at the Department of Energy; Curt Smith, NVIDIA’s Vice President, GPU Datacenter Architect; Jay Boisseau, AI and high-performance computing technology strategist at Dell; and Judson Graves, director of Analytics and AI at ViON, each explained where they’ve seen strides made in Federal AI adoption at MeriTalk’s “The Edge of AI: Federal AI Adoption” webinar on Jan. 12.
“The Federal government as a whole, from my perspective, is making pretty good strides when it comes to understanding how to apply AI towards the mission,” Isom said during the panel discussion. “And that’s the most important thing because if we address the mission, then we address the citizens’ needs.”
That’s an approach shared by Smith, who said it is crucial that organizations have a customer experience and mission-oriented vision of how AI can be helpful.
“One of the very first pieces of the work that an organization needs to start on their AI journey is to start with a vision and how that vision can transform the customer experience and how it relates to the organization’s mission,” Smith said.
Isom said she’s noticed a shift from Federal AI adoption being defense-focused to a more distributed approach across the Federal government. At DOE, Isom said the agency is looking at how AI can be used to aid the agency in its fight to mitigate the impacts of climate change.
Boisseau emphasized the importance of making data accessible in building AI models.
“First and foremost is make sure you always have access to all of the good data that can help train” AI models, said Boisseau. “This is one of the challenges in the current era.”
“We’ve created all of these data silos – and we tend to think of silos as a negative word now – but it had many good reasons for separating databases from different parts organizations, to reduce risk and to enable you to create the right schemas and tables and such that are optimized for just the people who were using it,” he said.
“So, there were many good reasons for specializing data access and securing subsets of data,” Boisseau added. “But as we adopt these more powerful data analysis techniques – up to including AI techniques – we often want more access to more of the data across departmental boundaries, and even agency boundaries while recognizing at the same time that that data has increased value to external threats as well. So, we have to balance that and make it more available to people who do need it, while continuing to increase the secure protection of that data from external threats.”
Graves cautioned agencies that are working on structuring and designing pilot AI programs to not skip any key steps.
“What I’ve stumbled on a lot of is transitioning from that demo to proof of concept to pilot transition, and honestly, we’ve seen a lot of people shortcutting that,” Graves said. “We see a lot of going into a demo situation and being blown away by some kind of capability out of the box, maybe even integrating a little bit with some of their data. And then kind of skipping the pilot phase and then moving straight into attempting to do it in production, and then really failing miserably.”
“The truth is that AI over the past few years” represents “an intersection of new massive capabilities of storing data, incredibly fast transfer speeds across networks, and the vast amount of computing power we’re suddenly able to unleash on these datasets,” he added. “That three-way combination is really what has enabled this kind of new capability in AI.”
Isom said her office is currently exploring how to utilize AI for projects related to decarbonization, grid and energy resiliency, and hydrogen and solar efficiencies. She said DOE is also working on using AI for autonomous vehicles and drones and sees a distributed-AI approach as the future of integration at the Federal level.
“My office is very focused on not only helping us make sure we stay focused on mission, which includes research and development, but also protection – protecting the AI as we build out or acquire solutions,” she said. “[DOE plans for] more work and more insights and more education and training on distributed AI. And that will complement the whole edge of AI concept and some of that capability. So, we need to put that more into action and operation.”
To hear all of Isom and DOE’s plans, and more from the rest of the panel, register here for MeriTalk’s “The Edge of AI: Federal AI Adoption” webinar and watch it all on-demand.
