“Artificial intelligence is touching every sector of our economy and has the potential to transform every single one of these sectors,” says Tanya Das, director of AI and energy technology policy at the Bipartisan Policy Center, adding that DOE can help agencies seize on that potential.
The department “has been home to advanced computers and supercomputers for several decades, and they contributed to key breakthroughs that have enabled our current AI revolution,” Das says. “They’re uniquely positioned to help the rest of the government become adopters of AI for their own applications.”
Breaking Down the Key Components and Goals of FASST
The FASST initiative has four pillars.
- Al-Ready Data: DOE will transform its vast classified and unclassified data sets into the largest high-quality scientific data repository for training AI.
- Frontier-Scale Al Computing Infrastructure and Platforms: FASST will build next-generation, energy-efficient, AI-enabled supercomputing platforms and the requisite infrastructure.
- Safe, Secure and Trustworthy Al Models and Systems: Models that DOE develops will be trained, tested and validated for their ability to accurately predict emergent behaviors while maintaining privacy.
- AI Application: Unlike the private sector’s models, DOE’s will be tailored to strategic and critical application spaces that might otherwise not see investment.
DISCOVER: These are the four biggest security risks to generative AI solutions.
Agencies will need solid data, strong infrastructure and reliable AI models to bring to life applications that deliver on the potential of AI.
“These are four priority areas that work well together,” Fleming says.
Measuring Success and Real-World Outcomes
Agencies will need to determine whether AI is delivering on its potential as they engage with FASST.
“We don’t want to come up with new tools just for the sake of using whatever’s popular at the moment,” Das says. “We want to make sure that it results in actual efficiency improvements.”
LEARN MORE: Artificial intelligence can help with hybrid cloud security challenges.
“There are really good, quantitative ways in which we can determine how effective a tool is in speeding up our processes,” Das says. “Let’s say it currently takes an agency four years to get through a review process. If we can cut that down to a year, or even a matter of months, we’ll be able to quantify and see those impacts in a very tangible way.”
Agencies can judge real-world outcomes by measuring speed, cost savings and even citizen ease of navigating government programs and accessing resources. More broadly, they can look at economic investment, job creation and new scientific discoveries that drive the economy.
“Have we been able to expedite things like the review of public comments or requests for information?” Das says. “Are we able to do things more quickly, using fewer people and resources? Have we expedited some scientific discovery that would’ve taken years if we didn’t have AI?”
The answers to those questions will help agencies track success.
MORE FROM FEDTECH: The Navy wants to prove zero trust’s ROI.
Adapting to Evolving Needs and Challenges
FASST can help agencies to adapt to a changing landscape by accelerating AI deployments.
“In national security, it might be about using AI to make sure our cybersecurity is as robust as it needs to be,” Fleming says.
Civilian agencies stand to benefit too. For example, AI can help solve pressing transportation problems, Das says: “How can we use it to make our roads more efficient? How can we use it to speed up our ability to build high-speed rail?”
“Agencies can learn from each other how to implement new programs, how to work through their necessary bureaucracies and regulatory processes,” she adds.
WATCH: The National Science Foundation is growing a national artificial intelligence innovation ecosystem.
That’s especially useful for responding to new challenges.
“FASST will help keep the U.S. economically competitive because of all of the things that we don’t even know about that will come out of that investment,” Fleming says.
