Federal civilian and defense contract spending for artificial intelligence and machine learning (AI/ML) will reach approximately $3 billion this year, according to Bloomberg Government.
This may sound small compared to the worldwide market for AI, estimated at $327.5 billion in 2021, according to International Data Corp. But, there are a series of indicators that point to rapid and significant AI growth in the Federal government, including:
- Creation of the National Artificial Intelligence Initiative Office within the White House Office of Science and Technology Policy
- A proposed $90 billion in Federal government-directed research and development (R&D) spending in technology areas including AI. The funding is part of President Biden’s $2 trillion American Jobs Plan
- Elevation of the Department of Defense’s (DoD) Joint AI Center (JAIC) so that it reports directly to the Secretary of Defense
- Introduction of the bipartisan Endless Frontier Act, which would authorize $100 billion over five years to strengthen research and development in areas including AI
- Introduction of the bipartisan AI Scholarship-for-Service Act, which would provide scholarships to college students studying AI and related fields in exchange for service in the public sector
- Launch of AI.gov, a government clearinghouse for information about AI in the Federal government
These actions and many others demonstrate the growing importance – and even urgency – of AI investment by the Federal government. In its 2021 report, the National Security Commission on Artificial Intelligence (NSCAI) called for the U.S. government to invest $200 billion in AI over the next 10 years.
“The United States must act now to field AI systems and invest substantially more resources in AI innovation to protect its security, promote its prosperity, and safeguard the future of democracy,” the commission said.
While planning and proposals continue, Federal agencies carry on with AI initiatives large and small.
Every large federal agency is executing on one or more AI proofs of concept, pilot projects, or technology demonstration projects. Agencies with mature data science practices are pursuing more complex AI implementations, including robotic process automation, fraud detection, and automated translation, according to Anil “Neil” Chaudhry, director of Federal AI implementations at the IT Modernization Centers of Excellence within the General Services Administration.
It’s important to start small, so agencies can ensure they have the infrastructure in place to support the intensive compute and storage requirements of AI implementations, said David Kushner, senior vice president of sales at enterprise IT solutions provider ViON. “To have the greatest success, agencies should start with one or two specific use cases and scale up once they’ve established some best practices,” he advised.
“A flexible infrastructure is essential to success”, said Chip George, vice president of public sector at Nutanix, which provides cloud software and hyperconverged infrastructure solutions.
“AI and machine learning can be unpredictable in their performance and demands,” he noted. “To make the right decision, AI and ML need to be exposed to more and more data – which means agencies are managing more and more data. Agencies need infrastructure that can flex with changing demands, so they’re not running out of storage or crashing servers.”
Data storage is also a key consideration for agencies embarking on AI initiatives. Because of the huge volumes of data that AI and ML applications require, agencies must carefully consider where that data will live. In some cases, operations will run efficiently and cost effectively in the cloud. In others, the cost of extracting data from the public cloud will be too great, and agencies will turn to on-premises storage and computing. Still others will turn to the hybrid cloud model.
“Hybrid cloud is going to be critical to support AI in an affordable and efficient way,” Kushner said. “ViON’s X-as-a-Service model enables organizations to move out of the sandbox faster – from test/dev into production. We can test, prove the model and move to implementation quickly. All AI models and programs are in continuous learning mode and the X-as-a-Service allows the organizations to scale. We also have 40 years’ experience working with government customers to plan their data requirements which are fundamental to AI initiatives.”
Kushner’s sentiment is echoed in Nutanix’s third annual Enterprise Cloud Index. Eighty-seven percent of Federal government respondents said hybrid cloud is their ideal IT operating model. Feds also said their plans call for more than doubling hybrid cloud usage within one year and expanding to 74 percent penetration within five years, up from about 14 percent today.
Feds said their top reasons for changes to their IT infrastructure were greater control of IT resource usage (59 percent), greater flexibility to meet organizational needs (54 percent); and better ability to support employees and other users (53 percent).
In addition to hybrid cloud, as-a-service AI can help agencies manage and scale their IT resources – and customize AI for specific use cases.
“The as-a-service model will enable agencies to scale as slowly or as fast as they want. That way, they aren’t over-buying or under-buying capacity; and they can ensure the projects align with their mission requirements,” Kushner noted. “I’m excited about the as-a-service model and the role it can play in helping agencies modernize IT.”