The Value Proposition of AI
It comes down to two things: AI automates or predicts. This is obviously going to drive many downstream applications, and in reality, the application of an organization's data is endless.
Mike Wiseman
Vice President, Americas Public Sector Sales, Pure Storage
The organization needs to use data science everywhere, across all of the company. They also need to know the limitations, the need for more accurate data, explainability and knowing how the AI makes the decisions.
Curt Smith
CP, GPU Datacenter Architect, NVIDIA
From predictive maintenance, to faster satellite intelligence and an optimized federal response to wildfires, agencies see artificial intelligence as an essential tool to meet their data-driven missions.
Achieving a strategic advantage through AI, however, requires as much commitment and investment in an agency’s data strategy, as well as empowerment of the workforce and data scientists.
Mike Wiseman, vice president of the Americas for the public sector at Pure Storage, said federal applications of AI include analyzing battle damage and corrosion on maritime equipment. What’s common to federal applications of AI, he added, is accelerating the time to insight.
“It comes down to two things: AI automates or predicts. This is obviously going to drive many downstream applications, and in reality, the application of an organization’s data is endless,” Wiseman said.
Curt Smith, vice president and GPU data center architect for the accelerated compute product group at Nvidia, said agencies see AI as an opportunity to “improve the signal-to-noise ratio for someone who needs to make a decision in real-time.”
For all these opportunities in AI to succeed, however, agencies must also invest in greater computing capacity and data infrastructure.
“The tactical side, it sounds really simple. Behind that, though, it needs to be recognized, in each of these organizations, that they have big AI model training tasks to undertake. That’s going to take a fair amount of compute, a lot of storage, fast networking and a commitment from the organization to drive through projects where they’re experimenting, putting AI into their products, and creating a culture that believes in the power of large AI to solve these kinds of problems,” Smith said.
Smith said agencies successfully use AI for tasks that include image classification, text and document processing and natural language processing.
“It’s being used to solve those types of problems, or to create systems that can process that kind of data. What we see is it can’t be done with normal algorithmic approaches,” Smith said.
In terms of balancing AI opportunities with the cost of upgrading computing capacity and associated IT and data infrastructure, Wiseman said agencies should recognize that AI isn’t the solution to every problem.
Agencies, he added, should also have an enterprise data strategy in place before making large-scale AI investments.
“Any time an agency has significant amounts of data and wide-reaching problem statements, they should review their data strategy and find a repeatable process that they can benefit from through an automated and predictive workflow powered by AI,” Wiseman said.
Wiseman said federal opportunities for AI include an acceleration of intelligence data and insights for warfighters, or optimizing federal response to natural disasters such as wildfires or hurricanes.
“Both of these applications require massive amounts of unstructured data to train and develop a production-grade AI. What interests me is how we create an easy-to-use, end-to-end infrastructure that can quickly process this data and produce a meaningful outcome that helps save lives at home and abroad. For us, it’s about accelerating and development of augmented intelligence and data fusion capabilities for public sector organizations,” Wiseman said.
To execute on the goals of an agency data strategy and build out enterprise-wide applications of AI, Wiseman said agencies must rely on the expertise of data scientists to develop AI systems with a high degree of accuracy and confidence.
These data scientists must also develop deep-learning algorithms that enable AI systems to learn and adapt to the environments in which they’re deployed.
“Developing production-grade AI is a challenging process with several nuances. You have to incorporate data strategy and data engineering, dev and testing. You have to incorporate data strategy, data engineering, development, and testing, deployment and sustainment into a well-thought-out initiative. Not considering AI infrastructure through this process introduces risks, and could derail or delay a project,” Wiseman said.
Smith said agencies need to consider training the workforce around a “new way of thinking about data,” and improving the organization’s overall data literacy.
“The organization needs to use data science everywhere, across all of the company. They also need to know the limitations, the need for more accurate data, explainability and knowing how the AI makes the decisions,” he said.
Smith said agency leadership also needs to set the right tone from the top and encourage a “culture of experimentation” from their workforce.
“The top of the organization needs to believe in the power of AI to transform the experience of the people that are executing the missions,” he said.
Listen to the full show:
Copyright
© 2024 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.