A panel of government and industry experts on artificial intelligence (AI) technologies is recommending that Federal government leaders focus on good data structure and governance policies as government agencies increasingly seek to put AI tech into play.
Ensuring data is well structured using comprehensive data governance across repositories is critical to integrating the use of AI in government organizations, the panelists shared at a July 24 AI event organized by GovLoop.
The panelists ticked off a number of recommendations on the data organization and management front, including organizing data outside of administrative files, such as PDF files and other documents.
They also said that strong leadership and sharing of expertise within the AI community and government organizations is necessary for good data management.
Without well-organized data, AI systems won’t work as efficiently, the panelists warned.
“If you’re not utilizing all of the tools that are in the private sector and the public sectors about how to set up those programs and how to use metadata effectively, and how to how to classify the data – if you’re not using governance to monitor all of those activities, you’re really not going to achieve any goals,” said Mark Wheeler, the director of AI, data & analytics at the General Services Administration’s Centers of Excellence.
Nick Psaki, the federal principal technology strategist of Pure Storage, said that data management includes knowing how and what data to store as generative AI becomes more frequently used in the workplace. He said that a challenge will be learning how to store the mass amount of data produced.
“When we’re doing AI training the data set explosion is ten to 100x,” said Psaki. “Some estimates in the industry are there will be more AI models than there are people on Earth. So, eight billion artificial cognitive synthetic machines are creating new data, where do we park that stuff?”
“These are the things that we really have to think our way through because the big explosion is going to happen at a rate that defies imagination,” he said.
Some AI tech can be used to manage administrative tasks while other AI systems can be trained and used for different government departments or functions, panelists said.
Gaine Nazareth, the solution consulting leader for ServiceNow focusing on the U.S. Defense Department, recommended that integration use either a top-down approach – by examining all initiatives and programs – or a bottom-up approach – by interviewing leaders and experts – to identify where improvements are needed. This can help organizations understand where AI can improve operations and efficiency, he said.
Other recommendations shared by panelists include being patient as technology improves and becomes more specialized, developing domain-specific AI tools to address data and security concerns, and collaborating between government organizations and AI experts.
