IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Generative AI Is Only as Good as the Models It’s Built On

At the NASCIO Midyear Conference, Maryland CIO Katie Savage said the state is still early on in its use of generative AI, gathering and cleaning data and building models to make it an effective tool for government.

Maryland CIO Katie Savage.jpg
Katie Savage
Government Technology/David Kidd
NATIONAL HARBOR, Md. — Automation is not a new topic in gov tech circles, but the recent buzz around generative AI tools like ChatGPT has perhaps turned the conversation in a new direction.

While opinions on how much generative AI will impact government — and whether it’s for better or worse — ran the gamut at the NASCIO Midyear Conference this week, CIOs and other state IT leaders agreed that these are early days for the emerging technology.

Brian Tardiff, CIO and chief digital officer of Rhode Island, stressed the importance of having good policy in place before rolling out any new tech, including generative AI.

“There is an appetite from our agencies to leverage chat agents with Health and Human Services, Division of Taxation, Department of Labor and Training," Tardiff said. "But we have not landed on a policy yet to even help us explore what tool would be preferred at this point.”

Iowa CIO Matt Behrens said his state is still learning about what the use cases for generative AI might be, but he was optimistic.

“We’re continuing to look at the tools and there’s lots of opportunity that I see,” he explained, “but I think we’ll proceed with some caution until we understand some of the capabilities and some of the guardrails as we think about it going forward.”

That cautious approach is also the case in Maryland, where CIO Katie Savage said they’re currently in the “leveraging data to be able to create good models phase” of exploring generative AI. Because the technology needs to be built on strong data models, those models need good data.

Savage also highlighted the continued importance of humans to monitor the AI and ensure it’s doing what the state intends without negative ramifications.

Lauren Kinkade is the managing editor for Government Technology magazine. She has a degree in English from the University of California, Berkeley, and more than 15 years’ experience in book and magazine publishing.
e.Republic Executive Editor Noelle Knell is a contributing editor to Emergency Management magazine.