State agency wants to spend $10M exploring use of artificial intelligence

LANSING — Traffic lights that turn red or green in response to traffic flows, rather than a pre-programmed schedule.

Systems that can scan signatures on absentee ballots and flag those that don't seem authentic.

Chatbots that can respond to residents' questions at 3 a.m., when most state employees are asleep.

Those are all potential uses for artificial intelligence — and, more recently, "generative AI," which can generate text, images, code, or other outputs, and in some cases is capable of a form of learning and problem solving — inside state governments. The technology is generating both excitement over possibilities and concerns about issues such as privacy protection and potential impacts on the state workforce.

Michigan's Department of Technology, Management and Budget wants to spend about $10 million to explore how AI might best be used to improve services to Michigan residents. That proposal, which would include pilot projects and other studies, is part of Gov. Gretchen Whitmer's state budget proposal for 2025.

"The opportunities are boundless," said Laura Clark, who oversees state technology as chief information officer in DTMB. "But also, we have to be hyper-ly focused on protecting the state systems and the state data that we have."

Rather than be on the "leading edge" of AI use in state government, Michigan wants to "be on the leading edge in learning," and start cautiously with pilot projects in which personal data and broader state systems are not at risk, Clark said.

Today, for example, state workers are permitted to use a free-to-use AI system such as ChatGPT to generate reports, if they wish to, but only if all the information used in such a report is already in the public domain, she said.

Simpler forms of AI technology have been around in state government for decades, from spell check on computers, to email programs that suggest possible responses to messages received, to virus scans and blocking of suspicious websites for cybersecurity, to data analysis tools.

Some uses in recent years, such as the MiDAS (Michigan Integrated Data Automated System) used to automatically detect cases of unemployment insurance fraud, have been disasters, with tens of thousands of Michigan residents falsely accused of fraud. The state recently settled one lawsuit related to that debacle for $20 million. The Michigan State Police's use of technology that could be used for "predictive policing" has also been controversial.

Clark said the exploration of future uses of AI is at such an early stage she couldn't comment on whether it might impact the future size of the state workforce, which is now around 50,000 employees.

Ray Holman, a spokesman for United Auto Workers Local 6000, which represents about 16,000 state employees, said workers always get concerned when the state explores the use of new technology, and not just because of the potential of lost jobs. In the past, the state has introduced costly systems that are difficult and time-consuming for state employees to use and not especially helpful for the citizens they serve, Holman said. He cited a glitch-ridden child welfare software system called MiSACWIS (Michigan Statewide Automated Child Welfare Information System).

"This is stuff there should be more dialogue with the union about," Holman said. "They need to work with us."

Clark said discussions with state agencies show the areas where there is most interest for using AI today involve chatbot technology, data analysis, document and video processing, and coding work that can modernize and extend the life of older IT systems that would otherwise have to be replaced.

Michigan government already uses chatbot technology on the state's website, michigan.gov, and has recently called for proposed solutions from vendors to enhance and expand the use of that technology, which can answer questions and perform other tasks in response to requests from residents, possibly to include translation services, Clark said.

"Today, several chatbots are being leveraged across specific michigan.gov websites," DTMB said in a request sent to vendors. "Multiple products are being used with varying levels of success depending on the technical capabilities of the chatbot and the quality of the curation. Transitioning to a single michigan.gov chatbot implementation will ensure that there is consistency across all websites."

As the technology improves, the threats increase, Clark said. For example, hackers who try to malevolently infiltrate state systems by using "phishing" emails sent to state employees are making use of AI to create phishing emails that are much more realistic and easier to fall for, she said.

Many states and counties outside of Michigan use AI technology as a tool to verify signatures on absentee ballots, Reuters reported.

In Michigan, the Department of State "continues to track the many different benefits and risks that AI presents to our operational work, but we currently don’t have any plans to pursue any AI-based technologies," said spokeswoman Cheri Hardmon.

The Michigan Department of Transportation has at least three AI pilot projects in the works, spokesman Jeff Cranson said.

Under one project funded by a federal grant, special cameras and other technology would be used to warn bus drivers and other equipped vehicles of potential hazards such as pedestrians and cyclists, Cranson said.

Another system would use flashing road signs to alert wrong-way drivers on ramps to I-75, I-375, and I-696, he said.

A third project, also involving cameras, is intended to improve safety for workers in highway construction zones, Cranson said.

Contact Paul Egan: 517-372-8660 or pegan@freepress.com. Follow him on X, @paulegan4.

This article originally appeared on Detroit Free Press: State wants to spend $10M exploring use of artificial intelligence