More than 100 mayors descended on Washington, D.C. this week to learn how generative AI tools like OpenAI's ChatGPT could help them better run their cities.
The mayors sought to learn how the cutting-edge technology could help them do everything from better predicting which areas would be most affected by natural disasters to making it easier for residents to navigate city services.
The roughly four-hour Mayors Innovation Studio, hosted by Bloomberg Philanthropies' CityLab on Wednesday, is an example of how generative AI tools are making their way into many aspects of life and every level of government. While only 2% of cities surveyed by Bloomberg Philanthropies said they're actively implementing the technology, 69% said they were actively exploring or testing it and 96% of surveyed mayors said they were interested in using it.
In Washington, where much of the discussion of AI has centered around how Capitol Hill should place broad guardrails on the technology, the session provided a glimpse into how local governments may be among the first to harness the power of AI to serve their constituents, even as federal lawmakers debate lofty principles.
"Cities are places of action, where new solutions hit the ground," said James Anderson, who leads government innovation programs at Bloomberg Philanthropies, in a phone interview prior to the event. Cities are "the last mile and often the first mile in terms of the innovations," he added.
Policies governing tech have increasingly become the domain of state and local governments as Congress has failed to pass many major tech bills, like those protecting digital privacy or creating guardrails for kids on the internet. Meanwhile, states have taken such matters into their own hands, which many tech companies fear creates a patchwork of regulation that's difficult to comply with.
The focus of Wednesday's session was primarily on how generative AI may streamline processes for cities or offer new insights to make them safer or more efficient. The pandemic showcased the power of local governments harnessing data, Anderson noted, when many built dashboards of local Covid cases and hospitalizations.
Early in the session, Mitch Weiss, a Harvard Business School professor and former chief of staff to a former Boston mayor, demonstrated how the group might use ChatGPT to better understand and solve a local issue. Weiss used the example of "storrowing" in Boston, when trucks scrape their tops off while going under a low-clearance bridge.
Weiss prompted the chatbot to channel various experts to give their opinions and come up with solutions to reduce the issue and asked the AI questions like why warnings for low-clearance bridges weren't working. ChatGPT said drivers were distracted, unfamiliar with the area, or over-relying on GPS.
At one point, he prompted ChatGPT to create a line graph from an open dataset of such incidents in New York City, and many in the room wowed when a graph with a steep drop in incidents quickly materialized. He asked for a hypothesis for what may have improved New York City's storrowing compared to Boston, and ChatGPT suggested improved infrastructure, better signage, modern GPS and awareness programs may have contributed to the decline.
In one instance, just for fun, he asked ChatGPT for wacky solutions to the problem. The AI chatbot suggested a truck catapult. He also asked for more realistic solutions inspired by the wacky ones, and ChatGPT suggested a designated detour route.Â
Using AI for summer job programs and town halls
Later, Weiss prompted ChatGPT to create a form advertising a new summer jobs program with a city and target it in a way that would appeal to teens. The AI came up with the branding of a "Summer Hustle," and Weiss then prompted it to create a colorful graphic to promote the program.
Weiss also showed mayors how the tool could be used to prep for community board meetings, by asking the AI to generate possible questions from community members, including follow-ups.
Some mayors who said they've already played around with generative AI tools said they've used it to anticipate town hall questions, summarize articles they haven't had time to read, create draft job descriptions or draft responses to constituents.
CNBC agreed not to quote individual mayors who participated during the event, which was marketed as a place where mayors could come to learn and freely ask questions about a new technology.
The mayors also heard from several cities already deploying or thinking about the use cases of generative AI in their cities.
The city of Buenos Aires, Argentina, for example, is working on a generative AI model based on ChatGPT with its existing Boti chatbot that residents can text with using WhatsApp. The new generative AI version of Boti is trained to discuss culture and tourism — topics Melisa Breda, undersecretary for evidence-based policies, said they determined to be relatively low risk. Still, the tool hasn't rolled out yet as Breda said it still needs fine-tuning to ensure its responses fit their criteria.
Boston's Chief Information Officer Santiago Garces shared the city's basic guidance to its employees for using the technology: review any outcomes, disclose the use of AI and don't input sensitive data. Garces said such guidance should weigh the risks with the opportunity for employees in different parts of the government to experiment with it to determine how it can make their jobs more efficient.
Garces also said Boston is exploring how to use generative AI to translate information into specific regional dialects to help enroll residents in services.
Bloomberg Philanthropies and the Center for Government Excellence at Johns Hopkins University announced at the session a new City AI Connect platform, where city staff could continue to share ideas and resources on using AI in their governments.
The event sought to give mayors a starting point for how to think about implementing generative AI into their processes.
"We were hearing ... oh my god generative AI, everyone's talking about it everywhere," said Anderson. "We understand it could mean a lot for local government. We have no idea where to begin."
Mayors who spoke with CNBC around the event recognized the potential of generative AI to solve problems in ways that weren't previously possible.
"This is not a trivial thing," said Chattanooga, Tennessee Mayor Tim Kelly in a phone interview ahead of the event. "If we can get this right and people can use AI to better access city services, I think it could move the needle."
Kelly imagined that AI could help residents become more connected to their local governments, by surfacing open board seats that match their interests or simply making it easier for them to get information about recycling.
Columbus, Ohio Mayor Andrew Ginther hopes the technology can someday be used to better predict natural disasters and areas of high traffic accidents as well as streamline the process for residents to obtain government benefits for food.
"There's great promise with AI really to help us transform local government operations and service delivery," Ginther said. "We think it'll allow us to help for better frontline decision-making, more real time information ... We think those efficiencies are going to save city staff time and money and we think there are going to be dramatic improvements in resident customer experience. But we're also going to have to invest in it."
San Francisco Mayor London Breed, in an interview after the event, envisioned that AI could help the city identify forgotten and contradictory laws on the books, so they could work on clearing unnecessary regulatory hurdles for things like housing.
"We can't approach it from a place of fear," Breed said of AI. "It's coming whether we want it to or not. The question is, are we going to move with it and stay ahead of it to a certain extent, or are we going to fall behind and get run over? ... My hope is that we don't get to that place where we allow this technology to run away from us."
WATCH: How A.I. could impact jobs of outsourced coders in India
