Could technology simulating human intelligence and likeness cause us great harm? Could it become an everyday problem-solving tool used by city agencies and businesses to improve our lives?
Tech experts say that both of these things can be true — and in many cases, they already are. What's important is getting ahead of the curve, according to city and state officials actively working in AI regulation.
On May 16, members of the city's independent Audit Committee, including Denver Auditor Tim O'Brien, gathered for their monthly meeting and discussed the current regulatory landscape of AI and the effect it could have (and already is having) on city operations.
One day later, on May 17, Governor Jared Polis signed Senate Bill 24-205 into law, requiring developers and deployers of "high-risk" AI systems to use "reasonable care" to avoid algorithmic discrimination when in use. This must be done through a number of recommended procedures, such as implementing a risk management policy and program for the high-risk system, and disclosing information to the attorney general and known deployers of the AI about any "known or reasonably foreseeable risk" of discrimination.
On May 20, Denver leaders and AI experts gathered for a public panel at the University of Denver to talk about risk management and the current use of AI systems to improve existing city services, such as 311.
"Cities are the first place where all this starts," Suma Nallapati, Denver's chief information officer, told those in attendance.
For 311, there's now an artificial intelligence known as "Sunny" available to customers through the city's instant chat system, which can respond within seconds and speaks 72 different languages, Nallapati noted. Chatbot can provide a comfortable environment for people who would be less inclined to call and speak with a human, she said.
The use of AI could extend to other areas of Denver life, saving city staffers time and money that would normally be spent on hiring and employing people to do "soul-sucking work," as some city officials call it.
Permitting, for example.
"There's a lot of bureaucracy in the work," admitted Jeff Dolan, chief strategy officer and counsel for the City of Denver, during the May 20 panel. "One of our goals that we had [for the year]...is cutting permitting time in half. AI would be an amazing tool."
According to Adeeb Khan, executive director of Denver Economic Development and Opportunity, AI could help deal with the weight of the city's ongoing migrant crisis and better inform residents and newcomers about immigration and legal issues.
Suma Nallapati, Denver's chief information officer, was one of four local leaders to speak at DU about AI this week.
Daniels College of Business
In order to get work authorization, Khan noted how some migrants must first apply for asylum, which can be an "extremely complex" process. A large number of Denver migrants also go through immigration proceedings with U.S Customs and Border Protection.
"It requires one-on-one legal support," Khan said. "It could take twenty to forty hours to fill out those applications. The mayor and his team are looking...at AI solutions that exist right now that are able to help individuals navigate that. It’s a great way that we can use intelligence to be able to answer these questions in a way that allows people to navigate the legal nature of what an asylum application is."
Al Gardner, executive director of the city’s General Services department, told the DU crowd that what intrigues him most about AI are the energy- and cost-saving aspects. "How do we learn to use predictive analytics in a way to measure the investment that we need to make in solar, in gas?" he said. "How can we run the city from a structural perspective a lot more lean?"
While there's a lot of positives regarding AI, experts and local leaders would be remiss if they ignored the potential risks and drawbacks, which is what Denver's Audit Committee is looking to address.
During its May 16 meeting, committee members asked Information Systems Audit Manager Peter Ulrich — who was leading the AI presentation — a number of questions about the risks associated with AI, ways to mitigate those risks, and how AI will shape the future of city audits and the way they're carried out.
Negative uses, as pointed out by Ulrich, include social engineering, the creation of political "deepfake" videos, cyber attacks, committing crimes such as forgery and financial fraud, the spread of misinformation and false news stories, and attacking and manipulating people.
One of Nallapati's biggest concerns with AI right now is making sure discriminatory responses and inherent biases don’t go unnoticed.
“These large-language models are still learning,” she said during the DU panel. “So how do you ensure it’s staying within boundaries you create? How do you ensure that its learnings are human-centered?"
Looking ahead, Nallapati expressed how vital it is for city leaders to study AI's impact in the coming years and see if Denver is on the right path. "Course-correct as we go,” she suggested.
The "people aspect" is also important, according to Nallapati, and "ensuring every voice is represented in the right way." One example that she brought up is how large-language AI models don’t recognize certain important aspects of people’s cultures, citing her own Indian background as an example.
“In India, we wear a dot on our forehead,” she said. “The large-language model does not recognize that.”
Nallapati noted how discrimination and inherent biases can exist in large-language models due to human error when a person is uploading the language. “You have to make sure every voice is represented,” she added.
Another major fear of AI is putting people out of jobs, which is something DEDO and Denver leaders are looking to address head on through the idea of upscaling and up-skilling.
"As we think about the jobs in the future and the effects of AI, we as individuals are going to have to learn how to work as effectively as possible with AI, just like we’re doing with internet right now," Khan explained on May 20. “As a city, what we need to start thinking about is upscaling, because we're going to have a lot of individuals whose jobs may be obsolete in a few years."
According to Khan, city leaders must provide new, effective pathways and an infrastructure to achieve this.
"Yes, it’s kind of scary, because jobs will probably become obsolete," he warned. "That happens with every new emerging technology. However, it’s how we adjust and do that effectively and take advantage of the technology to create more prosperity that’s important."
Information Systems Audit Manager Peter Ulrich went through a presentation on AI for the city's independent Audit Committee on May 16.
Denver Auditor's Office/Channel 8
With SB 205's guardrails not set to take effect until 2026, city officials have wondered if local agencies and businesses should wait to use human-simulated technology until the digital landscape is safer. Audit Committee members, meanwhile, fear what it will take to get entities to comply with using or not using it.
"We've seen in audits certain groups are not willing to change," said one committee member. "How do we incentivize government to be more adaptive to new technologies?"
According to Ulrich, the key is maintaining a strong relationship with the city's Technology Services department and making sure auditors are working with individual stakeholders to show value and create buy-in.
"They're very open to exploring the possibilities of artificial intelligence, and they're starting to take the right steps," Ulrich said of the city. "It starts there. Finding places to show that value and creating that snowball effect. When you're thinking about change in a large organization, it's very difficult to lay out, 'We're going to start using AI everywhere.' And that's where the governance comes in. Targeting the approach and a strategic plan."
While not every performance audit conducted by the Auditor's Office will be dealing with AI, Ulrich says city leaders should be aware of and prepare plans of action, so they can ask questions like, "Can AI make this process better?" and "Is it reasonable to ask that business unit [to use AI]?"
"A system may not solve your problem. A good process is still a good process; you don't necessarily need a technology system to help it. But 'Could that process benefit from either technology or AI?' should be a question auditors are thinking about," Ulrich explained.
It's certainly one that Nallapati and the city have been thinking about as they head into the future and get department staffers ready with AI and cybersecurity training sessions.
"We are ensuring and tracking that everybody takes the training," Nallapati told Westword after the DU panel. "It's very important in our training that we look at not just the technical aspect of it, but the business impacts, the resident impacts, the data impacts, of everything. So that's why our training is very measured. It's very targeted to ensure that more people understand the repercussions of technology."
