We are on a precipice of game-changing technology. Today’s AI apps will disappear, just as early social media did, but the technology is here to stay. In fact, you’re probably already using it. At light speed, AI is being embedded in tools used daily by government, including the entire Microsoft Office suite.
It’s been 15 years since hackathons, meet-ups and unconferences helped state and local government experiment with Web 2.0 technology. We didn’t care that so much of government still needed fixing — we saw light, user-centered, open technology as a vital tool to help us get the basics done faster. AI could be the same, if we embrace it.
There are valid worries about bias and hard-coded inequity in AI, but the mood seems to be that cities should focus on the basics of good content and well-designed services, and view AI with cynicism or open disdain. I strongly disagree with those urging caution and restraint.
We have seen what happens when government lags on technology, and corporations behave irresponsibly. Government doesn’t have to repeat this mistake. The digital government movement must get out ahead of this turning point. Build tools that embed equity, test and create principles for constitutional AI, and prove out use cases that show what human supervision looks like.
Right now we are at the early stages of AI, and seeing some fun but basic applications for the general public. Mostly it’s being used to answer our questions, or to do our busywork. Much more exciting will be the meaningful things AI can do for us and our constituents.
Here are five things that seem totally possible now, soon, or eventually.
(1) Build your digital services
- Make your content 5th grade English. AI can do this right now.
- Accurately translate your content into multiple languages better than the current tools like Google Translate. We’re very close to having high quality human-like translation.
- Eventually build out highly personalized user journeys based on clusters of services that go together, automated by AI.
- Even more eventually, let users navigate services through voice, and chat interfaces, done largely through third party AI assistants. Eventually people will get used to interacting with services this way. You can already use these to book you a table at a restaurant. This is not a call to design your own City chatbot (please don’t) but spend more time considering how your services will talk to these AI intermediaries and less time worrying about your homepage design.
(2) Create policy
Help you quickly pull together user interfaces and model services to test with residents. Think: upload a policy document and see an AI turn it into a prototype service. Test it out, tell the AI what to change, and watch it iterate. Use this prototype to procure the full service, cutting down time on procurement requirements gathering.
Simple prompts like ‘write a policy on AI’ will get you nowhere, but prompts like ‘Imagine you are a policy officer for a county. Write a policy governing the use of AI in the workplace that allows people to use AI to assist with work, but not if it causes discrimination’ will yield something you can work with. Of course we don’t know how much of Biden’s Executive Order regulating AI was written by Chat GPT, nor The Bletchley Declaration.
(3) Plan your city
Use generative AI to dream up public spaces, community facilities, and housing solutions, all while applying the constraints of zoning and planning code. Use these as inspiration, or to test public reaction. Consider prompts like ‘You are a nimby / yimby, design a new development that houses 200 families, with at least 50% below market rate housing’.
Have the public work with AI to dream up their ideal cities. Use AI to model the impact of new scenarios on traffic, quality of life, air quality, and more.
(4) Inspect things
Computer vision is when computers look at things (pictures, videos, the world) and interpret what is happening. It’s already being used in the construction industry to identify when maintenance is needed on equipment, and in agriculture where drones identify crop health. Imagine building inspectors using computer vision to analyze videos of construction sites, or the placement of solar panels on roofs, without the need for in-person inspections.
Computer vision can also be combined with generative AI to assess things and design solutions. Imagine a person with a disability needing assistive equipment and adaptations around the home. Uploading a video walk-through of a home, and using gen AI to create a plan for adaptations is a potential future that can help occupational therapists, and families supporting their aging parents.
(5) Handle some customer services
Someday companies could have AI answering calls 24/7 in any language, raising tickets, and providing tier 1 support for just about anything. For local government, this means high frequency requests like graffiti reports. There’s an interesting potential future where an AI assistant calls 311 on behalf of a resident, and is answered by an AI agent which fulfills the Tier 1 request.
Services like grants and permits require some up-front time with city staff before the constituent submits an application. What if these nuanced conversations could be handled by AI? Uploading a draft permit or grant application to be checked and even iterated by AI could speed up the process for everyone. For now, humans will supervise these kinds of AI tools. Eventually they won’t need to.
Maybe all this excites you, and you’re buzzing with possibilities. Maybe it fills you with fear, as it should do a little bit. Maybe you think it’s all hype and dystopian - if this is you, I urge you to put your healthy skepticism to one side and engage with AI to see if you can help chart an ethical and pragmatic path.
We should absolutely be conscious of the risks: models that replicate and amplify human biases, hallucinations that give residents inaccurate information, and inevitable cybersecurity risks are just a few things that we have to anticipate. If these things worry you - prototype some solutions. Try to create your own AI apps, and see what problems arise, and figure out how to fix them. We have an opportunity to give constituents fairer, more user-centered services, built with safety at their heart.
Much of this will require a whole lot of data. Perhaps you have a great use case in mind, but you don’t have enough data to train a model. This is where government silos need to collaborate – between departments, between cities, and between the different levels of government. Share training data, share ideas, and share use cases. There won’t be many local government AI startups, so the sector will be beholden to big consultancies who will try to convince cities and states that they have the silver bullet. The way to avoid this trap is to become conversant in AI, call BS, and prototype the use cases you want to procure.
There’s a lot to be excited about. And while we should definitely be conscious of the risks at scale, there is nothing to fear in this prototyping phase. If we don’t get messy with AI, we will miss the opportunity to shape its impact. Now is the time for bold experimentation. Let’s have some fun!