“Author’s” Note: This month’s column is Part 1 of focusing on AI in local government. The following article is based on the human author’s prompt to the generative AI application, Claude.ai, to write a “750-word plain language article explaining what a municipal government elected official should know about AI and generative AI.” Claude wrote 747 words.

As is appropriate for many AI outputs, the author subsequently edited the text to add context, accuracy, and style. The edits are shown in green underlines and struck out text.

Artificial intelligence (AI) is transforming how organizations local governments operate and provide services. As a municipal leader, it’s important to have a basic understanding of AI, including the newer generative AI models, to effectively leverage these technologies for your city.

AI refers to computer systems that can perform tasks normally requiring human intelligence, such as visual perception, speech recognition, and decision-making. AI encompasses a variety of techniques that allow software algorithms (formulas) to learn from data to make predictions, recommendations or take actions.

AI systems “learn” through M machine learning (ML). ML is a commonly used technique where algorithms are trained on large sets of example sample data . This data can include the content of the publicly available internet or an organization’s own stored data. The training teaches the AI program to recognize patterns , make connections, and make decisions. The training creates what are known as “neural networks.”

For instance, AI can analyze photos to identify signs of infrastructure damage or review documents to route them to the correct department. AI is also used in chatbots that understand natural language questions from residents.

Generative AI takes things a step further. Models like DALL-E and GPT-3 can generate brand new content like images, text, audio, and video based on short text prompts. This ability to autonomously create original, realistic artifacts content has enormous potential — and risks.

Claude.ai is the generative AI chatbot that wrote this article in response to the question (aka a prompt ). The more well-known ChatGPT is a competing chatbot system.

Human Author’s Comment:

Claude missed or ignored a key point. An element of how generative AI works it that the program uses its knowledge to predict the next word in responding to a prompt. This is the way the human brain works. But there is a catch.

Chatbot outputs have been known to invent information or “hallucinate” that it knows an answer but really doesn’t. When this happens it is choosing words that it thinks make sense but may not reflect reality. This effect initially surprised developers. They are now working on ways to minimize hallucinations. Thus, use of chatbot output must be carefully considered and reviewed by a human before it is used.

Back to Claude.

How can municipal officials harness AI safely and effectively? Here are key considerations:

1. Start small. Look for focused ways to improve specific processes vs. jumping into major AI deployments right away. Pilot projects in areas like service requests, inspections and document processing can demonstrate benefits and build trust.

2. Focus on augmenting staff. Position AI as assisting rather than replacing staff. For instance, chatbots handle common questions to let staff focus on higher-value interactions. AI should empower employees to do their jobs better.

3. Assess risks. AI bias can lead to unfair or dangerous outcomes. Rigorously test for impacts on different populations and mitigate risks before deployment. Document processes to ensure transparency.

4. Control generative AI. While creative, generative models also carry significant risks around misinformation, security, intellectual property, etc. Set strict governance policies on how and when generative AI can be used.

5. Acquire cautiously. Be wary of vendor hype and don’t assume AI is the solution. Carefully validate capabilities and start with a limited pilot. Watch for exploitative or unethical business practices.

6. Plan for change management. Adopting AI changes processes and job roles can meet resistance. Involve staff early in evaluating AI. Invest in retraining and upskilling employees to work alongside AI.

7. Build smart data practices. Quality data is essential for effective AI. Audit existing data for biases and gaps. Improve data collection going forward to better reflect the full community to reflect the full community accurately. It is also critical that any use of generative AI safeguards confidential information or data with personal identifiers. That information should never be uploaded to the cloud unless precautions are taken to ensure they are not ingested into the chatbot’s system.

8. Focus on public benefit. Apply AI to improve life for all residents fairly. Make decisions transparent. Protect privacy and security. AI offers great potential but make sure it aligns with public sector values.

9. Stay engaged. Monitor AI legislation and regulation which will likely expand. Participate in policy discussions to advocate for municipal needs and ethical AI standards.

Agencies with robust in-house technology skills can directly engage with vendors such as Microsoft, Google, and Amazon Web Services to support AI interests. For smaller organizations there is a wide range of dedicated municipal application providers. Many of them are integrating generative AI and other AI applications into their existing products. Thus, Claude’s item #5 above has added importance.

The public sector will lag behind adopting AI day-to- day use. But by understanding capabilities, limitations, and responsible use, municipal leaders can tap AI’s power to enhance services, efficiency, and equity across their cities.

The idea for this article was itself prompted by the author’s recent Op-Ed in NJ Spotlight, “What State and Local Governments Should Do About Generative AI."

An earlier version of this article appeared in New Jersey Municipalities, the magazine of the New Jersey State League of Municipalities.

MARC PFEIFFER, an ICMA Life Member, is a marginally retired New Jersey town administrator and state agency manager. He is currently a senior policy fellow and assistant director at Bloustein Local, a unit of the Center for Urban Policy Research at Rutgers University. (marc.pfeiffer@rutgers.edu)