A Human’s Take on ChatGPT
I can tell you that I wrote every word of this article at my keyboard with our family dog looking on.
You might never know otherwise because the quantity and quality of computer-generated content is growing rapidly. ChatGPT and its Generative Artificial Intelligence cousins work daily, outlining marketing plans, writing email templates, drafting computer code and so much more.
ChatGPT and Generative Artificial Intelligence
The public preview of ChatGPT crossed over into popular culture in November 2022 and will almost certainly launch 50,000 new startups this year. Many have said it already: this will define the future of search and the future of work. Thousands of companies are reexamining their strategies and thoughtful approaches to AI, and your city should too.
NLC’s Director of Sustainability and Innovation, Lena Geraghty, published a post in April 2023, “A Conversation with ChatGPT on Local Governments, Emerging Technologies,”where she invited ChatGPT to answer this question, “What should local government leaders know about you?” The response:
I am designed to understand and process human language to generate responses and provide information. Local government leaders can leverage my abilities to gain insights and make data-driven decisions. Here are a few things that local government leaders should know about me: I am trained on a vast amount of data: I have been trained on a massive dataset of diverse texts from various domains, including news articles, books, and academic papers.
Chatbots like ChatGPT and Bing Chat are trained on Large Language Models. If pointed to Wikipedia, they ingest and organize what Wikipedia covers. If pointed to US Tax Code and Regulations, they learn what the Tax Code contains. And, if pointed to your city’s ordinance, policy, budgets, and transactions (e.g., permitting, licensing, inspections, emergency response, and more), then they know a lot about your city and its daily operations and will respond with corresponding insights and specifics.
City Leaders Must Know: There are free and paid versions of generative AI chatbots. The paid versions can be further trained with your data and configured to reflect your voice and policies. Both are useful.
For just five minutes, let’s play and learn with ChatGPT.
- Point your browser to https://chat.openai.com.
- Click Sign-Up and create an account.
- First, one for fun. In the “Send a Message” prompt, type the following: Write a poem about my dog Spot. He’s brown and white and loves chew toys.That was cute.
- To change to a more serious topic, click New Chat (top left).
- Type the following in the Send Message prompt: I am an elected official of a medium-sized city in Florida. What state programs should we access to assist my community’s unhoused population? Make the response brief. Format in a table.
- Press Enter and watch for your personalized response. Mine looks like this:
- Now, enter the following (a continuation of the previous prompt and conversation): Which of these programs emphasize assistance for school-age children?
- Press Enter.
The “Regenerate Response” requests another, slightly different, answer.
Try different personas and different prompts. Remember to click New Chat (top left) when changing topics.
City Leaders Must Know: These Prompts and Responses are not private. In fact, the model learns from its previous interactions. To start a new thread of questioning, click New Chat.
The prompt above tiptoes into a potentially sensitive topic that warrants thoughtful consideration: the personal experiences of those experiencing homelessness. Yes, there are biases in generative AI just as there exist biases in its underlying training data and our own individual decision making. ChatGPT does work to warn or block unsafe content. The guard rails are not perfect but are part of a continuous effort termed Responsible AI.
In the world of artificial intelligence, “hallucinations” are misfires. These are cases where the model responds incorrectly (yet with authority). I’ve seen examples where the chatbot “invented” web site URLs that did not exist. This is why responses should be validated as accurate.
City Leaders Must Know: Although generally safe, it is possible for Prompts and Responses to violate your personal and city standards for respect and kindness.
City Leaders Must Know: Do not rely on these tools blindly. They can (and do) sometimes misrepresent facts.
Would we discourage an author from referencing a thesaurus or Google? Of course not. The same is true of ChatGPT. There are so many examples of good, productive uses which should be encouraged.
Staff can benefit by asking the tools to generate catchy social media headlines. The AI can summarize a historical event or even complex legislation. What better place to seed a brainstorming session than with a provocative ChatGPT prompt?
City Leaders Must Know: Carefully consider both the benefits and risks as you explore how to leverage the tool and regulate its use. Organize training and establish a policy and encourage staff to get creative.
Microsoft recently announced Office 365 Copilot, a promise to integrate generative AI into its Office 365 suite, including Word, PowerPoint, and Teams. Google recently announced “magic wand” for its popular Google Docs.
It is clear that generative AI is now part of our professional and political landscapes. Its role in city offices and its impact on city services is not fully known but is bound to be remarkable.
Note: This article was first published on the National League of City’s CitiesSpeak blog.