Some Guardrails to Consider with ChatGPT 

Posted on August 9, 2023

An iPhone sitting on a granite countertop. The phone's screen displays ChatGPT.

Today’s Morning Buzz is brought to you by Rafael Baptista, Strategy & Innovation Director for Buncombe County, NC. Connect with Rafael on LinkedIn.

What I’m Listening to: No Rules Rules Netflix and the Culture of Reinvention by Reed Hastings Audiobook

A Hobby I Enjoy: Bike riding 

What I’m Working On: Moving houses with two kids under 3  

A popular topic of discussion in local government revolves around the use of AI tools like ChatGPT for various purposes. Some individuals are thrilled and have eagerly embraced the ChatGPT technology, while others approach it more cautiously but are still willing to explore its potential. On the other hand, there are those who harbor concerns and fear that AI will lead to dire consequences for humanity.

Personally, I find myself falling somewhere between the first and second group. In my personal life, I do use ChatGPT to assist in editing my written work. However, when it comes to using it at work, I’ve been more hesitant due to privacy and security concerns. Nevertheless, there are numerous informative articles available, and individuals within the local government, such as Micah Gaudet, have contributed valuable content on how to utilize ChatGPT effectively.

Today, I’d like to emphasize some important guardrails to keep in mind when using ChatGPT. Firstly, it’s crucial to remember that whatever information you input into ChatGPT or any other AI model becomes the property of that model. This means that if sensitive information is provided to the model, it essentially gains ownership of that data. Therefore, it’s essential to avoid inputting any sensitive or confidential information into these models until adequate regulations and safeguards are in place.

Moreover, although these models have proven to be quite competent, they are not infallible. Mistakes can happen. For instance, I once asked ChatGPT to summarize an article on a specific topic, only to receive a summary entirely unrelated to my request. Hence, it is of utmost importance to carefully review and verify the output produced by these models before relying on it for any critical task.

Close window