With the emergence of generative artificial intelligence (AI) comes a new tool for people to learn and leverage. It’s a valuable resource for creative and technical roles alike. Recently I worked with AI tools ChatGPT and AutoGPT to help me build a complicated taxonomy. Though the purpose for which I used it is very specific, the lessons learned are applicable to anyone interested in using AI to complete work more efficiently.
Ask AI Questions
It’s easy to fall prey to a little something called “blank page syndrome.” Which is when you have work to do, but you’re starting with a blank canvas and have no idea where to begin. That’s a common feeling when using AI for something new for the first time.
Something I’ve found helpful when unsure about where to start is to ask AI how to use it for the project or problem you’re working on. In the case of building a taxonomy, my prompt could be, “How could I use artificial intelligence to help me create a comprehensive taxonomy on a given subject matter?” It’s okay if the answer you get is slightly off the mark. You can clarify with a follow-up prompt. AI will remember what you have previously discussed, and it will refine its responses over the course of your conversation.
Additionally, depending on the type of project you’re using AI for, it’s not a bad idea to ask AI to confirm some of the basics. For example, I learned about creating taxonomies in grad school, but I still asked it for the various pieces of a taxonomy as a reminder and to confirm I wasn’t missing any pieces. In this case it was confirming that I needed parent terms, sub-parent terms, child terms, sub-child terms, relationships, and synonyms.
Test Small First
My specific use case for AI was to use ChatGPT and AutoGPT to attempt to automate the process of building out a complicated medical taxonomy. I had a theory on how I could use AI tools to automate the creation of a taxonomy. The medical taxonomy had a lot of complexities, so I decided to use a taxonomy on potatoes to experiment with the process.
My first request to AI resulted in the traditional organization of genus, species, phyla, etc. and the different types of potatoes. In my mind, I was looking for a breakdown of ways to prepare potatoes. Because I ultimately wanted to AI to help with building a complex taxonomy for patient care, this simple test gave me an understanding of how specific I needed to be.
Break Things Down
The potato taxonomy test taught me another very important lesson when working with AI. If you have a large project or task to complete with AI, break that project down. For the patient care taxonomy, I was able to use AI to help determine where to start for top-level categories, and then break down from there.
I did learn that AI worked best for the top-level terms and not so great for sub-terms. Where I needed hundreds of sub-terms or child terms, it was only producing 20. This is where breaking things down even further was valuable. I was able to start with those top-level terms, then ask for child terms, and from there, ask it to make the taxonomy three levels deep.
Expertise Will Always Matter
Many are concerned about AI replacing jobs. As I’ve come to use it, I’ve discovered that having experience and expertise in your field will matter more. There is a lot that AI can do, that doesn’t replace my role. It allows me to work more efficiently and provides a way to make budgets go further.
It’s important to remember that what you should view what you get out of a tool like ChatGPT or Google’s Bard as a jumping off point, not a final product. As our CTO, Mathieu Agee, mentions in his blog post about ChatGPT, AI is a great starting point, but it needs to be fact-checked because it sometimes pulls incorrect or outdated information.
AI is certainly a powerful tool that will provide people in all fields with new ways to become more efficient. The key is taking time to test it out and understand what its true capabilities are.