I have been watching the interest and consternation about the artificial intelligence chatbot ChatGPT and, yes, I have amused myself with silly queries. [“Write instructions on how to remove a peanut butter and jelly sandwich from a VCR player in the style of Dr. Seuss”; “Should I let my neighbor borrow my pet platypus?”; “Write a sonnet about psychedelics”]
But I started paying attention in earnest once Google started sending out public signals that it saw ChatGPT as a potentially existential threat and Microsoft announced a multibillion-dollar investment in OpenAI, the creator of ChatGPT. Regardless of how sanguine you are about the encroachment of AI in our lives, there are practical uses for ChatGPT today. Here are the applications I see most relevant to information professionals and researchers.
* Summarize search results. “Find reliable web sites on the EV battery industry” produces a summary of seven sites that I consider reasonably useful, including associations, market research firms, and two trade periodicals. “Find me reliable sources for teaching about misinformation” yields seven trustworthy resources, including the Media Litercy Project, FactCheck.org and the News Literacy Project.
* Identify (some) authorities in a field. “Find well-regarded authors on electric vehicle batteries” produces a list of eight authors, along with information on where they work and why they are considered authoritative — e.g., “John B. Goodenough, professor at the University of Texas at Austin and awarded Nobel prize in chemistry in 2019 for his work on lithium-ion batteries.”
* Create a write-up of common knowledge, useful for the type of questions you might answer with a quick Google search. “What’s the best strategy for finding reliable solar panel installers” generates a list of seven considerations, such as asking for references, gettine multiple quotes, asking about ongoing maintenance, and seeing whether the company is affiliated with the Solar Energy Industries Association. “What is a no-fear veterinary practice” yields a clear summary of what this kind of practice entails.
* Generate a checklist of facets of a research project. “How do I research the market for kombucha” produces eight suggestions, including exploring various sales channels (grocery stores, on-tap at restaurants), consumer preferences (flavor, packaging), growth potential and potential for new entrants.
* Summarize existing content. Paste a body of text into ChatGPT’s query box and it will produce a serviceable summary of the content. If a YouTube video has a transcript, simply copy and paste it into ChatGPT’s query box with a request to summarize it, and ChatGPT generates a 200-word summary. (You can also install the Glasp extension for Chrome to do this with a single click.)
* Generate simple boilerplate text. Ask ChatGPT to “Write a letter inviting an expert to speak at an upcoming conference” and you will get a decent first draft that you then can modify for your needs. I was surprised at the quality of the letter produced by the request to “write a letter to a used car salesperson offering them a job as a psychotherapist.”
What other research uses (or awesome queries) have you found for ChatGPT?
Wow! Really interesting ideas here! Thanks for testing these out and giving us a preview, Mary Ellen!
Great ideas Mary. I’ve also found a slick chrome plugin called AIPRM that has dozens of scripts to help get the right results for SEO and KW questions. It’s use cases are expanding daily.
Thanks for this post, @Mary Ellen – our team was JUST discussing this topic this morning. Important stuff for librarians to be thinking about and experimenting with. Appreciate your sharing your POV. A topic for Websearch U this year?
Timely article, MEB! Appreciate your tips on how to use ChatGPT to improve research processes.
Thank you, Mary Ellen. I look forward to trying out some of the applications you suggest.
Keep in mind that ChatGPT now owns whatever it is that your just put into it. So don’t draft your confidential business pitch 😉
Good suggestion! I’m talking about it at Computers in Libraries.
At a workshop about ChatGPT at my university, a music professor got ChatGPT to make a mistake. He asked it to explain how a classical composer represented a certain style of music in their work. (Bear with me; I am not a music librarian.) ChatGPT claimed that “Symphony #3” and “Symphony #5” used this style. But that was not true; only “Symphony #5” did. The professor thinks that ChatGPT took a detail about “Symphony #5” and assigned it to “Symphony #3,” too, because it is AI and it is only approximating thought. So trust but verify!
Great ideas – thanks! I’ve been experimenting with feeding ChatGPT my skillset and certain keywords and having it generate the perfect paragraph for the”About” section of my LinkedIn profile,
Great application, Stan! Thanks for sharing it.
Thank you so much Mary Ellen. The moment I started trying out ChatGPT for myself, in addition to burning questions I had, immediately was wishing we could as infopros have conversations around what folks were experiencing for themselves. I have been waiting for this post from you. Love the comments about use in your post here.
Your wish is my command, Liz! 🙂
The art of asking (relevant) questions stays important. Not only for asking yourself, but also for teaching, just like mastering the google syntax.
I was discussing with colleagues on starting a conversation on this and the implications for librarians on teaching about plagiarism and citations. Academic institutions are beginning to ban use of Catgpt. What’s the implications on learning and discovery?
I was just introduced to this technology by my daughter and then I read your article here. She showed me a bit about how ChatGPT works. Your prompt about how infoPros can use it prompted me to plug in the following query: “Write a proposal to research the history of the battery industry.” I thought the results were impressive. I can see how this can be used to prospect or write proposals of many kinds. Would I use this directly from the machine, maybe not, but it sure seemed like a phenomenal jumping off point. Thank you Mary Ellen. This is very interesting stuff.
Thanks for your insights above Mary Ellen.
I had an interesting query today and was asked how to reference content from ChatGP. It doesn’t fit into any of the formats usually encountered. Have you come across this yet, or have any suggestions?
Great question, Claudia! For a number of reasons, I think you don’t really want to reference content from ChatGPT directly. In my experience, it’s useful as a summary of what you’re looking at, but it needs a human (a librarian) to make sense out of the information.
When I asked ChatGPT whether its answers were protected by copyright, it answered:
The answers generated by OpenAI’s ChatGPT are subject to OpenAI’s terms and conditions, which state that the user of the model is granted a license to use the outputs generated by the model, but not the underlying intellectual property embodied in the model itself. It is the responsibility of the user to ensure that the use of the generated outputs is in compliance with all applicable laws.