AI and Libraries: CONUL T&D Seminar blog post series, part 1
Welcome to the asynchronous component of the CONUL T&D seminar on AI and Libraries: What you need to know, being put on in partnership with University of Galway Library. The in-person seminar will be taking place here on the University of Galway campus 28 November and will welcome librarians from the CONUL member libraries.
This asynchronous series of blog posts are intended to get attendees of the seminar warmed up for the topic of AI and Libraries. Attending to these is entirely optional, but we hope you get something out of them!
The blog post today will be a general introduction to AI and will be the first in a series of five, with topics to follow including AI and library marketing, collection management, research support and student learning.
Each post will have a task or two to try out. If you'd like, you can write out a blog post or shared doc or similar of your own as a response to the activity prompt. You can submit a link to us via this form; this will count as an entry into our AI + libraries seminar contest. It will be looked at by our judges and if deemed the best entry, will win a 100 euro prize! Contest closes close of business 25th November, winner to be announced at the in-person seminar on 28 November. Entries to be judged on three simple criteria: originality, readability, and concision!
Please note that while anyone is welcome to follow our blog posts and try out our activities, only attendees to the seminar will be able to submit entries for the purposes of the contest.
On to our first blog post!
AI and Libraries blog post 1: AI, a general introduction
What is AI?
The term artificial intelligence has been around for a good while now, but the AI that is all over the news, our inboxes, and showing up embedded in the various apps, websites, and browser windows of our everyday lives is what’s referred to as generative AI or GenAI that is driven by machine learning. These terms together refer to a process where a machine is capable of learning, by processes like repetition, of creating things on its own, whether it be text, image, video, all in increasing degrees of complexity.
Libraries and library work have been profoundly transformed in the last thirty years by changes in digital technology. As a newly emergent category of digital technology, AI has generated a good deal of excitement and anxiety about what the implications will be for libraries among the library community. [illustrate this with data presented at CONUL T&L 2024]
This series of blog posts leading up to an in-person seminar on AI will service as an entry point into the space of AI and libraries, covering some of the key areas of activities for libraries, including information literacy, cataloguing and metadata, marketing and outreach, and (heritage) collections.
This first day will set aside the ‘and Libraries’ part of the equation, introducing a key skill, that of ‘prompt engineering’ as this can be applied to a few available AI tools.
Prompt engineering refers to the way that we tell certain prominent categories of artificial intelligence tools to do what we want them to do. The paradigmatic case is the tool known as Chat GPT. You can tell Chat GPT to do something like ‘write a paper illustrating the concept of hegemony through the use of ethnographic evidence’. You can improve the performance of the tool by, for example, being more specific about what you tell it to do – you can tell it to make the text a certain number of paragraphs for example, or use a certain number of examples.
In other words, good prompt engineering can involve an iterative process of improvements to the prompt and related improvements to the end product that the AI tool you are working with creates. A good approach to telling AI what you want is to test and refine your prompts. In doing so, you might wind up with a sequence like the following:
“Compose a compelling product description for a professional-grade camera that can capture stunning visuals in challenging light conditions.”
“Craft a persuasive product description for a high-end camera that can deliver stunning results across a range of genres, from wildlife and sports photography to studio portraits and landscapes.”
“Write an engaging product description for a premium camera that is designed for photographers who demand the ultimate in speed, precision, and creative control.”
Or, if you were using a ‘text-to-image' type AI tool, you could test and refine your prompts and wind up with the following sequence of prompts:
Generate an image of a serene beach at sunset.
Generate an image of a serene beach at sunset. Pay attention to the warm colors of the setting sun, the gentle waves, and the relaxed atmosphere. Ensure there are no visible human-made structures or objects in the scene.
Generate an image of a serene beach at sunset, where the sun is setting in a breathtaking display of warm, vibrant hues. Capture the tranquil ambiance with gentle waves gently lapping against the shore. Ensure the scene exudes a sense of calmness and tranquility, with no visible human-made structures or objects to distract from the natural beauty.
Exercise 1a
There are many AI tools out there, some very well known (like Chat GPT). To get a feel for text-based tools and using prompt engineering, cut and paste the first sequence of prompts, above, into Chat GPT. Note that you will need to create an account to use Chat GPT.
Exercise 1b
Now try a second tool, like the open source tool HuggingChat, or well-known tools Claude or Google Gemini. How does this second tool seem to be different to Chat GPT? Or does it seem mostly the same?
Exercise 1c
Now try the text-to-image sequence above in two text-to-image tools, for example Adobe Firely, Microsoft Image creator, or Craiyon
Task
Test and refine a series of prompts using any of the tools we have looked at today. When thinking about a prompt, you can cast a wide net. You can think about something that is related to the work that you do; or, you could think about something to a hobby or some other aspect of your every day life. How might you leverage AI to enhance how you do things in your day-to-day life?
Comments