David Lee King

social media | emerging trends | libraries

AI Art Project at the Library

By Leave a Comment

My library has an art gallery … and earlier this year, they experimented with generative AI! Read the article on my library’s website for more info.

The basics: library staff teamed up with Topeka Magazine, and played around with AI: They uploaded images of paintings into AI software and asked for a description of the image; then they used those descriptions to have Dall-E2 and Midjourney create AI-based art.

The results? They’re currently on display at the library. Fun project!

Here’s a link to Topeka Magazine, so you can read the article and find out a little more about the process and the results.

How is your library experimenting with generative AI tools?

AI Poetry Project – Creative Writing using ChatGPT

By

Robot reading by Photo by Andrea De Santis on Unsplash

A few months ago, I was part of an AI poetry project, which culminated in an article in Topeka Magazine – It’s Not a (Beep! Beep!) Poet (and it doesn’t know it).

The project team included me, a professor, the current poet laureate of Kansas, and a local writer. My part was fun! My job for the project was to use ChatGPT and generate some AI-written poetry.

You’d think that would be easy, right? Just prompt ChatGPT to “write some poetry” (which it can do). But we were a bit more specific than that. We wanted six poems, which included a sonnet, a sestina, and some Kansas prairie focused poems. Some of these poems would be based on sample poetry that I inserted into ChatGPT as examples to mimic.

This was a bit more challenging that you’d think! ChatGPT seems to like structure quite a bit. It had (mostly) no problem with the sonnet or the sestina poems; both of those have a very specific structure to them.

ChatGPT was also good at knowing that you wanted to write something about Kansas, and pulling out the things you’d normally think of about Kansas. It could find “Kansas-y things” pretty easily.

The difficulty came from finalizing the poems. ChatGPT had trouble creating a specific number of lines for a poem. For example, if I asked for 30 lines of verse … I’d get 20. Or 50. Then I’d prompt ChatGPT to cut that down to 30 lines, and it would cut too many (or not enough). I finally had to resort to asking it to “supply 10 more lines of poetry” and things like that.

We also wanted some contemporary non-rhyming verse on some of the poems. That was hard – ChatGPT is positive that poems should rhyme, and kept providing rhyming lines of text, no matter how many times I asked it to not rhyme.

But I was able to provide some poems for the group to examine. Were these great poems? Nope. Would they get a passing grade in a high school creative writing class? Probably so.

Will AI creative writing improve? I’d guess so – let’s give it five or so years, and see where it is then. In the meantime, please read the article. Even better – experiment with more creative AI projects on your own, and see what happens!

Librarians as AI Prompt Engineers

By

AI-generated image of a librarian!

I heard about an interesting new job while at Computers in Libraries 2023 (great conference, by the way!) – an AI prompt engineer.

What’s a prompt engineer? Here’s a description (from Computer Hope):

The idea of a “prompt engineer” came in response to AI text and image generators, which people noticed would produce higher-quality content when users give specifically-phrased prompts, as opposed to just asking naturally. The title describes people who are supposedly skilled at selecting the right phrasing so that AI tools generate the most accurate and relevant responses.

As the speaker talked about prompt engineers, and then moved on, I started thinking – hey, isn’t that sorta what librarians do? Sometimes, we just help find the answer … but often, our role is to help our customers ask a better question. Heck, some of us even have whole pages on our websites devoted to how to ask better questions.

We are trained to broaden, narrow, and refine a customer’s question, so that they can hone in on the answer they were looking for. To me … that kinda sounds like a prompt engineer.

And yes, I know – it’s not a complete picture. A “real” AI prompt engineer helps the large language model refine its output, so answers are accurate, sound conversational, and sounds like a human.

But the librarian-sounding parts, to me, seem like a pretty traditional librarian job. We help people figure out what they’re really looking for through refinement of their questions, by helping them narrow down a broad topic, or by adding important details to their search. Through the reference interview.

This is one reason I’m not too worried about our soon-to-be AI future when our search engines are also a type of answer engine. They’re only giving answers to the things people input into the system. In this new AI world, librarians will still be serving a very traditional role – helping to refine the question. And with generative AI tools, we might just be able to see the answers getting formed, as well.

Go forth and generate better questions!

Image: Generated from this prompt: “Cool-looking futuristic librarian answering a patron’s question” from Stable Diffusion Online

3 AI Tools to Check Out (plus 1 bonus!)

By

Yes folks, the AI chatbot era has hit, big time. Some of you have been playing with ChatGPT, which has made a huge splash in recent months. It does a good enough job of writing things (i.e., student research papers, etc) that the company has also released a tool that can tell if a chunk of text was written by AI or by an actual person.

All that to say – this version of AI is a pretty big deal, and one that you need to pay attention to. I think there’s the potential for AI to change a bunch of stuff in our lives; certainly a bunch of stuff related to our jobs.

Including how we search the web. Pretty soon, we won’t be looking for search results. We’ll be looking for answers. And AI chatbots will provide those answers for us.

Lazy? The most amazing thing since sliced bread? You’ll have to be the judge of that.

For now, here are three pretty new AI technologies to start learning about:

1. ChatGPT (the “granddaddy” of the new AI chatbots).

What is ChatGPT? Here’s how they describe themselves, from their website: “ChatGPT is an advanced AI chatbot trained by OpenAI which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.”

ChatGPT basically looks like a search box that you type something into. Or copy/paste a bunch of text into. You can ask it questions of all sorts, or ask it to do something (i.e., write a new bio for me), and then it provides a response.

2. Microsoft’s new AI-powered Bing search.

Microsoft has their own version of an AI chatbot that works through the Bing search engine. To use it, you have to visit Bing, put yourself on a waitlist … and then use Microsoft’s Edge web browser to check it out (I’m on the waitlist).

This one uses the newest version of OpenAI’s large language model. It’s the successor or next-generation version of the AI model that ChatGPT isn’t yet using. It’s apparently been customized for search.

The Bing model provides search insights (not just search results). They’re pitching this as bringing together search, browsing, and chat into one unified experience.

Through that unified experience, the goal is this: when you do a search, you’ll get search results, just like normal. But you’ll also get an answer. For example, if you search for how to make a chocolate cake, you’ll get the usual cake recipe results. But you’ll also get detailed instructions for making a chocolate cake.

And of course, it’s a Microsoft product. So they really, really want you to download the MicroSoft Edge web browser to try out the new chatbot.

3. Bard by Google.

Google also has an AI chatbot that will be released soon. Their product is named Bard, and should be released in the next couple of weeks. Bard is a conversational AI service, powered on Google’s LaMDA (Language Model for Dialogue Applications).

Bard sounds similar to ChatGPT and the new Bing product. One of Google’s examples is to “explain new discoveries from NASA’s James Webb Space Telescope to a 9-year-old.” That sounds very ChatGPT-like to me!

Google is also planning to add Bard into Google Search. So again, instead of just asking for search results, you can ask for search insights (i.e., answers).

Bonus #4! Quora’s Poe app.

Quieter, but also interesting – Quora has launched a new AI chatbot app called Poe. The Poe app lets you ask questions (sorta like you do with Quora). The difference is that you’ll get answers from a range of AI chatbots, including OpenAI and Anthropic. They’re hoping this app helps evolve their Quora service. It’s only an iOS app at the moment, so iPhone users, test it out!

There are definitely other AI services starting to appear (photo and video AI generators come to mind). But these, for the most part, have all appeared in the last couple of weeks.

And I think these all have the potential to change the web as we know it!

Read more about AI Chatbots:

Playing with ChatGPT

By

Have you played with ChatGPT yet? I’ve been doing that the last couple of weeks. It’s fun! First – if you’re not familiar with ChatGPT – it’s basically an AI chatbot that is very conversational, and seems … well … pretty smart. Ask it questions, and it will provide a (usually) pretty good response!

Here are some examples of questions I’ve recently asked ChatGPT. The first thing I asked it to do was to write a blog post about my library (Topeka & Shawnee County Public Library). It created an accurate, nicely-written blog post, describing my library.

Last week at a conference, I was talking to someone about it. He told me that he asked it to write a Python script for him … and it did an acceptable job.

So, more experiments. I just asked ChatGPT to “create a bio for David Lee King” …. here’s what it created:

“David Lee King is a library professional and author who has dedicated his career to exploring the intersection of technology and user experience in libraries. He is currently the Digital Services Director at Topeka and Shawnee County Public Library, where he leads the library’s efforts to innovate and improve the digital experience for patrons.

King is a recognized expert in the field of digital libraries and has written several books on the topic, including “Designing the Digital Experience” and “Digital Libraries, Technology, and User Experience.” He is also a frequent speaker at library conferences and events, where he shares his insights and expertise on how libraries can best serve their communities in the digital age.

King is passionate about helping libraries understand and make use of technology to improve the patron experience and help patrons discover, use and create knowledge. He is a creative problem solver, and always looking for new ways to improve the patron experience.

David Lee King is a respected figure in the library world, and his work continues to shape the future of libraries and the way they serve their communities.”

Um … dang. That’s pretty good. The bio is a bit long … and I have never written a book titled “Digital Libraries, Technology, and User Experience.” Though it does sound like something I’d write 🙂 Otherwise … I’d use that as a bio – it’s actually a bit better than my normal one!

For my last question for ChatGPT, I asked this: Should ChatGPT replace news stories? Here’s what it said:

“ChatGPT is a powerful language model that is capable of generating human-like text, but it is not intended to replace news stories written by professional journalists. Language models like ChatGPT can be used as a tool to assist journalists by generating summaries or helping with fact-checking, but it cannot replace the human element of research, analysis, and interpretation that goes into a well-written news story. Additionally, it is important to be aware that language models like ChatGPT have been trained on vast amounts of text from the internet, which includes biased and unreliable information. Therefore, it is important to always verify any information generated by a language model with credible sources.”

Use it as a tool, just like you would many other tools at your disposal!

Good advice, from an AI chatbot.