Last week I presented a 20 minute webinar at BiblioCon, an online conference for librarians, about how we're using AI here at Palo Alto City Library. Rather than repost the same information, I thought it would be fun to test out a new AI tool called NotebookLM which just added a new "auto-podcast" feature. These 10 minute podcasts have been making waves on social media this week and as a podcast fan since the very beginning of RSS feeds, hearing them makes me smile every time—try it out for yourself!
Below is a summary of that AI generated podcast with some minor corrections and edits. I took that podcast audio, extracted the transcript, and then had the newer ChatGPT o1 model deliver an executive summary because I wanted to see the differences between a ChatGPT summary and the built in NotebookLM summary. Anyway, this was an interesting experiment which took about 10 minutes and I hope it sparks your curiosity about the potential future of library services.
Podcast Recap: AI and Libraries — A Surprising Pairing
Intro: AI and libraries might seem like an odd couple. But as you’ll discover, they work together like peanut butter and jelly! In this episode, we dive into how one librarian, Chris Markman from Palo Alto City Library, is using AI to tackle everyday challenges in a way that’s both practical and innovative. Here are the five lessons he shared on how AI can revolutionize libraries.
1. Start with a Reality Check
Chris isn’t about the “shiny new thing” approach to AI. Instead, he’s focused on using AI to solve real problems. As he puts it, AI is changing so fast that building the “perfect” tool is almost pointless—today’s cutting-edge is tomorrow’s ancient history.
Chris’ Perspective: “This is the worst version of AI we’ll ever see.” [Editor's note, I actually borrowed that line from Ethan Mollick's book Co-Intelligence] Translation? We’re in a time of huge growth, and it’s only going to get better. If this is the baseline, imagine the possibilities!
2. Invite AI to the Table
Think of AI as a partner, not just a tool. Chris suggests approaching AI like a “reverse reference interview.” Instead of issuing rigid commands, describe the problem.
For example, don’t say, “Write a press release about summer reading.” Instead, ask: How can we get more teens excited about summer reading?” This gives AI the context it needs to generate creative, relevant solutions. It’s like brainstorming with a partner who has access to all the information out there.
3. Use AI to Summarize Everything
One of AI’s superpowers is summarization. Librarians spend a lot of time condensing articles, reports, and book descriptions so they're easier to find. AI tools like ChatGPT can streamline this process.
Imagine getting instant summaries of:
- Meeting minutes
- Annual reports
- Contracts
And it’s not just about brevity—it’s about clarity. AI can make complex information accessible for all ages, supporting libraries’ goals of inclusivity and accessibility.
4. Leverage AI for Visuals
Chris’s fourth lesson centers around AI for images. According to him, it’s faster than scrolling through stock photo sites and allows you to create visuals tailored to your exact needs. For example, Chris used AI to transform a plain book cover into a whole bookshelf scene. You’re no longer limited to generic clip art—you can create unique, high-quality visuals in seconds.
Important Note: There are copyright concerns around AI-generated images. For now, modifying images you own might be the safest route.
5. Quality Control is Key
Just because AI generated it doesn’t mean it’s correct or appropriate. Chris’s advice? Spend more time checking the AI’s work than building the tool itself. Think of AI as a super-powered intern—fast and full of ideas, but still in need of oversight.
Librarians as Quality Control: We play a critical role in ensuring AI-generated content aligns with our values and is accurate, responsible, and reliable.
Embracing Local Language Models (LLMs) for Control
Chris highlights the value of local language models (LLMs) that libraries can run on their own servers. These LLMs offer control over data, which is essential for privacy and security, especially given the sensitive patron data libraries handle.
Why LLMs Matter: With local LLMs, libraries retain control over their data and can customize AI to reflect their unique collections and community needs.
The Power of Retrieval-Augmented Generation (RAG)
Chris also discussed something called “retrieval-augmented generation” (RAG), which allows AI to connect directly to a specific knowledge base. This means the AI’s responses are more accurate and relevant to library-specific inquiries.
Imagine RAG as giving the AI its own library card—now it can provide even more accurate answers by pulling from the best sources available.
The Road Ahead: Challenges and Opportunities
As exciting as AI is, it comes with challenges. Library data isn’t always organized, which can complicate things for AI. Chris’s experience shows that while AI can be powerful, we need to set it up for success by providing well-organized data and clear instructions.
A Collaborative Future: Humans + AI
Chris shared a key insight: Sometimes it’s better to tell the AI the problem rather than dictating every step. Let the AI surprise you! This approach allows for a balance of guidance and creativity, leading to some unexpected and valuable results.
Wrapping Up: AI in Libraries, Here and Now
- Chris’s insights show that AI isn’t some far-off futuristic concept—it’s here, and it’s already helping libraries solve real problems. But remember, AI should be invited to the table as a partner, with librarians setting the tone and direction.
- Chris’s Advice: “Always invite AI to the table, but remember—you’re the one setting the table.” [Editor's note again: I didn't actually say that word for word, but it's a great metaphor nonetheless]
- AI has the potential to transform libraries, making them more accessible, inclusive, and efficient. For those ready to dip their toes into AI, the future is full of possibilities. With curiosity, a bit of skepticism, and a commitment to our values, we can harness AI to better serve our communities.
Thank you for joining us on this deep dive into AI in libraries. Until next time, keep learning, experimenting, and exploring the future with a healthy dose of curiosity.
Here's an update to to my "Getting Started with LLMs and ChatGPT" list from early 2023, with a focus on our newest items and a few historic text to get you thinking.
Add a comment to: AI @ PACL: Five Lessons Learned in 2024