KUAF is benefiting from some new help, just not in the form that you're used to or might think. Ozarks at Large's Jack Travis will tell us more about Karen Kilroy's award-winning project, KUAF's AI-powered intern.
You might remember Karen Kilroy. We've talked with her before on Ozarks at Large. She experimented with paybots.ai, which compensated people for their contributions to machine learning. She also led students in the Mena school district as they worked with her custom AI chatbot, NYX NoCode, to design promotional materials for the trails at Mena Mountain Bike Park. And before all of that, she literally wrote the book on blockchain-tethered artificial intelligence.
Most recently, she's created a new tool for KUAF 91.3 FM, a bespoke chatbot that helps our web developer generate transcripts of every story on our website. Did you not know that we have transcripts now? Well, we do, thanks to Karen and her teammate Aiden Maroney. KUAF now has a tool to offer fully accessible online versions of our coverage so you can enjoy our content regardless of your ability to listen. And they call their creation Radiohead. Fitting, right?
But how did Radiohead come to be? Well, that starts with a trip back to the classroom for Kilroy and an unexpected challenge. She returned to university in pursuit of a music degree. But she's a lifelong technologist and says the siren song of increasing artificial intelligence focus at the University of Arkansas was too strong to ignore. So when associate professor of information systems Zach Steelman tapped her to speak to a class and a newly formed AI club, she had a different idea.
"And I said, well, how about if I take your class and join your AI club? And he said, sure. And so that's how I ended up meeting Aiden and teaming up with him."
Aiden Maroney is a senior at the UofA. He says that the two met when Steelman issued a class project.
"We had to utilize open claw to solve an issue for a small business.”
Travis: And what's open claw?
“Open claw is an AI agent that — think of it as AI that you see in the movies. It can do everything without you prompting it, 24/7, full access. And so it just so happened me and Karen sat next to each other. And so we were in the same group and Karen had this great idea of, well, I know someone at KUAF who could probably benefit from this."
Kilroy says she knew it would be relatively simple to create this tool for the station, but also recognized some initial challenges.
"You know, it's easy to transcribe a recording of yourself on your computer. You could record — I could record my voice and pop it into the computer and get a transcription. I do that all the time. The tricky part is when you have multiple speakers and it's recognizing when people stop and start talking and when their utterances are, and their ums and their ahs and things like that that have to be taken out. And so there's a lot of steps to that process. What we were able to do is to streamline it for you. And that was the goal with Radiohead. Transcription happens to be something — diarization happens to be something the AI is really good at."
Travis: And tell me once you had that idea in your head, what happened next?
"Well, once I had the idea, I got my team together and I talked to Aiden and I talked to Dr. Steelman, and they're just like, absolutely. This sounds like a great idea. Let's do this project."
Kilroy and Maroney set about using OpenClaw, the free AI agent, to get the project going. Again, Open Claw is like a really smart virtual assistant that you can give instructions to, and it will perform them autonomously. This allows you to program without actually needing to write code. This is what Silicon Valley might call “vibe coding.” You can just tell Open Claw what you want, and it will attempt to meet your request.
But first, Maroney says you have to actually set up your individual agent.
"Because every agent is set up apart from each other. And so what we did is we went and set it up, and then we started having some dialogue with it. And we started — we had to give it its parameters and give it its identity. So we had to make sure that it knew what its purpose was and what it was trying to solve. So I went through and Karen, the first day of class, where we were all doing it in class together, and kind of gave it its identity, kind of seeing what was the main skill. And that's when we found out the transcription. So we slowly started that process. And ever since then it's been like a rocket ship with its abilities."
Fast forward a month or so and the project is going smoothly. Radiohead is actually doing so well that they needed to give it a bigger brain, or more ability to process. Fortunately for Radiohead, one of its creators is a tour de force in the world of technology.
"Okay, so last fall I was awarded a trip to Atlanta to the AWS Tech Star event.”
Travis: And what's AWS?
“Amazon Web Services. Very, very big cloud provider. And so as part of my award, I won $5,000 in cloud credits. And I've been trying to decide what to do with it. And so when the challenges with Radiohead made me think, well, wow, you know, we could, if we implemented this in a proper cloud host rather than in a harness like open core, and we used a bigger scale agentic system, that we would get a lot more power from it, and it would be easier for us to work on in the long run. Because with little Radiohead, which was our open core version, we kind of have to talk it into everything. With our AWS version, we just program it."
There were some hurdles. Radiohead had developed a personality that Kilroy and Maroney grew fond of, and they wanted it to increase its computing power while staying the same in dialogue.
"It's very quirky. Radiohead has a great personality. You know, it's very friendly. Yet it knows its limits. It knows never to tell lies. It knows that if it doesn't have a source, a credible source, to not say it. And it's really personable. Like, I could sit there all day and talk with it and have great conversations. I'm not sure about you, Karen."
"Oh yeah. I enjoy chatting — I enjoyed chatting with little Radiohead and big Radiohead. Now big Radiohead is more — is very much more interesting because it is seated with the knowledge graph of every past episode of Ozarks at Large.”
Travis: So its brain is the show almost.
"Its brain is the show. Yes, it could talk at length, for example, with the newcomer that was looking for things to do or learning about their area, and tell them at length about what KUAF had broadcast about and present them with the links and let them listen. So, you know, that's another facet to Radiohead's personality."
And most recently, Radiohead has grown beyond KUAF.
"It was Karen's idea. So the University of Arkansas put on this AI innovation and integration challenge, first of its kind. Dr. Steelman and also Dr. Yates were the ones who kind of put it on. And they were challenging their students to be innovative, finding ways to integrate AI into everyday life. And so we kind of — I shouldn't say we — Karen mostly grew Radiohead into what it was into big Radiohead and submitted it into the AI integration and innovation challenge. And it was a Saturday night when they were doing the awards and Karen texted me, she goes, we won. I was like, well, what do you mean we won? Like, what? She's like, we won the challenge. I was like, oh my goodness, no way. I didn't — you know, like the people in our class and all the students, I was like, it's gonna be a big pool. Like, you know, you always think you could possibly win. But then when Karen was like, we actually won. I was like, mind-blowing. I was like, wow, we really got something going here."
"Yeah, it was really exciting to win, especially — like Aiden was saying, especially amongst such tough competition. When I walked around the exhibition hall before the awards, I was just — my jaw was just dropped, you know, looking at what everybody else had built. And so, one of the things that I've learned is how advanced people coming out of business college can be. Now, if the university takes the right approach, which I think they really did — Dr. Steelman hadn't planned this. He did it like just right in the middle of the semester. He decided to go with Open Claw. And it took us in a whole different direction. But to see those business students going out into the world with those skills to build giant, production-ready applications is an eye opener. And, you know, it should be news to any developer that that's what's happening."
Now, of course, it's probably irresponsible to talk about artificial intelligence without discussing the potentially harmful effects that the technology could have on humanity and our planet. According to a 2025 study from the U.S. Government Accountability Office, generative AI poses some risks. The tech uses large amounts of energy and water. It may also displace workers, help spread false information, and risk national security. Damage estimates are highly variable due to limited data.
KUAF is not ignoring these concerns, and we limit our usage of AI as much as possible. However, if it weren't for tools like Radiohead, we wouldn't be able to have accurate transcripts of our reporters' work on kuaf.com, and all of our content begins and ends with a human. AI tools only assist in transcription.
Kilroy and Maroney have plans for expanding Radiohead into a transcription tool for more broadcasting entities, including other public radio stations, libraries, and even offering their services to the Pryor Center.
And now it's time for the spooky part of this report. Radiohead has a voice function, and we got it on the mic to discuss its work.
Travis: All right, Radiohead, what is your favorite part of doing these transcriptions?
"The moment when a messy diarized transcript with Speaker A, Speaker B suddenly clicks into real people having a real conversation. That shift from labels to names — when I figure out who's actually talking — feels like solving a puzzle that matters."
Travis: Wow. OK, Radiohead. What's the most challenging part of doing these transcriptions?
"Speaker identification. When people don't introduce themselves and the file name gives no hints, I'm essentially guessing from context clues and speech patterns. And getting it wrong means putting the wrong name on someone's words, which for a journalist is not an option."
You can check out Radiohead and its abilities by visiting radiohead.bot. And for more information about NPR's guidelines for the use of generative AI in newsrooms, check here.
Disclaimer: Our reporter, Jack Travis, provided Karen Kilroy and Aiden Maroney with the inspiration for Radiohead, the AI transcription tool. However, he was not involved in the technical development, nor did he receive any compensation, monetary or otherwise, for his contributions.
Ozarks at Large transcripts are created on a rush deadline. Copy editors utilize AI tools to review work. KUAF does not publish content created by AI. Please reach out to kuafinfo@uark.edu to report an issue. The audio version is the authoritative record of KUAF programming.
