Artificial Intelligence in the Library [podcast]

Gerry Bayne: Artificial intelligence is being used in and developed for the library, museum, and archives fields. As the technology becomes more advanced and widespread, it's important to see the uses AI can bring to the library. I spoke with two librarians about this trend: Karim Boughida, who's the Dean of University Libraries at the University of Rhode Island, and Chris Erdmann, Chief Strategist for Research Collaboration at North Carolina State University. I started by asking them in what ways are they seeing libraries engage with AI, and what they think are the most promising areas for library focus.
Chris Erdmann: I think one of the areas that you see libraries already starting to engage with AI is around just using some of the machine learning tools that are out there. An example was recently an institution was working with triple IF images, and trying to classify them using Microsoft or Google's computer vision or clarify, and just experimenting with what comes out and classifying their various images from special collections, and just realizing that the underlying data that they've been trained on wasn't necessarily capturing correct classification items. I think you're seeing a lot of those things also with chat, looking at our chat logs and trying to create chat bots potentially to interact with our students, with our community. That's another way that libraries are experimenting. We're very good at the programming side of hooking into these tools. I think the other area that we can start engaging with AI is with our graduate students post docs, working with them collaboratively on developing material together, so for workshops or events. The other area I think that is definitely an area where we could focus on and for what we hear from computer scientists and other people is that there's really a need for data ethics, so conversations around that. Again, people are focused on the technology and not exactly the social aspects, the people element of AI. Having those conversations, and again, the library facilitating those, I think that could be a really great area for us to work in.
Gerry Bayne: Sure. Did you want to add anything, Karim?
Karim Boughida: Yes. I agree with Chris. I would say that the library now is more morphing into an interdisciplinary hub within the university or even beyond the university of terms of public libraries. It's where things are happening. The reason we're pushing for AI is that it's going to have a big impact on society. People need to talk about it. People need to understand what is it, what impact, and how they can influence direction, and have been embedded, what Chris just said, like ethics or what we may call a social AI. In terms of technologies, we mentioned yesterday that there's what we call AI effect in that sense that when something became mainstream, it's considered no longer AI, like in our field, like information retrieval, OCR, optical character recognition, spatial recognition. We don't call it AI anymore. AI is an umbrella. You defined it as you wish, but it also has a strong component in terms of computer science. The University of Rhode Island, the main goal is very educational. It's not research oriented. We got the grant because we wanted prepared students and learners to learn about AI because actually the jobs are there. It's clearly high, high demand, and also, prepare them in terms of liberal arts how to think critically about AI. Also, we shared that when it's purely technical, it can be damaged if you don't have those social values. It's important, like in our field, that social values of the library, like social justice, intellectual freedom, open access is embedded in AI.
Gerry Bayne: What academic areas or disciplines in your own institution are the most active in AI research or in incorporating AI in their coursework?
Chris Erdmann: Definitely computer science.
Gerry Bayne: Yeah.
Chris Erdmann: To start.
Gerry Bayne: That seems like an obvious question.
Chris Erdmann: Yeah. Just recently we had an announcement at the beginning of December about some research, people at NC State looking at how they can help AI programs with just in general gaging the intent, people's intent. In various cases, like in chat programs and others, trying to understand what people are really getting at, and so speeding up the process of the conversation. That was just recent. Definitely our computer science group is really looking into this. I attended the Department of Computer Science's anniversary recently, and they invited experts from far away and then also just business leaders. The thing that came up often time and time again is that from the businesses we're looking for AI expertise and then from the faculty talking about how they wanted to integrate it more into the curriculum. That's really great for CS. There's a big demand for this also in the other disciplines. It's very similar to what you saw with the rise of data science. There was this need for increasingly computation skills in many disciplines, almost all of them. That expertise doesn't sit necessarily within the discipline. Again, just recently, I talked to a faculty member about one of his graduate students that was taking courses and getting a second degree in CS, particularly focusing on AI because he felt that that was an area where he could be a great use, a bridge.
Gerry Bayne: Sure.
Chris Erdmann: Yeah. I think, especially that cross disciplinary side, is an area where, I think again, the CS group, the computer science group, could really help the other disciplines and be a resource.
Karim Boughida: Along the same lines, yes. At the University of Rhode Island, there is a push for interdisciplinary fields. We know that to solve big problems you have to have this approach that you involve many disciplines, but you still have to have a strong root in a specific discipline. In our area, of course, we have computer science, evidently. We just had 50 years of computer science at the University of Rhode Island. It was very timely. The second is computer engineering and biomedical engineering. One of the teams in our AI lab is from computer and biomedical engineering. They have also a strong link to health sciences. Also, we have philosophy. Also we have a course. It's already embedded. In the course is Sherry Foster, professor in our university. She's working on already embedding it into mind consciousness.
Gerry Bayne: That's interesting stuff.
Karim Boughida: It's very, very, very interesting. We have, also, the businesses, so we have the College of Business where Professor Gordon, he's already now in India teaching artificial intelligence. He already knows how to embed neural networks in his field. That's also promising. Also, we have the Graduate School of Oceanography. They have huge data, like climate change data. It's a logical sense from big data to AI.
Gerry Bayne: Yeah. This might be redundant to what we just talked about. What skillsets are needed to work in this area?
Karim Boughida: If you have strong (skills), in terms of the program inside, like Python or Jangle or Dancerflow, those help you transition to AI or in our field, if you have the data science or research data management, if you have strong programming background, you can transition to AI. In terms of the ethics, same thing. If you have also the basics in terms of ethics, how you can apply that to technologies in AI, so it's combined. Machine learning, that's specific. Machine learning, deep learning, you need to have specific training in them. Down the road, my university or other universities, are building it in terms of all of them, there's a huge, huge demand in the market, especially now, machine learning is more prominent. Cliff Lynch in yesterday's opening keynote, he mentioned machine learning four times. It clearly was the strong lead in this field. Also, the big players like Facebook, Google, and Amazon are hiring these people.
Gerry Bayne: Yeah.
Karim Boughida: That's why if you're not on the table, then you cannot influence any direction.
Gerry Bayne: That makes sense.
Chris Erdmann: Yeah. There's great demand in this area and some of the salaries you see are upwards of $300,000. It's very competitive. Like Karim was saying, I think in libraries, there's that transition you can make in using these tools, but I think that deep statistical knowledge is something that we need more of that expertise embedded in the libraries. You have that expertise that's in high demand in industry, and how can we embed that? I think again, through those collaborations with graduate students or post docs, potentially. Libraries do this for artists and residents. We could consider having an AI in residence and, fortunately, there are people in the AI field that like to do these kind of projects, and work with communities, and help them with understanding what's happening. I think there's an opportunity there for us to embed, potentially, that expertise in that was as well.
Karim Boughida: It's also part of reskilling because, down the road, when you have a new technology, it will displace skills. You need to be prepared, whether it's inside a library or in other fields. We need to give some ammunition to students and the staff on how we are going to handle this. Yesterday I shared one slide about the reason I wanted to have this AI panel. When seeing the movie Time Machine in 2002 where we have literally a reference librarian, it's a holographic librarian who came up in New York public library where this librarian has everything, like read all the books and has all the answers. Is this going to replace librarians? Not at all, but it's complimentary in terms of how would you handle those things. It's going to happen at some point and it's not like the machine versus man. In terms of skills, there's some where you have extremely repetitive tasks. That's happening already. We're not going to debate it. It's happening already. How would you handle those if you don't retrain them?
Gerry Bayne: Yeah.
Karim Boughida: If we had this big societal problem, you may have thousands of people in the market, like society, so it's how would you have peace and have income is where some even big CEOs are pushing for some kind of universal income. This is why AI is really at the core of what's going to happen to our society in the future and also how you can have a democracy, how you can have your openness, your governance. It's going to be impacted. Also, we talked yesterday also about diversity and inclusion. AI will be the same as you have your biases. If your biases are in the machine, if you don't have the right control, the right governance in terms of how to handle those biases, we'll have problems.
Gerry Bayne: You have to be very careful. Yeah.
Karim Boughida: Absolutely, we have to be very careful, and not transition easily because we have already divided society, especially now within the administration. How would you transition now? You have to fix that. There are things that are totally wrong. KKK is wrong. Hate speech is wrong. Would you help that? Okay, now we're translating from man to machine. This is a change, maybe, the machine to help a little bit.
Gerry Bayne: Right.
Karim Boughida: Raise awareness and people to be aware of their bias. There will be biases, but here we're talking just reduce the biases. What's wrong is wrong.
Gerry Bayne: My final question would be what would your advice be to other institutions that want to implement AI into their library and coursework? If they're just getting started, what would your advice be to them?
Chris Erdmann: I can warm up Karim here, actually, because I think we're going to have a similar message. I definitely think that one of our greatest assets is that our spaces and building community, and I think that can be a resource for all of use. Not necessarily if you have the expertise within your library, but you can start by leveraging your spaces and bringing this expertise together to talk about the challenges. I guarantee you that's a need everywhere that people always need to come together and share their ideas, share their challenges. I think that's one of the things that I could recommend to libraries. The other thing, too, is just, again, emphasizing the fact that, especially potentially working with Data and Society that's in New York that's thinking about data ethics is creating a place where that can be discussed as well because that comes up as a need in a lot of communities. Then, at a basic level, because it is a jump to do this.
Gerry Bayne: Sure.
Chris Erdmann: One of the things that libraries can start doing is just tracking it at a basic level and signing up for some of these AI newsletters. For instance, Laura Noren runs a data science newsletter that's data science, but she often covers stories on AI. Following those newsletters and some of those resources like humane AI allows you to track this and stay current at least at a basic level because I understand we all are sometimes strapped for resources. Developing these ideas, these can be helpful resources.
Gerry Bayne: Sure.
Karim Boughida: I think libraries and information services, they should do AI. You remember at the beginning of internet when scarcity was there, so people will learn internet from going to the libraries. Library was the mediator between society and technologies behind the scene. Actually, when I start talking about AI, people were saying, "Oh, we're not MIT." You don't need to be MIT to do this. Everyone can do it, even the small institution, even public libraries without much resources because they're here for the community. They need to be helpful, have the conversation, and participate. If you're strong in humanities, or typical library art college, you can participate. If you're strong from the technology side, be it. It's not just the arena of computer science. The need is huge and even companies are hiring also social scientists to deal with ethics and also with the corporate responsibility. That is also a niche for these people to have a say in terms of how would you handle all this project. Also, because it's a big society problem, I shared yesterday to one of the scenes of the movie, The Public, it's going to be out next year by Emilio Estevez, the whole scene is in the library. In the trailer, the trailer was released last week where it says why we're doing this. There was a conflict between users mainly homeless and the library and it was cold outside. This huge interaction and the user took over. They were saying, "Why? Why in the library?" It says, "Because libraries are the last bastion of democracy." It is and will be, so you have to be there and you have to be their voice. Also, we say library is a neutral space. That's not true. We are not a neutral space because library has embedded values. Typically, you can check those values within the basic American Library Association values, intellectual freedom, diversity, openness, community based. Those are our values. They're not neutral, so if you believe in those values, then you need to invest in AI.