Exploring AI Implementation Challenges in Libraries: Insights from a Panel Discussion

hi my name is rosin um I am moderating today this panel came out of um some conversations that we were having

actually in the Fedora Community um about Ai and their impacts on the practicality of our systems and also

conversations I've had with my own team who is currently working on a project with um Amazon uh Greg Grider is here

today you can he'll tell you about it or I'll tell you about it um hopefully they'll present at the next cni

hopefully um but uh I'm going to start off with a question and each of the panelists are going to introduce

themselves as they answer the very first question it's a general question um so the first questions are what are the

biggest challenges you faced in implementing AI initiatives at your institution and how have factors like

funding governance politics and broader campus interests influenced or a interest in AI influenced your progress

so we're going to start with Annie and go down the line awesome hi everyone uh my name is

Annie Johnson and I'm the associate University librarian for publishing preservation research and digital access

at the University of Delaware Library museums and press and I would say that our situation at the univ univers of

Delaware is maybe a little bit different than um some of my colleagues on this panel and and maybe um other panels uh

across the conference and that our university has not invested a significant amount of money into

AI um and so as a result all of the AI efforts that have kind of taken place around campus have been you know pretty

decentralized different groups experimenting doing research in in different ways and that has certainly uh

been the case with the library and I wanted to kind of say all that because I think that um especially at conferences

like this we often see some really amazing projects and we think well that's great but I don't have the money

to do that I don't have the staff to do that um that's not possible for me um and I am here with an example from the

University of Delaware to say that that's not true um I really really believe that Innovation around AI can

come from everywhere it doesn't just have to be uh folks with you know millions of dollars and multi-d

developers to make something happen um and so I I hope that uh this can maybe Inspire some folks who haven't yet

jumped into experimenting with AI tools um because they think they can't so what are we doing at the University of

Delaware well our big project is that we have um launched an AI based chatbot we started work on the chatbot in summer of

2003 and it went live in January 2024 it's currently um on our website you can check it out right now uh the chatbot is

called UD stacks and um it has been a deliberately collaborative project uh between Library it and our reference

Librarians and it's involved staff from across the organization we've really tried to get as many people involved

looking at the chatbot examining answers helping with the prompt um as possible to kind of get people that Hands-On

training um with the tool and um I don't have uh a lot of time today to go into our whole training process but if you're

interested in learning more um my colleagues and I recently wrote an article in information technology and

libraries um in their recent AI issue and that talks all about kind of the the training process for us um we are

currently running on uh chat gbt 40 and while the intermediate tool that we use for the chatbot is not a fancy

Enterprise solution like um ivai or um element 451 that many other you know institutions of higher education are are

using um it has worked really well and since we went live it's only cost us $400 so um I would say that as a pilot

project considering the small investment it's been really really uh valuable because again it's helped us upskill our

staff and it's also helped us position ourselves as kind of a leader on campus when it comes to understanding and

implementing AI based chat Bots um and that kind of gets me to I guess where our biggest challenge has been and that

has really been um that we haven't had any kind of centralized as a university we haven't had any centralized mandates

around um the use of AI tools and so when we wanted to start piloting this chatbot um there wasn't a

university-wide solution for us and there wasn't even really any guidance for how we should kind of go about this

so we really did it all on our own we we did the research ourselves you know we we skilled up ourselves and um we tried

to be really open with our University it about what we were doing uh but I will say that not everyone within it was

super thrilled uh with the fact that we just kind of went ahead and and did this ourselves um we didn't really ask

permission in part because we wanted to move quickly and and kind of seize the moment and so I would say that

navigating the the the politics around um our ability to have a chat bot and other units and ability you know that

haven't been able to have a chat bot has been one thing um that has been kind of an interesting ongoing conversation

between us and University it at this point it does look like there is going to be um an Enterprise solution for

campus when it comes to to chat Bots because again every every unit wants wants their own um so I think one

question that we're trying to work through right now with udit is do we join them do we join up in this um

solution would it work for us would to not work for us um and that's ultimately going to depend on a lot of different

factors so we're kind of working through uh that right now but I would say those those kind of ongoing conversations with

campus around campus regulations have been kind of the most most challenging um part of this so far and I will turn

it over to you thank you uh good afternoon um uh my name is Todd gone I'm an associate University

librarian at UCL La um and uh our um conversation with AI is actually about a decade old we've been working with uh

some Communications faculty members for a very long time uh who have been um using um these sort of tools and

techniques to extract speech and text from large corpora of um Broadcast News media so they're really interested in

going through hundreds and hundreds and thousands um actually hundreds of thousands of hours of Broadcast News

looking for um speech patterns that match up with gestures though little bit you know people saying something and the

gesture that goes along with it and I just for years I thought oh that's isn't that really interesting these um tools

and techniques that they can use to go through a giant giant corpora uh to find these very specific um points um

and so we um I've been working with them for for quite a while um and I became a little more interested in some of their

techniques um one uh rainy morning in Los Angeles when I had to go to the basement of our library where we had um

a room about this size filled with gift books and no one um uh no one to really look at them and process them and they

were a large portion of them were in um uh Asian languages and I thought to myself why can't we just take a picture

of the cover and use uh some of these AI tools to extract something that looks like a title and something that looks

like an author um and maybe we could sort through uh these these very quickly um uh and we you know this technique um

we started some experimentation at UCLA on um an unprocessed uh collection that we had uh digitized uh some years ago of

um uh political Flyers uh collected in Los Angeles um and so these flyers uh we had digitized them and we hadn't

actually catalog them so um uh our team started looking at uh some of these tools and techniques uh to see if we

could extract um text uh from these images um and it worked okay you know I think in a lot of ways um uh like

University of Delaware I was really interested in trying to get a um a person in the loop to work with the

engineers to make the tools better um uh fast forward till um last year UCLA um signed a contract with open AI um to

um uh work with their tools and my colleague Josh who's right over there um uh and I um they put out a call to

campus um to use the tool and we put a proposal in um and we're able to get one of these coveted licenses I mean I don't

know why they're so hard to get but um uh we we um we got the license and we're going to start this spring on uh doing

something a little more in Earnest and we're very interested in using these techniques um not only on scanned images

um but also to um uh uh on multimedia content again uh you know we've got a lot of uh UCLA uh part of UCL library is

the film and television archive the film and television archive is a huge collection of um media and we're very

interested in using these tools to see if we can automate um for uh Ada purposes

um so yeah I think that's kind of what we're doing um thanks hi everyone uh I'm Aviva WIA I'm

the vice Provost for libraries at the University at Buffalo which is part of the State University of New York System

also known as Sunni um I'm going to talk about a couple of things so first of all um uh I will start off by saying that

the University of Buffalo has been involved in AI research for many many years

uh for those of you who are um aware of one of the very first large language model implementations um this is uh

connected to the US mail service um our VP for research is one of the primary architects of the ability for the US

mail service to be able to read the handwritten addresses and get your mail to your house so our our uh commitment

and activities within that AI space have been as I said very longterm um a the University of Buffalo is also home to

one of the NSF aai institutes so we have been really investing for quite some time in AI as part of our kind of larger

research Enterprise um so uh I say all of that to sort of give you an idea that you know

as an institution AI is part of the work that we've been doing but we're not necessarily kind of like what Annie was

saying we don't have a ton of money right we haven't had as a library a ton of infrastructure for Technology support

we don't necessarily have teams of people who are interested or excited not that's not a fair way of saying that um

who are not necessarily um imbued with the idea that they that they can try some things with AI So within the

library we really only have one project that's going on within our special collections which is around um basically

using AI tools and it's actually co-pilot um I think to to come up with program descriptions for a collection

that we have that's um with um uh wbfo which is um our NPR station that's uh Buffalo Niagara so Western New York and

and Southern Ontario um so that's kind of one project that's happening in the library but I

think more importantly and kind of what I really wanted to kind of talk to you a little bit about was um so New York

state in the spring of uh 2024 the governor in Buffalo came to Buffalo and made an announcement about Empire AI um

so Empire AI is a public university and private university partnership to create a research infrastructure an AI research

infrastructure on Prem at the University so this is meant for all of the major research universities within

New York state to be able to have access to This research cluster to be able to do whatever kinds of AI research they

want to do uh it actually went online last month um so it is officially up and running down on our South uh actually

our downtown campus which is where our academic research cluster is currently located um they are building out uh on

our North Campus um a larger data center to be able to handle um uh AI research we are expecting within about four

months four to six months that we'll be able to bring that online and it will be four times faster than the

infrastructure that we currently have on Prem which is still pretty darn fast um so uh I I'm just going to mention this

because Tim made sure that wanted me to make sure that I did mention this that so U also has a commitment to be carbon

neutral by 2030 we are on track to be carbon neutral by 2030 and that means everything including like carbon offsets

for my travel right so we are committed to it as an institution and the building of the AI infrastructure on our campus

will be buying will be playing into that larger goal to be carbon neutral so we are actually trying to build an on-prem

infrastructure that will be carbon neutral as best as we possibly can which is pretty cool um so uh uh with all of

that I also wanted to kind of talk a little bit about how we've also been thinking really at the system level so

across Sunni around research data and what we can be doing to make sure that we can build like we can have data that

we can then feed into Empire Ai and use it as training data so one of the things we've been talking about at the system

level is actually creating a systemwide Data curation Network into a centralized data repository and then we can use that

data to train our own AI um so uh I think I'll maybe stop there that's a bunch of stuff that we've been kind of

got going on and we can dig into it a little bit more um and like I said you know we don't have a ton of stuff going

on within the library oh sorry I did want to talk a little bit about um campus stuff so the libraries have been

heavily very heavily involved in fact we have been um on the sort of Campus leadership team around all of our work

around um uh teaching and learning so we were on the original the initial task force around this we're on the

implementation task force around it we're in leadership positions across it I'm on the steering committee um and

we're also heavily involved in the way that AI will be playing out in the research Enterprise um and so we've been

heavily involved in conversations with our vpr's office and again and this research data piece is kind of playing

into that larger conversation with the vpr's office I'll stop hey good afternoon everyone um my

name is Tim sheerer I'm at UNCC Chapel Hill with the university libraries um I've decided to call myself a small llm

so what I say today is the product of the training materials and I just wanted to mention some of my colleagues so uh

the vice Provost and University librarian Maria sterino my colleague Michelle col who's the a for health

Affairs Amanda Henley who is part of data services and a grad student Cole Hargrave so a lot of what I'm going to

say it's product of that that training I would say um uh I'm going to talk first about

campus and then about the libraries so I decided that I was observing on this campus that this is the first time in in

my experience that what I think of as a technology project with vendors and risk and all those things um had product

ownership uh from the Deans when we moved from for instance on Prem to the cloud for Office 365 nobody cared about

that except how annoying it was to them but now we have a lot of product owners with conflicting um ideas of what they

mean when they say Ai and um I found myself without being asked saying hey can we have Grace for each other as we

figure this out um so I wanted to to mention that um uh there are a bunch of committees

I'm sure y'all are experiencing that on your campuses we have a a new chancellor who was just installed but he was in

term in the beginning of this year and one of his signature uh programs was Ai and he had a Chancellor's committee

Maria esto our University librarian managed to get on that committee and I I I'll mention that um as she was hearing

the the work and the excitement around it a lot of it was around research um a lot of people who are already Miles

Ahead of other folks and have access to tooling and um experience um uh and then a lot around operational excellence what

can we do to improve business processes and data yeah but she wasn't hearing as much about students um and so she began

to think about students student support in the libraries r on campus um there's also a Deans I would call it interest

committee so there's a group of Deans that meet regularly to talk about this and they actually went up to Microsoft

uh Innovation Hub uh for a two-day uh retreat and then there's a provos committee and I would say that's really

where the rubbert is hitting the road in terms of actually doing AI on campus which has five sub committies but I'll

I'll reel them off in case they're of Interest so there's one around pedagogy how how can students use uh AI in the

classroom there's one around uh campus strategy um there's one around uh communication how are we going to

actually bring communication together on this campus um I'm going to blank on one and then the last one is the

acceleration program so at this at the time all this was coming to a head uh we were revisiting

our relationship with Microsoft around Azure and so we actually did a a commitment I think of around $10 million

over five years uh just for AI and this actually was new money which I think is a significant way to think about it and

we needed to figure out how to distribute that so there's a AI acceleration program there's a way to uh

propose things there's help in proposing things because a lot of people need to help craft that and then we make

decisions recommended the Provost and they get funded uh hopefully um uh two Library staff sit on those Provost

committees and and then on um the subcommittees the proos committee and the subcommittee um I should also say

that uh we they negotiated for the co-pilot uh GPT for all UNCC Affiliates and also did a pilot for the uh the

co-pilot I wish they would name things differently uh the the co-pilot within Microsoft 365 so thank Microsoft Excel

teams Outlook um which they're pushing hard to to recoup money and I'm not sure we're totally seeing uh it's it's nice

to summarize meetings um maybe how uh so so back to Maria and and kind of where our heads are um we're thinking of

ourselves as the big tent we're not supporting a program or a school we support everybody uh we already have

spaces and staff and expertise and so we're thinking of ourselves a potential home to to help people kind of get um

off and running um so then to the libraries in parallel to this Maria had um adopted this idea of Road mapping

which is supposed to be sort of a Nimble way to figure out what you're going to do in the near-term and Midterm Horizon

and we came up with an AI road map um and we have five planks in that so one is Staff Readiness and I can talk about

that later but it's the idea that within a year everyone on staff will have some exposure to at least Ai and folks many

folks will have Proficiency in it um moving our library Services uh and Collections and people to meet the AI

needs of Campus is the second plank the third is collections which has its own uh plank as it were because we're

thinking about the context of our unique materials in llms and what that means we're looking at our risk profile that

we used to have for the repositories where there was a certain tolerance for risk but it it seems to go up a lot when

you start to think about millions of eyes looking in in in a minute at at your stuff um and then also the

licensing stuff that you've probably heard some about today um the fourth is Library work so how are we going to use

AI for our own work and we've been thinking about uh which you've heard about today if you went to yal but also

is thinking about um John dun and Indiana's work around basically scaffolding access to to special

collections but also um digital accessibility which in a couple years for public institutions is a big problem

to solve and AI seems to have promise for that also systematic reviews and the health

Affair side of campus and we're also uh doing a fair amount of work around research impact so there's some

possibilities there and the last blank is engagement which is um simply we're we're going to be wherever it is

including cni panels so well um thank you everybody and I also want to call out um Deborah hanken

CTS was supposed to be on the panel but she wasn't able to attend um today so um we have an empty seat for her um so our

next question Annie I was wondering if you could talk about it since you were talking about your chat service and I

think this um goes nicely with that is um how do you see AI impacting the roles of Library staff particularly in

balancing between enhancing services and potential concerns about Job displacement because I know that's

something that people often talk about within our own organizations yeah and I will say that

when we um first kind of announced the chatbot that we were were working towards a chatbot to staff there were a

lot of concerns particularly from um our reference Librarians who were worried about um you know their had a lot of a

lot of worries worried about um you know misinformation worried about their kind of their

their jobs being taken away um worried about impacts on the environment worried about you know all the things we worry

about um around AI um and we took that to heart and I think really tried to from a very early stage involve our uh

Librarians in this process so that they could see they they weren't just presented with here's a finished chatbot

that's on our website but um we explained to them we needed we need your help um and so what ended up happening

is the reference Librarians ended up um not only kind of originally they created the kind of the original knowledge base

for the chatbot which was based on our uh FAQs but then as we were adding more and more um questions and getting new

questions they have been through this process of looking at those answers that the chatbot gives um telling us if

they're you know sufficient not sufficient um giving feedback helping us adjust the prompt again so that we are

um kind of staying within our lane so to speak it was very important for us in part because of the concerns that we

heard from reference Librarians that our chatbot really focus on spaces and services and not try to answer uh

research questions now I will say we're still getting research questions uh it hasn't stopped uh students from asking

those questions but at least for this pilot period we've really tried to limit it to um you know your your again your

your garden variety um FAQs that that you're seeing from from patrons as they come in and that's been one way that

we've mitigated concern and again just really bringing in the Librarians into this process so that they're a part of

it they're learning about what the chatbot can do what the chatbot can't do um I think that's been really helpful as

well um I will say that originally we had this kind of idea that perhaps the chatbot would be an efficiency tool it

could maybe free up the some of the reference work take that off the plate of our Librarians um thus far that has

not proven to be the case in part because we've just invested so much time in in training the chapot and and

refining um maybe down the line hopefully down the line that might be the case um but that's not where we are

now so it hasn't been an efficiency Tool uh thus far um but I think we're going to have to continue to have uh

conversations with our staff about the role of these tools um in the library and um what that does mean for the

future of our work and how some of that work um might change as a result of of the use of these tools for sure does

anybody else have any thoughts you know what we're you know I mentioned before the little project we're doing at

UCLA to extract metadata uh from uh images and and um you know um unprocessed books and I I do think you

know the um the idea that it might um uh create some real efficiencies and some of the collections um uh work that we uh

that we want to do is real it's a real it's a real possibility that we can um get closer to the um uh uh to actually

processing unprocessed collections that um uh but I think of it like this I mean those those books had sat in our

basement for about 20 years um you know that work was not getting done it just flat out wasn't getting done so we're

not you know if we can use these tools to uh increase inre the efficiency so that we can get to things that we

haven't um been able to um previously then I think it's going to be uh a wonderful thing because we'll get to do

you know um uh things that make those collections richer and deeper and and even more

accessible um I I guess one of the things that I might say is that um you know and I think most of us in this room

know that right that with any technological Revolution your job changes right but it doesn't necessarily

go away aspects of your job leave they come back like they they move they shift like that's been the the story I've been

trying to tell people with this is just to say you know yeah your job will change but your jobs changed 700 times

since you took it three years ago so it you know it'll it'll modify itself over time um I started an AI Action Group in

my library um which has a number of different responsibilities uh one of what which is to um sort of hold the our

AI strategy um but that actually came later if I'm being honest um but part of what they're doing is is Outreach and

connection and and sort of helping our staff learn about AI um and really get a better understanding of it so that

there's not so much anxiety about it and maybe a little bit more excitement about what you could get from it um uh one of

the things and this is a little bit of an aside but I just wanted to say this cuz I I think it's actually a really

important like just thing to think about um so uh this fall at our our senior leadership team Retreats this is the

Deans the vice presidents the vice provos the provos the President right all come together every year for our

annual Retreat and in preparation for that our provos had all of us do that eight hour uh Arizona State um uh teach

Ai and teaching and learning program so if you haven't seen this or haven't heard of it um and you're looking for a

way to kind of get your team all on the same page this is actually a really good primer on like the basics of AI um and

having the entire senior leadership team go through that process together actually help some of the Deans

alleviate some of their anxiety about Ai and have them have a better understanding of what AI was and how it

might impact their their VAR like the various parts of their work um and really I think you know a lot of this is

about kind of mitigating anxiety right and so the best way to mitigate anxiety is to give people information that helps

them not feel quite so anxious rley can I add one other thing that was just it just occurred to me as

you were talking um you know a lot of times um uh we you know um we process a collection so we go do the whole thing

so we we process a a manuscript we digitize the manuscript we put it in the hotti trust and we go oh okay done we're

done um and then faculty members come and they say oh I want that table from that

book or I want you know this uh I think um you know the these these Pages or you know something out of this um thing that

you've digitized um and this is where AI can really help um you know to to extract data from something um that's

already been done uh it saves a lot of um time and hand coding um transcribing you know spreadsheets of data from an

old book into something that we can use and so I do think it also adds a lot of value to um uh to the collections um

just by um uh being able to work in new ways uh with things that we've already figured we've finished with and I will

add to that if you haven't seen it yet um Northwestern has been prototyping something that will just pull that

information from their itiz collections which is pretty cool um so anyway um actually I'm going to uh ask Tim a

question if he doesn't mind starting um what steps should libraries take to ensure the ethical use of AI

particularly regarding data privacy faculty rights and the risks associated with third-party companies and if you

know Tim he likes a soap box and that is my favorite thing about him so I love his

answers um so so I I really am going to come at it from a very specific perspective that that I'm at least um

tracking on so in May mid to late May of 2024 we had what looked like a distributed denial of service attack and

it was on our repositories and our exhibits um and ours our ears picked up first it was what are we going to do to

keep the servers up uh and that was the starting point um uh we ended up in less than a week blocking 4.3 million IP

addresses just to keep the servers up and running so I checked in with my campus colleagues and nobody else that

does technology on campus was getting slammed and then I checked in with my colleagues and in libraries and it

around the nation and everybody was getting slammed um and so I I was going to our leadership team saying I I think

we've got to block these for now just to keep the service is up and running but I'm starting to get a little bit anxious

and this is where I'm coming from uh UNCC collects the South and you can imagine what comes along with the South

including the Civil War a great deal of information about enslaved peoples um protest imagery and video oral histories

of living human beings um and it's not that we digitize that stuff with the intent to keep it closed our goal is to

make it open that's why it's there it's for collection is data it's for researchers um but this hunger to train

llms on human uh voices and human imagery is is such an extreme hunger and you start to think about how your

collections are going to be used in a way that you NE didn't necessarily intend for them to be used so yes I

would love this information to be in large language models for the big three or four that you are aware of but we're

also concerned about deep fakes I don't want an oral history with a very convincing uh voice from a real human

being to say things that they didn't say I don't want to see an image from uh tearing down a Confederate Monument a

few years back uh with somebody in it that wasn't there wearing regalia that's that's not theirs there's just a lot of

things to be concerned about so we took this opportunity not just to block the um block the scrapers for survival but

also to give us time to step back and convene a group to start talking about how do we want to engage with this and

the answer to that could well be leave it open but we wanted to have the time and space to have those those

conversations um I also mentioned uh risk so we have uh we University archives comes through the libraries on

our campus um and so our uh our digital repositories include campus records everyone clicks the little box that says

yes I read the terms and says that they have don't have pi eye in it and we always look at stuff but we only have a

limited number of folks but our our risk stance has been it's probably okay we've done a first pass and we'll crowdsource

this and if one person in six months finds something we'll take it down uh at the same time that's true for thesis and

dissertations as well and then um in the Carolina digital repository the CDR which is an IR the same thing can be

happened someone has an appendix with stuff they didn't mean to leave in um anyhow we're actually taking this time

also to re our rist stance because the the the opportunity to mine that stuff for pii just shoots shoots way up and I

just I just always keep sorry I always just keep thinking that they that that they are not paying attention to what

the right statements are on it anyway so they don't care yeah so anyhow that's that's less about bias or other kinds of

ethical things um I will say that from my perspective our staff largely weren't worried about losing their jobs a lot of

existential angst about the environment um sustainability about deep fakes about um I don't know the Terminator um and

then a lot of other folks who are having trouble seeing the relevance for their job so there's I feel like there's a lot

of translational communication understanding what's possible and helping people see how it fits their

jobs I will also say um I mentioned that the Fedora Community was having this conversation and that's um conversations

between Tim Scott Prader who's also here um we were talking about that and what Tim and I just had a conversation about

this morning was our catalog is what has been attacked um and my understanding from

vendors and vendors who manage infrastructure is your catalog is also being attacked you just don't know it um

and I think that is an important thing because it does go back to the ethics of of AI um who's getting access to your

data I think it's it's actually very relevant to conversations that have been happening about oclc as well so I think

there is some tie in there and I can just add that um our chatbot doesn't collect any information about users and

that was a very deliberate choice that we that we made um and we specifically tell our users not to include any um pii

in their responses however we have been thinking about um you know one of the things we've been seeing in in questions

that users ask the chatbot is about you know their own their own Library accounts about the catalog and so if we

in the future if we continue the chatbot and we want users to authenticate so that we could connect the chatbot to

various other systems um that would that would make this a lot more complex because then we would be dealing with um

with pii and um that would have a whole another layer of complications that we haven't haven't gotten there yet but um

we've started thinking about it if I could add just a little bit um because you know Tim's comments really

uh hit home with me at UCLA we have this thing this wonderful Digital Library program called the modern endangered

archives program uh and what we are is we give um uh small digitization grants for preservation to um folks in the

global South and um often those um those collections are at risk because of social justice issues or political

issues um and so and we've always been Advocates strong Advocates of Open Access um so we uh part of the terms for

us providing these uh funds for digitization is an Open Access uh license um and um uh it's become very

apparent in the last couple years that um uh these digital libraries that we put

up are uh experiencing um uh new fundamental use um from these uh AI Services uh that we didn't really intend

and a lot of our uh the folks whose content we've digitized you know they are happy to provide it to us with Open

Access under Open Access terms but they aren't really sure how to feel about them becoming um those Collections and

becoming part of the U monetization effort from Microsoft which um is a real struggle that uh Open Access piece and

um AI is a real um uh uh Confluence of issues for libraries because uh not only of the things that I think Tim and I are

talking about with regard to these these collections but also you we've been pushing open access for so

long you know um and now a lot of our authors are coming back going oh I didn't you know I didn't know

Microsoft was going to use my uh Open Access publication um you know they were anyway

I mean yeah all our Publishers are selling their content to these places but um it's become uh something we as an

industry as a cultural heritage industry I think need to talk a lot more about is what does open access mean uh in the

world of um uh hugging face yeah I think um aiva you had mentioned something about you notice more and more faculty

coming back about um agreements I think it was or yeah I mean we're starting to see

faculty who are coming in and asking questions about well you know I signed this agreement 15 years ago to put this

article here and I didn't know that AI was going to be a thing that I was going to have to worry about and now my stuff

has been included in these training models and and that doesn't seem appropriate and how do I get

compensated and what does that look like um and I I actually it's funny as Todd was talking I was thinking about do you

all remember the the on our backs brewhaha about like 10 plus years ago where there was a on our backs was a

reveal media we worked with reveal media to basically digitize the entirety of that collection make it available online

well this is a small lesbian magazine that was sort of shared with a relatively small group of people it has

naked images of people who are real life human beings who have jobs and you know all sorts of stuff that they never

expected that naked images of themselves would be available on the internet for researchers to use for whatever reason

um and so I think we sometimes are find ourselves in this situation where we have things that were created or

collections that were created with uh with an intent right or with a a hope and um and you know then we find

ourselves sort of further along in this scenario where we we like well we didn't intend for our faculty's research to be

used to train a a closed system that you know we don't know what it was trained on I mean it's you know I I think that

that what a lot of our faculty I think are starting to struggle with is this idea of like I really have no control

over what I've created um and that there is a real desire to claw back some of that control um and I'm not sure that

our structures are necessarily going to make that happen I just wanted to jump in with

three additional thoughts about um this one is that I just want to acknowledge this is an arms race that we're going to

lose um but we're we're doing our darnest the second is I've I've been in conversations with people like the whole

point is to be open and I think that's true when you're talking about research teaching learning uh and and I don't

know that it is true if it's repurposing to create new things that could be harmful or to monit Iz this back back to

us and the last thing I'll say is that I'm the most naive person in the room and the least likely to understand my

leverage but this is one of the few times where hundreds of years of collection development by relatively

underfunded organizations and tons of Labor to describe and make that accessible and digitize it uh is the

very thing that they're desperate for and uh I don't know that we're thinking about um how we leverage that if it's at

all possible because they're mon izing it back to our campuses and it just drives me

baddy um so the next uh question I wanted to ask was really about campus Partnerships

um how can libraries a Eva maybe you can take this um since you have a multicampus

partnership how can libraries collaborate with campus Partners like it and and the office of research to drive

effective AI adoption sure um I mean I I don't know that I'm going to have some magic bullet

for you and I think it obviously depends on the institution that you're at and the administration that you have at your

place of work um but I will say that I think one of the things that we've done a really

good job of is um kind of to Tim's point is sort of articul cating what we actually bring to the information

landscape right so this is our job this is our role this is where we work and this is why having us in this

conversation now is vital and important um and I I recognize that I happen to have a Provost who is a a very smart man

who is very sympathetic and really tries to understand what we're doing and recognizes that that's that like that

we're trying to do good work for the institution um and a vpr's office who wants to

collaborate and an IT office that wants to collaborate right we're a scrappy institution and so we have to work

together um and we want to um and so I but I what I will say is that I think you have to come in with a a not a like

yes to everything but a you know a okay maybe it's not us here but how do we help you figure out where this works and

then you get seen as sort of that that glue piece of of trying to get people together and talking about the right

things and bringing people in I mean I think one of the other things is we often ask the right questions um I think

Tim actually has a a good example that we were talking about earlier about like you

know you have a group of researchers who come together they're working on whatever it is that they're working on

they think they know the answer to it um and then you come in and you say but what about the students right how does

that impact students or or okay what happens if you want to transfer that data someplace else right

like we ask different questions and I think that if you can find Opportunities to just ask questions sometimes um that

that can really open up doors and invite you into conversations um I'll also say that for us on the system level um you

know what we H so look New York State generally speaking is a state that views investment in higher education as a

driver for Innovation and change right that is a rare thing I think in the current higher education Marketplace

to hear me say right so so recognizing that that's that that's my institutional that's my system scenario right um what

that means is that you know three months six months ago the chancellor said I want to put $10 million toward um toward

figuring out data right what do what do what do the the university centers and the PHD granting institutions

need around data and here's $10 million to try and figure it out sorry $100 million to try and figure it out right

okay well that's operational effectiveness that's research that's you know Rim systems that's all of that

we're getting that kind of investment and so we're invited into these conversations because we keep saying

here's where we have here's where we have value and so we do keep getting added into the conversations but I think

part of it again is that that all of the Deans um you know I don't know where Kareem is but like we've all been s hey

Kareem like we've all been sort of pushing on this idea that like data is the future of libraries and you can't

train AI without data and you and if you want that curated properly and you want it something that you can use you need

libraries and the work that we're doing to make that happen and that seems to be resonating so I don't have a good

answer I will just jump in and say I think that when it when it comes to the teaching and learning efforts AI on our

campus early on it was recognized that the library had a clear role to play and we have been um involved in those you

know organizing seminars and things like that from day one in terms of the libraries involvement with um it efforts

in the research office that has been really more me attending all the meetings I can attend to try to position

ourselves and say we have things that we can add to these conversations we have expertise we can bring to these projects

um again I think the the our chatbot we showed as an early example on our campus that we know how to do this work and we

can help you do this work too so we're trying to to show our value and and get into some of those spaces where we're

not naturally recognized as as partners or innovators um and I think we'll just continue continue knocking on those

doors and continue trying to be in all those meetings um um Todd's gonna give his

answer real quick but we're going to start taking questions so if you have a question please feel free to come up to

the mic okay um I'll keep it brief I just wanted to reflect a little bit for myself um uh this morning uh uh I guess

it was was this afternoon Cliff was talking um and he did a little bit of uh Tod's from the West Coast

everybody um he he was talking a little bit about sign posting um and sort of what it looked like when cni started and

I just was sort of struck by you know it's called the Coalition for networked information um and there have been just

a few I think seminal Technologies of the history that um I feel like have had um uh the potential to have a giant

impact on the work we do in libraries networking um uh was one of those and I I really feel like artificial

intelligence is the next one um it does a lot of the same things it wants to do a lot of the same things that libraries

do and uh it wants to do them as well as we do it um I don't know that it will ever get there but um the fact that

we've been doing this sort of thing for a really really really long time answering reference questions uh

describing um uh uh content we're incredibly well positioned as an industry to have a really big impact on

how these things impact our campus um one of the things that struck me when we were talking about our project at UCLA

last week um was just kind of how far ahead we were from a lot of the other units on campus um uh I was just kind of

flored you know because we have been doing uh this kind of experimentation for a while um so I guess I would just

my final thought on this is just do it just just do it I think all of our panelists would agree um you know a

large llm a small l l m um just do it because uh it'll be great I do want to grab the mic just to

say that at UNCC Chapel Hill the libraries uh to meet this need for students are creating something called

an AI studio and the principle is to democratize access to close the D digital divide to brive a home for

people to get from zero miles an hour to 20 miles an hour we're not necessarily going to push them into the stratosphere

but we never know where the next great idea is going to come from we also know that it'll have to evolve the campus

will have a different makeup and experiential level in a couple of years um but this has gotten great traction

with the chancellor the provos the Deans and campus it uh and the libraries are a great place to host it we have people we

have buildings we have hours we have expertise and um and we're really excited and I can talk to anyone who

wants to know more about that and it's about information uh are there questions from

folks oh here we go yay hi um my name is Yas shures I'm at James Madison University thank you for the

panel um my question comes from a context of working on collections as data for several years seemingly forever

um and Todd you were saying like our profession has a lot to offer into this AI ecosystem and I'm wondering if you

all are having conversations about how you wield that power in these conversations on your campus where

there's a lot of cross-campus interest in implementing a Solutions or platforms or different

agents but we have professional values um that can sometimes be in conflict with that and I also think this

is a time to really lean into uh what we learn from practices of community archives and that centering of

communities and the objects that the collections are about and then um kind of bring up the care data principles

into that equal footing with Fair um how is how are you tell me you're having those conversations I guess first of all

and then how are they going thanks well I can say that those conversations we are having them they're

very actually very very Central to some of the collecting efforts um uh but your your other question is how do we wield

our power on campus which I think is a really interesting question um you know uh I'll let my other panelists talk to

about this a little bit but you know um uh I I just did um um over the course of the last couple years a research data

working group to develop sort of a research data infrastructure roadmap for UCLA right um and part of uh the ask

that we made to invest um from campus was really an investment in the library to um uh

digitize um and process our um uh um uh Library collections um we we made that ass because a lot of folks wanted to use

our collections for uh AI experimentation so you know um we will That Power by being a partner we will

That Power by just being and doing what we good good work you know answering questions you know there is a human

needed in the feedback loop uh in order to make um AI better um you know we do that really well I think it would be um

wonderful if we had um uh Librarians you know prompt engineering um you know anything from a

reference question to um uh data management plan uh there's a lot of there's a lot of ways we could um uh we

could um be partners in the organization uh in and around our collections for AI are there any other questions we we

have about five minutes left oh have to it's not a question it's a comment K

bua ston Borg University the other Sunni hey yes it's a thank you not to the panelist

and many of you here because I've been advocating for a for a very long time and finally I feel really high level of

maturity so there's a lot going on now before like it was barely kind of conversation kind of now it's happening

and also in term of C campus Partners it's happening and people listen to you and they know you have the data you know

the know how and you know how to manage things and get things moving forward so just a big thank you and meta

note so thanks Karim and you know um these tools are hungry for the data the data that we have um our oral histories

our books our manuscripts our Maps everything these things are really hungry for it um uh but you know um you

you got to get out there and let people know you've got these things um otherwise your uh your integrated

library system looks like it's under attack metadata is the other thing they really want

hi uh I'm Jia from UC Riverside uh I also want to just make a comment uh that it seems to me that uh the library AI

Focus currently is rather leaning on our collections uh uh metadata uh it feels to me a little bit too uh inward looking

I also want to mention that uh large language model is not AI it's only small part of AI and CN and all the other

things Afra photos not llm um so I want to mention that we don't want to forget the other side of of the AI uh spectrum

and lots of our uh faculty members research does not lean on our collections uh the data came from

microscope um and Telescope came from lighter data collected from the road I just want to mention we don't want to

forget that and the library also has a role to play there thank you

yeah I mean we didn't talk as much about it in the session as um uh maybe we could have but um I think the teaching

and learning part of um of our work with AI I think is really significant and while as a panel I think we spent a lot

of time talking about our collections as a sort of training piece of of into AI I think there's been a lot of work and I

would imagine many of us but you know we like at UB have done a ton of work in teaching and learning and a ton of work

on um uh so we have um uh a portion of our English 105 which is you know our our freshman English course um we do a

significant portion of that we're responsible for all the the DAT the you know uh information literacy portions of

it and research and and whatnot um and we've already started doing more around um AI uh sort of understanding AI in

that context and we're expecting that that's just going to expand so recently um the system made a decision that AI uh

literacy is now a requirement for all of our students to graduate and that is something that is expected that the

library will have a much bigger role in that part of of teaching um so I I your point is well taken um there are a lot

of other pieces that we're doing and of course teaching and learning is also just one other small piece of where that

is but um you know we've also been doing um uh AI work with faculty about actually like integrating AI tools into

their pedagogy and like actually doing it in the classroom um and so we can talk about that too but we don't have

enough time to do but you're right 100% um we'll ask let you ask but I just want to thank everybody since we're at

time for attending um and hearing our panelists speak um so thank you all uh CLA do you want to um yeah hi I'm Clen

ol from the University of Leeds and it goes back to Tim's point about paying again I think for our collections being

used um in AI models and how can we work together to create some of the tools based on our collections rather than

paying the vendors again for the content yeah I don't have a great answer but I will say that the scale of

spending I'll just choose Microsoft is is insane and the need to recoup those funds is insane and I feel like it's an

order of magnitude different than some of the things that we've talked about many many times in our communities so I

I feel like there ought to be something we should be talking about collectively but I don't

know well thanks again everybody [Applause]