 Good afternoon, everybody. First of all, we, as we sat here and we thought, holy shit, they put us in this huge room. I wanted to say thank you for coming. We're here to give a brief presentation, some slides, about some of the work we're doing at the University of California. I'm Todd Grappone. I'm an associate university librarian at UCLA with me here, so Ismail and Peter Brantley. And together, we've been talking about some practical uses of artificial intelligence in libraries. I'll give you just a little bit of a background. If I, here we go. Okay. Thank you. We had a, at UCLA, a donor come in and want to give us a rather large collection of documents they had collected over the years. It was specifically, it is specifically political cartoons and decades and decades worth of these cartoons from all over the world. And of, like, every donor you have to the library who wants to give you stuff, they want to see it up online immediately. And I thought to myself, how are we going to do that? You know, there's really very, it's a very daunting proposition. So I talked to the donor a little bit about some of the, some of the things we've been talking about with regard to artificial intelligence and machine learning. And he was interested. So I got interested in talking to some of my fellow colleagues here about the work we're doing. We put a little working group together to look at what we might do. And I'm just, I'm going to just briefly talk a little bit about what we're doing. We're doing a study in the capabilities of artificial intelligence for academic libraries. The study will investigate artificial intelligent capabilities of academic libraries. Specifically, the study will focus on the ways in which libraries are using artificial intelligence to enhance services, to improve user experience and enable more efficient and effective library operations. The study will include a review of current literature on the topic and a survey of academic libraries to assess their use of artificial intelligence. The findings of the study will be discussed in terms of the potential impact for academic libraries to become AI aware and how implications and the implications for how libraries can best leverage AI to improve their services. The study will also suggest areas to further research in the area of AI and academic libraries. So that's kind of what we're up to. We pulled a group together from across the University of California from several campuses. We've had a few conversations and what we're, what we're going to do is have conversations like this. I mean, this is not a presentation where we are going to go through a bunch of slides. We're here to really have a conversation about what this technology is doing and potentially how we might be able to leverage it a little bit better. So I'll talk about what the flow for the next 45 minutes will be. I'm Salva Ismail. I'm the Associate University Librarian for Digital Initiatives and Information Technology and the Associate CIO at UC Berkeley. And what Todd forgot to mention or cleverly did not mention was that our opening statement that he gave was actually from OpenGPT. So right there, we wanted to bring in some practical use. So our flow for the next 45 minutes is we'll go ahead and introduce how this group was formed and what we're looking to do. We will then explore some ideas that we've been throwing around. And at this point, we invite you on this journey with us as we're exploring these things. Think about how at your library with the services that your teams, or you provide for that matter, what are the things where we can use artificial intelligence for more operational cases? And at that point, we will also want to learn from you, as Todd said, in a conversational manner. Where can this go next? So with that, I will pass this right back on to Todd. Yeah, so I mentioned a little bit about the impetus behind what we're doing. It's all you can stuff on my punchline, but that's okay. We'll soldier on. We started, I just want to say a couple of things, is we kind of believe in the quantum future. So we believe that a lot of the challenges that we have with regard to mass processing of machine data, or we feel like there's a light in the future. We believe in a future where storage is ubiquitous. And I know Cliff mentioned this morning in the plenary about sort of the challenges to the amount of research data that's being produced. We share those issues, I mean, our campuses, just like your campus, the amount of data that's produced every single day is astronomical. But we also believe that if we engage correctly with artificial intelligence, a lot of the fundamental library issues can be really improved upon. And we're hoping that through these dialogues and through this conversation that we end up doing something that's important and impactful. So, and I would say that, you know, just to reflect a little bit more on the plenary this morning, you know, the state of the industry with regard to artificial intelligence and machine learning is, it's the speed that that industry is improving is very impressive and it needs constant attention. Are we risk being subsumed by some other industry and some other community? And that's really not what we, not the vision that we share for the future of academic libraries. So, what happens after, you know, the digital deluge? You know, we have all these collections that we've digitized and we have all this data how are we supposed to keep up with it? We have faculty coming into our data science center at UCLA who want to use the collections we've digitized for research, for traditional humanities research, but also for text and data mining, for image processing, image visualization, sorry, image, I can't remember the word. They want to use all of these digital files we have in ways that, you know, we didn't really intend when we digitized them, you know, we did it for preservation purposes, we did it for sharing purposes, but now they want to take all that stuff we did and they want to do even more with it and we need to figure out how we do, how we provide access to our things through those workflows, but we feel like those tools can also help us with things like digital preservation and other systemic and challenging problems that we have in academic libraries. So, we did form this working group, all my notes are showing up. Anyway, we formed this working group to talk about, specifically our use case and to pull a bunch of different use cases together from libraries across the University of California. We came, we had a meeting, we had one special guest that came in from a non-UC campus and we spent a day ideating about kind of what these things could do for us and we came up with a nice list that we are currently sort of curating that has to do with things like representation and metadata, things like taxonomic browsing, using different taxonomies on different collections. Just lots of really good ideas. Okay, and so that's what we are going to be doing is coming up with a strategy to make sure that the libraries within the University of California are AI aware. It starts with a strategy, that's kind of what we're doing, we're putting together these ideas, we hope these ideas will stick and that we can rally behind a few of them in order to really find some impactful work. We've also become very aware through this process that we need to do a lot of educating ourselves and educating the people around us with regard to what these products are and what the work is that we can do in order to really have a good conversation. Yeah, the other part is really, as I mentioned, we do have these collections. They're really there for us in a very traditional way like other library collections are but as we have faculty members coming into our libraries who want to use text and data mining tools, who want to do computer visioning, who want to do some really machine intense processing with our own collections, we haven't really built those services and part of what we're trying to do is imagine what those services might look like and how we might as a library not only use these tools for ourselves but also provide services and access for our faculty members. Yeah, so that's really what we're trying to do is define our own future. So as we kicked off this discussion, the three of us, I was able to sit at a presentation from the CIO of the University of California System. His name was Van Williams and he was encouraging all verticals, all IT verticals, and I consider the library an IT vertical, within the system to talk about how artificial intelligence might be deployed within their market and how we at the University of California might help define the product and the product suites that come from these new technologies. And so that's exactly what we're doing. We're really, really, and working, hopefully, with everyone in the CNI community to think about how we might dictate what the future is for these technologies within the academic library context. So the academic library context, I wanna emphasize that a little bit. I think very often we talk in broad terms about libraries but academic libraries, I think, are really interesting and deserve a little attention in this specific domain, just because of where we are and where the researchers are who are doing this kind of work and how we might reach into those labs and pull out those tools for ourselves. And what that's kind of what we're trying to do is have that conversation about the academic library. We're gonna play a little bit of laptop musical chairs here. So when we started this conversation as a group, Peter, Todd, and I had several ideation meetings, disgrained storming meetings, and we really wanted to define a path forward. What was the path forward that we wanted? One of the things that we were certain about was, we wanted to define a path forward. We wanted to make sure that as we began this work, we wanted to underline that this was not a replacement for any kind of service that libraries provide or any kind of work that our library staff members do. It's actually just enhancing their work. So for those of you who may have used a chat bot on say AT&T or T-Mobile, you go in and you're like, hey, I need this help and somebody goes in. At my university, you see Berkeley, our central Berkeley IT system has chat bots for chat transcripts where you can go in and say, what time is the gym open? So questions, basic questions that our librarians generally get to answer. What time is the library open? Where are your bathrooms at? Where's the stapler at? I've heard that's a very common question. So how do we enhance the work that we're already doing so that our professional staff can engage in more complicated tier two, tier three, and more complex issues around information science and the work we do. And once again, as Todd said, wanted to underline here, we're focused particularly around operationalizing academic research library services. While AI and the products that could be developed or tools that could be developed using artificial intelligence can have benefit for the entire glam profession and the information science profession, our focus is particularly academic research libraries, born digital, digitized content, metadata services, reference, instruction, literacy, and others. So we did a little bit of a landscape analysis. We're aware of the work that's already going on with leading the future of AI and public archives. The AI for Lam community, that's from Library of Congress, Smithsonian National, the cadre, Aeolian, and other projects that are out there in the field. So why particularly the UCs and why our invitation to learn from you here the University of California system as the 10 libraries are one of the largest collections. We do have massive print collections, but we also have massive digital collections when we look at ourselves as one system, one library. And we do actually have one of the largest AV holdings outside of Library of Congress at an academic library. We also, for the University of California, nearly one-tenth of the national academic research comes out of the UC system. And looking at our users, the user base is really large, one of the largest. And finally, the past year in 2022, it's 2021, the UCs implemented the first ever system-wide ILS which was bringing all our collections together and having one general discovery platform for all the 10 UCs, our research library facilities, the two RLFs, we have a North RLF, a Southern RLF, and California Digital Library. So that gave us an unprecedented amount and access to not just the collections, but all the resources that went into putting CILs together. So our model here is to develop a series of project proposals that are not just impactful, but can be operationalized, something that we can actually implement and talk about like, hey, with X number of resources, X amount of funding, this is what we can do and then take it to our UC leadership and something that we can then accomplish within either bite-sized chunks and within a specific timeline. So with that, we have about another 20 minutes or so and this is where we would like to invite the members who are sitting here to help us ideate possible solutions for what this journey could look like as we embark on discovering and discovering, discussing, and collaborating on solutions that can be produced, collaborations we should look at. So as Todd mentioned earlier this year in December, November, when we brought together a gathering of folks from across different UCs, we had UC, we had members from UC Davis, UC Berkeley, UCLA, of course, UCSF, UCSD, San Diego, and from IU, we had ideas thrown out, we had a jam board, people could just throw ideas out there, we discussed those ideas, we had members from across metadata services to digital practitioners, to data practitioners, to pure technologists and I'll throw out some ideas here to get the conversation started, which was around, can we process and evaluate born digital collections such that there are certain templates that could be created where the machine could just learn that these born digital collections, these are the authors, this is where sensitive information or private information is so that the processing can be made easier or faster. Can we extract tabular metadata from digital files or how can we identify contents of an image to augment metadata and enhance discovery? This one's particularly interesting because in one of my PhD programs, we were working on image mining, but it wasn't just image mining, it was helping then the computer understand a person sitting on a sand or a beach, how does the computer understand that is a person and then tags the metadata of that photograph with beach, sand, person, so that the next level of cataloging or metadata creation can happen. Using digital libraries, creating subsets of digital identification, sorry, identification of specific individuals across university archives in multiple formats, like how do we do OCR for example on non-Western languages or non-Roman languages or how do we automate the extraction of text from not just PDFs but maps and other items that are being digitized or being given to us in a born digital format. So with that, I will open, we want to thank you for hearing us talk about how this group came forward and what we're looking to explore and invite you to discuss solutions, thoughts and ideas with us. Yeah, so I'll just, so I decided that one of the things that I wanted to do here was just try to put a different kind of frame together around AI, because I think one of the challenges that we have as libraries is that we're coming from a librarian background and so it's natural for us to think about our collections and metadata and the kinds of things that we've traditionally been strong at and to try to utilize AI to do better at that and to bring new insights and that's really a critical role for us and the kind of examples that a solid just gave are increasing, as they get more sophisticated they're increasingly non-obvious and that's really important but I think that there's also, there are different frames and I think we need to be cognizant of different frames and so I wanna sort of finish off our part of the presentation with an encouragement to not just think about what are your ideas but what are the different frames that we can think about AI in. So for me this was really crystallized in late October I helped with some colleagues including some friends that you see put together a small publishing conference called Page Break in San Francisco which was in part the successor to a Mellon funded program called Books and Browsers and Page Break brought together a whole bunch of different kinds of innovators in publishing but over a course of two days and we invited Tim O'Reilly to close off for us many of us have connections to Tim often had worked for him in various capacities and we had no idea what Tim would come up to the podium and talk about so you know he sort of came in sort of late on the second day and made it up to the stage and what he decided to focus on was AI and he talked about his history in publishing and thinking about various types of innovation and many of you may know Tim also has a pretty deep interest in economics and sort of disruption of economic landscapes and sort of economic ideologies in a way through technology and through greater societal change and he recounted some of in his career how there had been several times where at O'Reilly they had made decisions to wholly pivot from one course of action of one course to another course of action and so they did that early on with the guides that they developed programming guides, books with the animal series that you all very aware of they pivoted again in part through Safari books online where they started producing a great deal of video content as a way of reaching out to people to train them and they pivoted again in COVID when they decided to shut down their conference series O'Reilly doesn't do conferences anymore so in Tim's mind AI is one of these moments it's essentially a burn the boats put everything you have into this innovation set of technologies because it will really change fundamentally the way that you're doing business it's not just do better it's really change the way we do business and the example that he gave which really stuck with me was partnering with an external firm that's specializing in AI enhanced content basically they had their start working with sales catalogs online you're shopping for winter shoes can I show you a scarf that kind of associational trip to help sort of build a context around a user experience and so O'Reilly gave them all of their video content and then what they did was instead of thinking about it as a black box here's this video content we'll use this one tool and produce this output the innovation that they engendered basically the work of with a very small number of programmers was they set up a sequence of tools and that's what I think what we should think about is setting up sequences of tools not just a box that you put content into and something comes out so in their case what they did was they took all of their English language video content and then they had a tool that generated a transcription of it then another tool did a translation of that content into Spanish another tool married vocalized that Spanish transcript into the video another tool altered the facial pattern of the voicing of the word so it appeared as if the speaker was a native language Spanish speaker and behind that they had the transcripts with the kind of enrichment that we know how to do a lot of enrichments over tags and subject areas and this R.I. and so forth so ultimately a user could sit down and say or do a query like how do I set up a containerized instance of Fedora and then there would be several in-context representations in English or Spanish so this is an empowering update of beyond what we typically think of and I think one of the things that we are particularly challenged to do is to step outside of our libraries as my colleagues mentioned we need to develop services not just for ourselves but for our research communities for the larger community that we work in and serve which includes the public so one of the things that I want to try to figure out is how do we broaden the conversation that we would normally have just with ourselves who are the data engineers who are the AI practitioners who are the people working with other kinds of content and other kinds of associated technologies that we can bring together in a room outside of our discipline there are people thinking about this in broad areas of publishing there are people thinking about this in broad areas of other types of media content there are other people trying to think about this and intelligence and military intelligence why can that broader community help us understand about the fundamental revolution that AI can provide and how do we supercharge that conversation so there are a lot of smart people here we'd really like to hear your random or structured thoughts on your impressions of AI and if you've had things that you've been using about we'd love to hear them now or later but if anybody would love to or would like to start a conversation with us now go for it, please come on up go ahead Peter, thanks all right, thank you Peter Leonard from Stanford University Libraries I guess one question I would have for you guys is that many research libraries are different some of them offer services for faculty in providing AI solutions, ML solutions, data science solutions for faculty research projects sometimes this is called digital humanity support sometimes other broader terms and those for the research libraries that are lucky enough to be able to offer those services the things that are offered are often extraordinarily customized and hyper fit to the peculiar research agenda of a professor or a grad student it strikes me that AI in the entire sort of library work cycle is that we need things that are perhaps not so hyper fit and instead reusable, generalizable because we don't know if the next set of images being digitized from special collections are handwritten in English or Portuguese or something else so I guess if you guys have any thoughts about this tension between sort of really boutique services towards faculty research or institutional research projects versus a type of universally applicable AI solutions that might be relevant to all of library's work I have to stand but so yes I think that's one of the insights that AI can provide is that we often customize first and then try to think about what kind of flexible more general solution can be yielded from that it's like wow we've put in a lot of effort into this relatively baroque package how can we leverage that and I think one of the things that AI can provide us is to reverse that workflow it can help us understand or work with a more generalized context and then permit us to wring out of that very specific instances and indeed there is like with all of the AI that we've seen very much of an iterative process here you know so as we learn more about more detailed use cases it helps us understand a broader generalized case that could serve a broader set of problems but again I think that one of the things that we need to bear in mind is that O'Reilly example of tools in sequence or in parallel right it's not just you know we don't want to build a big flexible box we want to understand how to associate tools and services together in as many different ways as possible and I think that's one of the really hard things that right is to understand what that you know what that dynamism is in AI it's not just you know building a new you know 370 mainframe that can do an awful lot of massive computing it's about you know it's think of it more like a web stack in a sense right that here we've got lots of little pieces of software you know we have this maturity now and software design online that's really transformatively greater than it was five years ago and AI is that in a far larger magnitude kind of way so how do we get to that point? Yeah I would throw a couple other thoughts in there and in that you know libraries have excuse me have always had standard ways for people to get into our collections whether that's the front door or the catalog or a lot of other ways that people would get access to the website you know these I think if we use these technologies we talked about a couple here tonight this afternoon we have the same paradigm you know we tell people how to get into the library you know the new way might be you know a conversation you have with the library you know in the way we're having a conversation now I mean that's really interesting Peter talked about that the O'Reilly example I mean what if we had that a service where every student regardless of their native language could talk to the library and get a response back in the language that they spoke in a conversational way those are the things we're hoping to get to you know I think we do have kind of a really an aha moment here in libraries in that this technology will be there it will be there in our lifetime and I you know part of what I really worry about is that we've built these systems based on a kind of a paradigm around sort of words and concepts and these technologies are really a lot of smart people on my campus and yours figuring out ways where you don't have to do that anymore as we turn you know our entire information system on its head yet again in my lifetime and in yours how are we gonna position cultural heritage and science and that sort of discourse that's been going on on our campuses in a way that continues to make it not only relevant but accessible as well so this is why we have to have these conversations because that's gonna happen that's gonna happen Todd if I'm Todd and Peter if I expand on that so taking into account the example that you were giving Peter which is let's think of it as tools right like it's a toolkit we're building little things we have small pieces and together it builds this structure like Legos and Todd you brought in like how about a student in their native language could talk to us whichever language it may be and gets a response back in the language that they understand the jargon can be that could be looked at looked at as one Lego block and then could that be expanded further than based on what we've learned from that tool to now interpreting documents that we may have in our digital collections or documents that we may have we may have from our born digital collections and enhancing that piece by piece because more often than not I've done in my career in libraries and what I've done my colleagues and I've seen we do is we look at the final solution oh we need to get to X now let's build this behemoth complex product tool platform solution initiative whatever keep going to get to this major thing and what we're looking at is let's talk about reverse engineering that let's have these little blocks that eventually get us there but let's not think about what that X might be talk about small Lego pieces and then we'll see if it builds as the Eiffel Tower or the Washington Monument hello I'm Jason Clark from Montana State University Library I'm going to try to put this into words I'm terrified and excited by this technology I think and one of the things I think I would give advice to anybody who's starting to think about this Todd hit on how quickly this is moving and how it does demand a certain amount of attention I think it also demands a certain amount of honesty there are components of this it is successful that will sunset parts of our jobs and it's okay to say that out loud in fact I think when you're trying to be an advocate for this it's important to say that out loud so when I can go to a place like github co-pilot and help ask them in a chat interface to write code for me what does that do to software development or our application developers so this is not just about generating metadata it potentially touches all parts of our organization so what I would what I would guide us towards or something else to think about is you're bringing this these initiatives forward is honesty and also starting to build out computational literacy for for what this means what and how do you prompt these engines how do you build with this technology I'm excited to see where this goes thank you thanks I'm I just before I one of the things we were talking about earlier uh... in this conversation not necessarily today uh... was really the uh... uh... sort of that what we're all sort of feeling in this post pandemic world where everybody feels like there's too much work to do you know everybody's talking about how stressed you are and I was thinking about ways me as a technologist in libraries might actually do something practical uh... and I thought about some of this stuff you know where you can kind of uh... do things at a scale that you you couldn't really do without this technology and how that might necessarily replace everybody uh... but allow people to work in a way that they always intended to work you know something that's not quite so deadline-driven crushing but you know uh... that's more thoughtful and reflectable and I think this kind of technology can help us to do that as an industry I would also say that I think research libraries at least are in a stronger position than we might have been just it took a it took a revolution in network technology to get us to a point where we can talk quasi-intelligently about AI and so there I think we do need to respect what the impact of it will be and there there is an incredible stress here we are learning that cliff talked earlier today about the uh... rapidly evolving uh... understanding about rights intellectual property rights associated with uh... some of the creativity that can come out of AI and it's something that we really do not understand and we certainly do not have uh... law you know decided law on many of these aspects yet uh... but there's also this there is a an incredible tension here in this is a very fast-moving technology and we don't have two generations of staff refreshment to have suddenly a fully AI trained cohort walking into our libraries and museums in our uh... national agencies we need to train ourselves to open our own eyes to what can happen here and that is going to be a very high impact transition for us and we don't know yet what that will mean I'll add Jason that we've been following the work Montana State has been doing through the IMLS grant of the AI and ethics uh... work that's been doing so as I've probably mentioned it briefly that we're looking at it from an ethical point of view but the very statement that we taught and I were discussing this in the morning and we said was our staff are overburdened they're telling us the morale is low they can't keep doing everything they can't keep wearing fifty hats that we're asking them to wear right five hats in some cases are fifty so how do we actually provide use this AI technology maybe to provide some practical solutions again I go back to uh... why are we should we be waste should we be using utilizing our time answering where the stapler is or where the bathrooms are way-finding solutions don't work so are there ways in which those things could be operationalized or meet more practical so our librarians and our staff can actually do more enhanced work maybe they're not cataloging level one metadata musical scores anymore now they're providing more enhanced work on this again it's evolving we will be learning and I think I read a tweet earlier today which was we can't you know we'll need to experiment and learn and I'll finally end by saying when open gpt the chat thing on twitter took off the past two weeks and everyone's talking about it and we mentioned it was mentioned in the earlier plenary around well will we have authors anymore will books be published open gpt can totally write books and there was this author and I can't remember her name right now but she tweeted about it and she said you know I will still have a job I'm still the one coming up with ideas and thoughts and creating the words that build a story maybe in a hundred years it might be different but at least in the next my lifetime I'll say we may still be learning on how this technology that's evolving needs to change be updated iterated and improved I'm Derek Devnich I'm at University of California Merced and I want to talk to you later about a very specific project I'm not going to take airtime for that right now I want to tell an open gpt story and then ask a follow up question for him the open gpt story for those of you who have not been following on twitter there's a bunch of people who are writing queries to have it generate all kinds of things like a mission statement for your university which is tragically on point or a business plan or you know all or poetry all kinds of things but a lot of people have asked it to write code so if you ask open gpt a question in the form of an SQL statement it will hallucinate a database with the appropriate columns and then give you back tabular data that actually answers your question just FYI try it out it's amazing and the follow up question so I'm following up on on the concerns of our colleague from Montana State the problem with all statistical learning systems is that they do the best they can in all circumstances so they typically need adult supervision this is the issue like we get c plus student essays out of open gpt right and so the workflow question is a very practical question for any of these engines that we have is what do we how do you insert appropriate supervision at the appropriate point do you have someone basically copy editing vetting the stuff I'm wondering if you have any like like concrete thoughts about how that would be applied particular projects that you have or that you're working on could you repeat the latter part of the question just the end do you have any particular thoughts about how you insert like actual human supervision and vetting of sort of an AI process that's sort of doing the grunt work but maybe need some supervision I mean I think at first that's that will be implicit because we don't know what we're doing and so you know we don't have to try to figure out how the tools are responding and the output that they're generating and try to figure out how to improve the symbiosis between you know us as instigators of that sort of technology and and then the outputs that we get and there is a broad class of of AI systems where the one of the greatest challenges is we still don't know why they work or how they work more precisely and particularly I think in in fields where this is really critical like biomedical applications there's quite a bit of work trying to understand how to develop AI models that actually generate the optimal outputs that are preferred in you know in a given scenario in an applicable scenario and so that's in some ways sort of a meta level of understanding or administration of AI systems is is not to assume that the AI is the best AI right and and to try to figure out how AI is do the learning that they do and how to improve that learning process to get different kinds of outcomes so I think we have to not just work with the tools but understand you know how the tools get constructed in the kinds of outputs that they generate and you know leverage that so you know it's always creative analogy suggested you know this is very much going to be the case too in sort of the engineering aspect where we have to have our hands involved making it very basic from I was chatting with a reference colleague of mine from Florida that's where I'm from and one of the things we were talking about since you mentioned OpenGPT and I will say that we've been talking about GitHub co-pilot because it helps you develop your code you can basically write your code enter functions and real-time code with you but my colleague and I we were discussing around information literacy how do we bring in this adult supervision that you were talking about which is maybe at some point our instructional methods will change how we instead of asking students let's produce a paper that's three pages around a thesis topic that you've chosen we actually tell them yes use one of the AI tools to write a paper to get a paper written and now go ahead and critique that paper so what you're doing is bringing in adult supervision in a way which is you're using the technology to give you an output but then you're critiquing that output and then you can continue to feed the system and I will say I just learned this the other day I was looking at some academic plagiarism code of conduct and for the most part most academic code of conducts include talking about it's considered plagiarism or dishonesty if another human being another person writes the paper for you and as was discussed earlier somewhere uh... AI is not a person and AI is definitely not another human being so I want the implications in that case uh... as these technologies are developed will be beyond just libraries right it will be in our academic code of honor and conduct and other policies that our universities use I mean the only thing I would add is you know it's really part of what we're trying to do is answer that question because it's impossible to answer right now I don't know what how this technology is going to be deployed but I think a library website you might see in ten years is not going to have not going to be centered on the search box you know we might not be doing semantic searching across the catalog uh... you know we may be doing a lot of different things uh... and I think that's kind of the point of what we're trying to do is you know what does it look like when we don't do boolean searches when it's really this other kind of input uh... and a different kind of output and whether that's uh... you know we spent a lot of time talking about archival collections uh... we're also worried about this uh... other kind of data collections and how we might make those accessible our human uh... uh... interaction with people reference etcetera i mean all of these things i think are going to be uh... fundamentally changed by this technology in the next few years and we really need to we need to really think about these things uh... and uh... just to the degree we can as a community experiment with solutions uh... i know we're right at time but we'll take one more question we can take one more question but members of the audience please feel free to believe if you have to because we know we're right at time and thank you everyone well i'll keep this very brief then and it riffs on some of the points you've been making my name is Nathan Gerrith and the University of Nevada Reno and i wonder if we can reframe it from adult supervision to really capturing that synergy because i think the key in this is that we think of AI and humans and what has struck me over the past couple of weeks as i've seen my wife who's an academic grappling with some of these plagiarism issues as i've seen other people just experimenting like dumping data into these things to see if it hallucinates a data set or something like that it's that creative synergy that's really going to drive some of these developments and i think building in a way that we can uh... embrace what's the term uh... the capability overhang right and find those moments captures those moments when people are a little bit creative with this and do something that's oddball is supposed to kind of doing the obvious because i think that you're spot on when you talk about this not being a reformulation or refinement where we at as library professionals but it's really fundamentally turning it on its head and so you know you can choose to comment that if you're on time but i do think that it's that formulation of AI versus humanity that is sort of misleading that really is the combination is that synergy that's going to really be transformative i think that's a great way to end this panel so thank you for that and i would just you know sort of underline no one of the things that i think that we all need to look for is to be surprised and uh... and to look forward to that because that is what will open up our eyes thank you all for coming we really appreciate it