Ep. 6.1 ADHC Talks Podcast: A Conversation with Xabier Granja

Welcome to the ADHD Talks podcast produced by the Alabama Digital Humanities Center at the University of Alabama Libraries. I'm your host, Sarah Whiter. This video podcast documents digital humanities research through semistructured conversations with DH scholars and practitioners. Our guest today is Dr. Shai Granja. Dr. Granja is associate professor and director of undergraduate studies in the department of modern languages at the University of Alabama. He also works closely with the artificial intelligence teaching enhancement initiative. His research interests focus on historical crime specializing in 16th to early 18th century Spanish. His work deals with personal identity in the early modern period, the literary and sociopolitical implications of individual identity formation and variations in didactic literature. Shavei is co-founder of human AI, a multi-institutional research organization that explores AI in the humanities. He is project leader for Renaissance, which seeks to leverage the power of AI to transcribe early modern Spanish printed texts. This project has been chosen for inclusion in the past two years by the Google Summer of Code program. >> Hi Shabi, I'm so happy that you are here with us today. >> Happy to be here. Thanks very much. >> I uh I'm really excited about talking about your work. Uh we haven't talked a lot about any computational projects on the podcast and um I'm especially curious towards the end to see if we can talk a little bit about your Google Summer of Code participation because I think that's something that a lot of people >> should pay more attention to. >> Absolutely. >> Yeah. So I'm going to jump into our first question. It's a standard question that I ask everyone. It's my passion question. It is the one that I love the most. What makes you feel nerdy? >> You know, I feel like when when you feel nerdy all the time, there's not one thing that makes you feel nerdy. Um I do like to say that learning is what interests me really about anything. Uh all the podcasts that I listen to are essentially me learning history about things. Like there's one that I love that is called the spot was this podcast will kill you which is about viruses over the history of humanity and they give you like history and then the biology of it and it's so entertaining to me right which is probably very nerdy because it's just learning about something has nothing to do with what I do but I will learn about that or I will learn about just doing plumbing in your house which I have done myself but just YouTube and so I think that makes me pretty nerdy to just do the plumbing myself before I call a plumber. >> So learning learning to maybe to a fault. Yeah, I think curious mind. I am currently listening to an episode of the Nature Podcast on a book called Apocalypses >> and it's about all of the different identifiable apocalypses in >> in the history of the universe >> and how humans have survived and reacted to them. >> Um, instead of like putting the apocalypse as something that has never happened but we're all just waiting for. >> Right. Right. Right. We survived what? 2000. We survived 2012. What's the next apocalypse? >> That's right. I'm sure. >> That's right. So, so curious person, tell me how you got involved in digital scholarship because Okay, first doing before that can you um summarize just a little bit of sort of what what your research is and then go ahead and talk about how you started uh using digital methods. Yeah. >> To to do your your research. >> Okay. Yeah. So um what I do is I teach um I teach a combination of literature and history although I am essentially Hispanics literature professor. Uh I combine both of them because I can't just uh work on literature devoid of historical context. So I I rely heavily on the history. Um so I teach a lot of historical crime. I research a lot of historical crime mostly has to do with marital crime. So that kind of also gets me a lot into gender issues usually of um how men were supposed to behave, how women were supposed to behave in the Renaissance, in medieval times, which also kind of pulls me back to even like Aristotle, like Greek times, right? Because those are the bases then kind of evolve um up to the Renaissance and and and after. Um so I I look a lot uh into all those things. Um a lot of killings, a lot of uh sexual violence. So this is kind of my life and my research, right? This is what I look uh into, what I research. Uh and then I kind of cross reference that with literary works that are supposed to represent specific given time and what that says about either the aristocracy or the lower class. This is kind of the analysis that I usually do. Um so this is not usually something that screams, you know, high technology and like artificial intelligence. Um I've always been drawn to technology since I was a child. Um, I think much of my interest comes from video games themselves. I've always been a video game player and that kind of got me interested. I'm fascinated by the advances in technology. The 90s were very good to us cuz the the jumps in in performance were just out of this world. Now it's different, right? It's it's it's slower. But, um, that kind of got me interested in how things work, how things process, how we process data, which is not, you know, so um, divorced from how my brain process data, right? when I do my research. So, how does a machine process all of this? Um, this is kind of how I've always been very much into the technology side of things, the digital humanities, how I can uh make my teaching more digital, my research more digital. I was very fast into getting comfortable in working in the cloud and then using different digital uh tools with my students etc. Uh so the way I kind of got um much more into artificial intelligence I had used it a little bit other than you know you have like your other Google home speakers where you ask it things and they can answer you and this and that. Um so I'm currently an associate professor and my dream project really is about this one woman who's called Louisa de Pada. She was an aristocrat in the in the 1600s who was very very religious and she a very heavy Christian and she saw all of this vice in society that she really wanted to eradicate because the nobility of the time went through many many changes in what's aristocratic values and what is not appropriate for aristocracy. Uh so she wanted to reook all of this with religion and she has four beautiful toms that are super interesting but they only exist in like Google books scans. There is no transcribed version of this. It's a little hard to transcribe. Um but there it doesn't exist. And so I want the world to have that in a transcribed form, published format that anyone can actually access and read in either a digital um element or even to facilitate research, right? So that you can search through the text body etc. Um this was my project for after I become full right um a full professor. So then I started looking at all of this uh optical character recognition and the things that AI can do which again kind of started with my video games where I can there are tensor cores that can kind of extrapolate from a small image because since I play my video games I can render them at a lower resolution and then blow them up to 4K with tensor cores that are predicting the detail that is supposed to be there. And I kind of started thinking well can I do that but with text when when we do that optical uh character recognition and it turns out that there's not much work on that but there there is some work being done on that and I thought why instead of waiting until I become full professor and actually start working on this and manually transcribe hundreds and hundreds and hundreds and hundreds of pages can I train an algorithm to do the bulk of the work um and then I can just, you know, uh, fix, edit, and get it to a final polished version. And it kind of gave me a win-win scenario where I can get the dream project that I want to do, whether it's by the time I get full or after or meanwhile or before. But at the same time, it's a really interesting research prospect where I get to learn more about how computers process and how AI processes data and how then I can apply that to the humanities because AI and humanities are not usually a connection that you see. um the world seems to think that it's only about computers and only about other science uh you know STEM projects like no no we can definitely use high-tech technologies in the humanities we need them we just need to remind the world that you know for the project for which we can use them and this is kind of what converged those sides of my life that have always been parallel and now I'm able to kind of like put them together really >> it's always so nice when that is able to happen >> absolutely >> yeah I'm thinking about like specific concrete examples of what you're talking about with the like the video game ex like expanding the so I um I'm a big Skyrim player and I play on my Switch >> but my Switch is docked to a projector that is a full wall >> and so usually after I leave my office in the afternoon I go for about an hour and play Skyrim which is my entire living room wall >> and The fact that it can render so beautifully on my living room wall is just it just blows my mind every single time. >> And that's part of what the Switch 2 now has actually moved to tensor cores because it is using an Nvidia Tegra processor. In this case, it was before, but the new ones basically take a smaller resolution and through AI through machine learning and using those tensor cores, it's blowing it up to 4K. So, it will look even better. That's what I have in my computer, too. And that's that's why, you know, I have a piece of hardware that I use for gaming, which is great, but I can use that piece of hardware. The same tensor course can process machine learning. That's what I mean, that's what uh blowing up an image and resolution is. So, you know, it kind of doesn't feel like I'm being efficient if I'm just using it for gaming when I could actually uh as we talked later about the Google Summer code, I can use that kind of processing and do it at home in my gaming GPU. >> Yeah. So, you've brought up you've brought up, you know, how you've gotten to this new project. Um, can do you want to talk technical stuff about the project? What what would you like to tell us about the project? And sort of um because I think last time I heard you speak, you were talking a lot about how you don't code, but you're involved in this huge coding project and like conceptually you have this vision and how have you made those connections? I guess just walk us through how you as a non-programmer >> figured out how to get started and how to make connections because that's I think one of the big things in digital humanities is that a lot of us have these conceptual visions and no technical knowledge. >> Right. Right. Right. Right. Yeah. No, that's me. I have no idea how to code. I I know about humanities, not about coding. So, okay. So, >> I'll let's go through a little bit of how I got to Google Summer code, which kind of then explains a little bit how all of this happened. And then I can show you maybe like a couple examples so that you see a little bit of what we're working on. So, when I started what we were talking about before, right, like I had this realization. I was like, well, how can I make this happen? And of course, now the next step is well, how how does how do you get to that point, right? How do I get to do this if I don't know how to code anything? So, I am the type of person who will never stop asking questions to an annoying point. But I I've always maintained that I would rather ask a million questions than not know and lose an opportunity or something doesn't get done because I didn't ask a question for fear of looking ignorant or stupid. Like, no, if I don't know, I'm going to ask. And people who know me know that even though I can be a little overbearing with questions, it comes from a place of improvement, right? How can we move to a better place? So I ask a lot of questions. And so when I started thinking, well, I don't know how to code. I don't know how to do any of this, but I do have the idea for the project and I have an idea of how we could work on this. So I'm going to ask the people who know about this stuff. So I started talking to um several people in the sciences. This is how also why when your university organizes u campuswide events or like collegewide events attend because you're going to find people that you usually never find because they're not in your you know in your immediate sphere >> and you're going to talk to them and I will ask them questions and this is kind of how I ended up talking to a couple deans in the sciences and they told me you know what like this and that person might be good people to contact in these departments and this is kind of how I um I met uh my colleague Dr. Sergey Glazer, right, who is in a science field, nothing to do with mine, and he was interested in machine learning. So when I asked him like, "Hey, I'm trying to do this. I'm trying to do a kind of an machine learning based transcription method. Um, do you kind are is there any avenue that you can think like how could we make this maybe work? Like if you have any ideas, I would love to get you a coffee and like talk together." And that's exactly what Sergey and I did. we got together and this is I told him this is what I want done and then he told me like okay these are the several ways that we could go about doing this right technically speaking and so that coffee kind of turned into another meeting turned into another meeting and we started getting a process like okay this has legs right um this is when he suggested something that he had been doing for several years with other uh science programs which is the Google summer code and he told me this might be a really good way to start this work and see where it takes us I had never heard of Google Summer code before. And what it is, it's a pro, it's a summer program that uh Google organizes every summer. It's not guaranteed, but it's been going on for like 22 years. So, you know, it's probably going to keep going. >> Um, >> and they organized this uh event that lasts for it depends. There's like small, medium, and large projects, but anywhere from like a few weeks to like three, four months more or less. Uh we are actually right in the middle of uh G-So Google Summer Go GOC uh 2025 uh which will end exactly in 11 days. So my projects will end October 6 is the my students final exam presentation that they have right evaluation. Um so the way this works is that at the beginning of the year you kind of propose a pro a project right uh you have to create kind of like a um um company of sorts. So they have to have a website all the projects etc. and you want to have a good amount collaboration between different institutions if you can because that will make you you know a more stable proposal. Um and we proposed it like this then Google accepted it as a project and they're like okay we're going to include you in in the summer program. This was this is my first GS. So last year was my my first one. This is my second. Uh so after that candidates from all over the world are going to apply. If they like your project, they they will look at all the many many pro projects that are proposed. If they like your project, they're you're going to kind of give them a test of sorts. They're like, well, try to solve this. And in our case, it was a very baby version of a transcription of just a few pages. Uh we told them like this is what we want. See what you can do. And so with that, we then evaluate those candidates which come from all over the world. Um and we submit to Google. These are are are the people that are finalists. Um, would you be willing to fund any of these? And then they'll tell us how many they fund. We've been lucky. Uh, in the past couple years, we've had a good number of them funded. Um, we had four the first year and this year we had more, which is not usual, but I think they're they're liking our project. So, they're they've given us that um that weight. Um so we have um several programmers from um from Japan, from India, from the United Kingdom, uh from Finland. So and from the United States, a little bit of everything, right? Um these are usually either fourth year or so undergraduate students in in the sciences or master students. Some of them already have finished their master and like are also they have a job now. They're actually employees now, right? So they they're kind of juggling a little bit of everything. the summer code is a very good opportunity for them because it's great for their CV and it's great for us because then it enables a project and a and a group a research group to exist that when I where I don't know how to code they are my coding team and this coding team enables me to um get that um infrastructure ready and I kind of guide as a humanist I guide where I want us to go um but there is a lot of uh I mean most of the bulk of the work is done by them to for I'm very grateful to our team and we also use this research to then work on publications, presentations because most of them are still students. So we push them a lot to present this work at a conference. Let's work on small papers like we also published an article last year from our first year of work. Um, and one thing that I really like about the Google Summer code is that they put the students first because you could see how this dynamic could end up being exploitative in some cases where some faculty might just, you know, do all the grant work with the students and they get all the credit. You must must must credit the students first in any presentation or publication. That's a Google rule. Otherwise, you risk that they will not renew you because then you're yes abusing the system. Um so let the publication we did last year all our coders are credited as first authors and the professors come second right and so they all now have a pretty decent publication in their record as undergraduates or you know master students that can only help them right so that's kind of a little bit of the of the general how I got into this and how it's working so far and let me show you a couple examples let me share my screen so you can see a couple of these I'm going to share my desktop up. Do you see it? >> Yes. >> So, this is what we started with. These are This is the Padia text. This is what I really want to transcribe for the future to create for the world. Um, you see that it's uh it's print text. So, it's easy in a way to to read. It's it's still difficult to for a machine to read it because it's not um the expected printing that we have these days. There was sort of a standard printing um back in the Renaissance, but each printer kind of had their standard. So there's many standards and also it kind of depended whether they had coffee that morning or they went angry that morning or happy or they went for a walk because their standard also varies from day to day, right? They make many variations oftentimes because when a when you use a press, each of the characters um would be made first of wood depending on the money that the press would have and eventually later in history towards indust industrialization it would be made of metal. So metal you can press and press and press you're going to be fine but the wood will crack eventually and so when you lose letters you kind of start interchanging letters. So you change V's and U's are very similar so you kind of interchange them. Fs and S's also kind of changed. So that is and there's no rhyme or reason, right? So if you see here for him, this this word obispo which means bishop, that's an F, which really that is an S. The word is spelled with an S. Sometimes you will find it with an F. Sometimes we'll find it with an S, right? So there's no way to to really predict and that's what makes it hard. And this is kind of like the the way the whole text will look, right? So this is this is where we started. Uh from this we've moved on a little bit to some other more complicated things like handwriting. This is something that we've been doing this year. This is an easier one. I could myself transcribe this manually somewhat comfortably. It's I can recognize the words quite well. I can read it, right? Whatever it is, right? I can read it. If I can read it, I can describe it. And it's quite clean and it's pretty well preserved, right? All of these documents come from different archives uh in Spain that I have visited and I've been compiling and compiling data. This, for example, was not um digitized by anyone. This is just me taking a decent picture and then processing in Photoshop so I can have a decent quality material to work on later. Right. >> Yeah. >> This is kind of how it looks. This is where we've moved on this year to try to do print and handwriting, right? Because we have more coders. Sure, we're having we have more projects. Uh, and getting a machine to read this is very very hard. >> And then I'll give you one more example for you to see. This is this is my nightmare. This is what doesn't let me sleep at night. I can't read this. I mean, I've tried it. It looks like an alien ship brought aliens to this planet. They gave this to the planet and then they left. They're like, I I don't know how to read this. Some of these are really complicated. um this we're not even bothering with this right now because it's too complicated, but these are the things that we're trying to read eventually, right? So that gives you a little bit of an of an idea um of the materials that we're working on. And the whole process is just to have our coders get the code so that um it reads several lines, different lines, like it kind of cuts the text in lines, processes it. We use a lot of ground truth. So we've um and I don't know maybe the listeners might not know what ground truth is. So that's just it's a piece of text that basically says this line says that this line says exactly that. Right? So that that way the code will learn what an R looks like, what an M looks like, what a B looks like, right? And that way start to learn how to read. It's not unlike children learning to read when they're young, right? And that way I've done a lot of manual transcriptions also with my colleague um Dr. Harrison Meadows from the University of Tennessee Knoxville uh who is also my colleague on this project. And so we transcribe a lot manually to to have more ground truth. And the goal is to not have so much manual ground truth eventually so that the code can actually learn. But for it to be good enough to learn itself, you need to help it. >> And we're kind of in that middle ground where like it's getting quite good. We've gotten so far to like 95 97% accuracy. Sounds really good. sounds like an A+ and it is, but that's not good enough when you're doing thousands of pages. A 3% error rate is hundreds and hundreds and hundreds of things you need to fix. So, I want to get to 99.99 perfection. We are getting kind of um I I don't want to say yet, but we it kind of looks like we might have gotten to like 99 point something% just a couple weeks ago, but I need to I need two more weeks for this specific contributor that's working on something, some updates. We'll see about the final project in two weeks when they finish because they they have to give me the whole presentation for the whole summer. Uh the the last code that we'll evaluate to together with Dr. adviser who's the you know the machine learning specialist uh we evaluate the code uh of like the quality of the transcription and then this year a new thing that I really wanted to do this year is whatever code we have I've tasked everyone with making either a web-based or local offline app and we're going to release several little programs where you can use the the AI algorithms we've trained and release them to the world for free all of this is going to be shared where um and I'm going to be sending this to a lot of academics all over the world that I know and like just try it. Try it. Try to transcribe whatever your transcription needs are. Try it. Let me know what the issues are so that then next year I can go back to Google and say, "Okay, this has been really good, but these are the issues. Let's make another summer program project to fix this so that we can evolve and improve." Sorry, that was a very long answer, but this is kind of the the >> I love it. I appreciate you um explaining some of the more technical things like what is ground truth because um I know what ground truth is but I didn't a while back right and I think that >> um I think that there are a lot of people es that I know on our campus but I think this is a big thing in digital humanities right now people wanting to transcribe old documents and be able to produce a critical edition or, you know, some sort of um either just to to promote access to it or >> or you know, some of the stuff is just really cool, right? >> And OCR for historic primary sources is so complicated even if >> like your first example, it's in print. Um, and for example, the stuff that I'm working on is uh much newer. It's from the 80s to 2000, but it's still >> Mhm. >> producing the OCR for it is so difficult because just like your, you know, Renaissance printers, it was a small publication. it was a historic newspaper and it, you know, it doesn't have standardized formatting from page to page and it doesn't have standardized heading or titles or and so you know you you run that through Python and it just gives you from this side of the page to this side of the page no matter how many columns are there and no matter what the the like the font change or the text change or or you know there's half of a title along with all of these other things. And there's it's it's been very challenging to train a model to recognize this is a column like there this white space around the column equals it doesn't go to anything else. >> Correct. Correct. >> And add to that complexity the fact that we don't have millions and billions and millions of dollars to either pay people or train things. We don't have that. Yeah. So, this is kind of where something like Google Summer code is so useful because those students need the experienced opportunity. >> They get paid a little bit. Yeah, I don't know exactly how much it is, but they get a little bit of monitor incentive, but it's really just about the pedigree of, hey, I was accepted at this Google program and I'm doing this project, which is really cool. It's very good for them and it's good for us because then we get the support we need. Yeah. >> Um with because we don't have that kind of money. We don't have that kind of funding in the communities, right? >> Yeah. I wouldn't be anywhere with my project if it wasn't for Jeff Turner um who was a previous guest on this podcast as well. Um but Jeff is one of our colleagues and um has spent quite a bit of time uh working on training machine models for um OCR of historic text as well. His example is um the US congressional record which also is not >> um super old but you know early 20th century late 19th century is still >> and even with the congressional record it's non-standard formatting because they go from you know publication to publication and change the columns and change the margins and change the font. And if you think about it, it kind of boggles my mind to think we live in such a digital world where everything is now preserved in some way digitally that the the vast amounts of knowledge that are recorded on paper that are either digitized but we cannot access because it's a picture. It's not actual reachable text. It's it's mind-boggling how much knowledge there is that we need to transcribe not just for preservation purposes which is an important goal unto itself but to be able to access it because to me it's the same difference as history. History is no good if you don't know it so that you don't repeat those mistakes. This is the same. You need that knowledge. Not just because it's nice to preserve materials. And I think it's crucial and fundamental to preserve them, to know who we are as human, as a humanist. That's obviously what I'm interested in, but also because you will not be able to grow as a person, as a human, as a society if you don't have acute knowledge of what you've been before. Because you don't know, you know, it's like if I start driving my car, I will eventually drive into a tree if I just don't have a direction. I I use Google Maps so that I know where I'm going. So, I need to know where I'm coming from to know where I'm going. It's the same thing for humanity as a whole. And the fact that all of these documents are >> not digitally preserved in a reachable way >> blows my mind that there has not been more uh work done on this. But it also at the same time it just, you know, I like to say that when when you're kind of pushed against the wall, if you don't have the funding and you don't have the knowledge, like I don't know how to code. If you don't have if you're against the wall, you can do two things. accept that you're against the wall and then end or re reject the wall, break the wall and then go to the other side. But just yeah, >> you need to not be in that position and find what can I do to actually advance in any way. >> You see what I mean? >> Yeah, absolutely. I think that's one of the true gifts of generative AI at this point because while we have all of these, you know, privacy concerns and ethical concerns and climate concerns, we also >> suddenly non techchnologists and non-coders have so much more access to some of these very technical problems that we can solve. and and they were not solvable even even if we had people with the technical expertise. The amount of time that it would take them >> to produce the coding without machine learning and without being able to train a model um it it would just it's it's just completely out of reach >> until now. And so suddenly this stuff is is in reach and it's >> it's a whole new world. the the opportunities are it kind of sometimes hurts my brain when I get excited about this is like there are so many potential opportunities that I don't even know what to do like you can actually get a little peerless by how many things you can do with it so just like focus on okay these are the things that I can do these are the things that it enables let's go get them >> right right yeah the scope creep is a little difficult to deal with sometimes >> but you know you start it's kind of like uh people have always been a lot of people have been afraid of the white the blank page. I've never had that cuz I just start writing. I don't expect what I what my my first sentences to be good. They will probably never exist in the final version. But you start somewhere instead of being blocked by the possibilities. Just start somewhere. See where it takes you and then you'll backtrack. You'll go back and fix and and change because anything you do will evolve. And our code even has evolved and and the reach of our project has evolved. like we would never even thought of doing handwriting transcription last year, but as if we've gotten the work we've done and we've seen like this has actually quite a lot of uh uh of potential, we we've gotten a little greedy like well let's let's expand to see what happens, right? >> Yeah, absolutely. Right. I have one more standard question for you and that is >> what's next? And I think you've sort of hit on it a couple of times, but right like >> uh you're welcome to include like bucket list things or like pie in the sky dreams. And I know >> from what I hear, you just decided that your pie in the sky dream was going to be your real dream and you're just sort of like I'm going to do this. So like once you finish that, what's going on? It's kind of like when I fix something at home, I will try to fix it. Even I'll do electrical and if then I fail, then I call the electrician. Yeah. >> At least I'm going to try. So, I'm trying this. It might hopefully I'm thinking I'm going to get me the bulk of the of the prescription done that then I can I can fix. But, um, what am I doing now? What's coming next? Well, right now I'm still focusing. I'm doing this. I have a lot of things. My my figurative stove has lots of pots and pans going on at the same time. So um I'm working on my second monograph uh which is all about uh all these historical crimes etc that I'm working on. I'm work a lot with the power of um of blood. Blood as an element in like judicial documents. You know it's not the same if someone uh hurts someone in either a sociopolitical way or a physical way. It's not the same if there's bleeding, right? Uh you there's like aggravating factors to different um two different processes. So, I'm working on that, finishing that book, which is really one of the records that I need to do to become a V professor. Um, that's about twothirds of the way done. Hopefully by next year, I'm expecting to have the manuscript completely finished. Uh, and that's as I'm working on this other thing, right, on all the as. So, the goal then is obviously, as I said, to get to the point where I can just completely transcribe these four tones and release them in some form to the world. And after that, I'm already kind of like now that that's become more tangible in a way because it used to be like I don't know how I'm going to do this to I'm doing this slowly, right? And I'm I'm finding all this new research while I'm doing this. Um really the the the next step after that and I don't like to get too far behind because sometimes we can get we can think too far far into the future and then we forget that we're doing things right now, right? Um it's a kind of growing beyond that um four tone transcription project which is okay now what can we do with this that can not just transcribe for other things because really my my goal has become I want as many human humanities researchers to have access to transcription tools that are free as possible because I always joke that you know I thought of starting to do this I don't have the code and knowledge to do this and I'm surprised that you know multinational companies are not already working on this. They don't seem interested. I'm guessing because it's it's more it sells more to have some kind of image recognition that tells you like you know this is a car, this is a cat or whatever it is like people use that in the consumer world more than this is a pretty specific uh specialized professional application of a >> um but if baffles me that no one's really doing it right. So we are in like as as little programs like you are doing it too like it's many of us researchers kind of doing this in little programs. >> Uh so you know if I can become a millionaire out of it then why should anyone else if it's if I have this idea why should somebody become a millionaire for my idea. So I want everyone to have it for free because then everybody wins right it's a much bigger win for everyone. >> Absolutely. >> And that's why we want to release the apps for everyone and then see what what use people have. Um, but yeah, this will kind of bleed into over the years and into not just OCR, but I want to get more into like the creative side of things. There's a lot of um text to video and AI happening now with like um Lumal Labs, Google V3, Runway, like all of these they are producing really interesting um results. They're also quite expensive and you have to have a subscription for it, which again funding wise we don't always have. So, >> right, >> I would like to get into some of that because then it's not just creating transcriptions of text. Then it's creating I could actually create video of whole like scenes or or things like that are really good for teaching materials and also just to absolutely more right. >> Yeah. I'm working on a huge um project because you know the library is in different stages of renovation. Soon we're going to be doing the first floor and the first floor is going to turn into the digital scholarship zone and we are going to have VR equipment. >> I was just thinking of VR. Yeah, same. >> So I am deep into the researching and writing a proposal for you know what UA libraries is going to do with our VR program as we move forward. and the timing is right. >> Yeah. >> Right now, um, and I think, you know, we've had some VR headsets that have circulated in the past and we've had some that we've run a little bit of programming with, but we haven't had like a a formal VR program. >> Sure. >> Um, and I'm excited to see how that shapes up. I really am excited about the period of time that I have in the next couple of months to sort of circulate around campus and see what people's VR dreams are and like >> what people feel like they need and and then figure out, you know, what capacity we have to help with that, >> right? I mean, >> we won't be able to help with all of it. >> No, of course we can't do it all, right? You only have one lifetime that I my my husband often tells me like, stop trying to fix the world. You can't fix everything. You don't have time in a hundred years if you get to live a 100 years if you're like >> so you have to choose something. But as you were talking about the VR headsets and the VR space, it also again it brings me back to humanities and it brings me back to video games because there are multiple um I mean I I'll tell you for example one of a very well-known uh um video game franchise is Assassin's Creed. And assassin happens throughout times in history. And we have >> not not um faithfully reproduced but very close to faithful um maps of like Renaissance Rome and that also exists in museums and that is something that can be read by machines and then that can be applied to a 3D world that can build be built more or less automatedly but also it will require manual manual input. But this is what video game companies are also now using machine learning to like create textures, right? So you don't need necessarily a full team of like 500 coders and builders and artists and engineers and you can have something at a smaller scale create that environment and then use that in educational purposes where yeah, you can have students read plays or whatever it is that you're doing and then have them jump into a VR space and understand the physicality of what they're reading. >> Yeah. Yeah, I think it's going to be so cool. >> Yeah. Again, the opportunities are so many that it kind of like >> Yeah. And where do you start? >> Yeah. >> I mean, the more I dig into it, I the more I'm like, "Oh my god, where where do you start?" >> Uhhuh. But, you know, that's my advice is just >> pick one thing that you like about it and start with that. And then it will it will take you to other things because you know, you're never going to be able to do it all. So pick the one that interests you most now and then work through it and it will open up. It's just tentacles start happening and then it starts >> reaching other things and that's the beauty of it really that that's how you know that you know all you're like oh the humanities are doomed or like the humanities are relevant like not really like cuz everything that I keep doing keeps opening more doors and people keep people keep being surprised like oh you guys in the humanities can do that like yes we can and we are doing it. Yeah. Yeah. Okay. Final bonus question that I didn't send to you. >> What's your favorite video game? >> Oh, that is difficult. That is very difficult to answer. What is my favorite video game? >> I do like a lot of envies. Indies are really fun because they're usually like less than 15 hours. Um, I do like I'm I am a sucker for a good Metroid vania kind of platformer. So, I will say this might be a weird choice, but I will say the the two AI games. There's um I forget the first one is AI and something of the forest and the second one is AI and the Will of the Wisps. And it's kind of a sidescrolling Metroid Vania. It is beautiful. It has fantastic mechanics and I appreciate when a game can have good fluidity and it just keeps letting me discover new environments and secrets. >> Yeah. >> Ori games. O R I. If you've never heard >> R I. I am definitely going to look it up the minute we get off of this. >> Fantastic. Easily. I don't know if they're my favorite because it's very hard to choose. But actually, >> you choose. It's like, do you choose the one that you spent the most time on? Do you choose the one that you come back to again and again? Do you choose the one? what makes it a hard choice because I was so I I get together with a friend of mine and and we usually talk about video games and we go for and we like to go for walks and like explore nature and uh we were looking at my Steam account and I have more than 500 games in my Steam account over the past 20 years that I've >> So how do you choose? There's just, you know, different types of your life, different stages of your life, there's a favorite, but AI, I think, is a very good thing for people to try. >> I love it. I love that recommendation. Thank you for that recommendation. All right. Well, I am going to wrap up. I thank you so very much for joining me for this episode of ADHD Talks. Um, and I did not say ADHD talks, which sometimes comes out of my mouth. >> I myself struggle with ADHD quite a bit. So, it's been so good to talk to you and I have been enjoying getting to know you and I I really uh look forward to more conversations and doing stuff together because I think we have we have we have some common interests. >> Absolutely. >> So, thank you. Thank you for your generosity and >> um and I will see you later. This episode of ADHD Talks has been brought to you from the campus of the University of Alabama. This is Sarah Whiter. Join us again next time to hear more about DH folks and their super cool projects. Goodbye.