 So, what we now heard was about AI in general, we talked about AI in the economy and the problems that may arise. But what we do here at the ReED Co-op is use AI for a very specific purpose, we're using it to unlock the history. So the question that we're trying to discuss now is, is AI the key to history? And I think we have to switch on the microphones and then we can start with the discussion. Okay, I hope mine's working. It's a bit in the quiet side, is it? Yeah, now it's better. Thanks a lot. Yes, yeah, welcome to this panel discussion. We also wanted to have a somewhat interactive part here during the first day and the plenary session here. And we thought it would be really nice to talk to the experts in their fields, both from application of AI in the digital humanities or social sciences and or social sciences, the experts on the technology itself, and also someone who is sort of a little bit in between those areas. So I have the great pleasure to present the panelists. The first panelist to my right here is Dr. Annemiek Romijn. She's an early modernist, so historian, and works on legal sources and in the area of the digital humanities. She works at the Hauchens Institute in Amsterdam and Freie Universität Amsterdam and frequently provides training workshops for transcribers. You might know her from there. And she's a long standing and often skeptical friend of transcribers. And we really appreciate the skepticism and the candor, obviously. And this is also why you're on this panel here today. Then at the far end of the panel we have Dominique Delandre. She's a full professor of history at the history department of the University of Montreal. Has taught European and North American early modern history since 1992. So also a very experienced member of the scientific community. And she works on material from judiciary and civil archives and researches, especially the women and their men who made Montreal's history, especially in the 17th century. So someone from the area of applying AI in the field as well. Then we have Professor Patrick Glauner, who Matthias already introduced. Yeah, I'll just go over this again quickly. He's a full professor of artificial intelligence at the Degendorf Institute of Technology in Bavaria, Germany. And besides academia, he's founder and CEO of his company Skyrocket AI, a consulting firm. He's an expert witness, as we've seen to many governments and institutions. And yeah, an expert in the field of the actual technology, we could say. And then last but definitely not least, Dr. Günther Müllberger, the chairman of the board of our cooperative. He's a Germanist by training. For more than 20 years now, he has advanced digitization and digitalization agendas here at the University of Innsbruck. And he has coordinated the REIT project, during which a large part of the development of transcribers happened. And he also initiated the founding of the REIT cooperative and now is chairman of the board. So yeah, digital humanities or humanities background and a very strong focus on digitization and digitalization. So we have a very nice spectrum of both application and expertise in the field of AI. And who could unfortunately not be here is Melissa Terrace, our third board member. I'll say a couple of words about Melissa as well. She is a professor of digital cultural heritage at the University of Edinburgh. There at the College of Arts, Humanities and Social Sciences. And her research interests are digitization of cultural heritage, technology or technological procedures and their impact. And yeah, she was really crushed not to be able to be here and we really wish her well. So let's start discussing today's question, is AI the key to history? So this is a question that maybe doesn't come up that often during AI discussion. So we mainly hear about the medical sciences, for example, or autonomous cars, robotics, yeah, anything you like. But history is not that much in the limelight when it comes to artificial intelligence. So I've prepared a couple of questions. The first one I'd like to ask is, yeah, we keep hearing more and more that AI can do things maybe better than humans, especially during Professor Glauner's talk. We've seen what AI can really do today. And my question is, can it really do a better job and do you think it can do in non-exact sciences, such as the digital humanities or the humanities in general, to which also history belongs? Professor Glauner, what do you think? Well, thank you very much for the question. Yes, AI can do better than humans, but we always need a good ground truth. And I think that's a challenge in every science. So you need a good database. And if the database is wrong or biased, finally, you will also make wrong predictions. I see a lot of potential in the humanities and, for example, in history. Like you talked today about search and all of those advanced search features you wish to have, AI will be key to that. So we have this exponential amount or exponentially growing amount of data around us. And the only way we can keep track of it is search. And there's so much historic data out there that is not digitized yet. And we don't have enough experts to read these old characters and to type them into a computer. So an AI is certainly very, very helpful to do this. But an AI can also help, for example, to restore images or text that is partially broken. Maybe part of a page got cut off or an image got partially damaged. And maybe the AI can help us to complete that information. There's obviously always some uncertainty included. Humans make mistakes and so does an AI. But the AI can allow us to do all of this much faster, much better potentially. And humans can finally still take a final look at it and try to improve it. And what do you think about the room of interpretation that often exists in the historical sciences? That's one of the problems, I think, also why I'm asking these questions. Sometimes it's difficult for humans to decide, is this a correct representation of the historical facts? Can this really be reduced to solid ground truth, do you think? So maybe the AI could propose different scenarios and could also give an explanation why the AI has come to that conclusion. And then maybe experts can look at that explanation and wonder, does it make sense? Okay. Yeah, because I think one problem that AI has maybe or that it doesn't do such a good job of yet is explaining things to humans. So it's quite good at providing answers, I think, and speeding up decision making but really getting a point across. That's maybe still a bit in the future. Maybe let's move on to the experts on this actual field, history. Do you think AI could be a historian and could it be a better historian than a human? Well here I'm very skeptical but that's why you asked me. So yes, artificial intelligence can help us understand the data but it can't do interpretations because we as historians can't provide ground truth for the interpretation. We can say someone was born then and someone died then, that's quite obvious. But we can't have the computer interpret what an impact someone has. That's something that's, well, you have a humanistic circle for that and by reading a lot and sure the computer can deal with a lot of information but it can't tell for now how much of that information is relevant and if you want to be able to propose a new vision basically ground truth would train the computer to come up with traditional visions because that's what it would hypothetically be trained with. But if you were to come up with an alternative view that wouldn't then, that would not count as ground truth but it would still be an interpretation of history and then I'm not talking about extremist views or Holocaust deniers or these kind of things but if I look at my own research to 17th century political structures in Germany and France I came up with a different interpretation than was commonly used and that would not be something the computer would come up with simply because it's not part of the curriculum. So with the humanistic circle you keep thinking about what is going on, you keep reinterpreting, you might come up with comparisons to different areas and you as a human can make connections that you can't feed the computer yet as being ground truth and if you do then you would have so many diverse ideas that the computer couldn't make sense out of it because well what is true then? Maybe let's try to get some other answers from the spectrum. Maybe Professor Gleuner again, do you really think the AI cannot do that? For example let me use a very simple example here, maybe one that everyone can understand. Why did Napoleon invade Russia? Are the common explanations that are given the true ones or is it just a matter of not having enough data that the interpretations are the ones that they are currently? What do you think, how would you respond to this view? Well computers and AI obviously give us the opportunity to gather a lot of information and analyze it, that's something a human could not do in his or her lifetime and that's obviously a huge opportunity. Yes the AI is trained on certain data, nonetheless there is some creativity now in these models and the AI could also possibly check if things we see as their ground truth if they are actually conclusive by adding more information. So I see that as a huge opportunity, but yes the AI will not ultimately decide about history. We should never let an AI to decide ultimately such very serious things, same in healthcare we don't want an AI to ultimately decide if I'm sick or not. It should always provide advice to experts and explanations and they will make the final discussion. That's a very interesting view I think because I think one thing that is also very discussed very hotly at times is what should we do with AI and what should AI do, not just what can it do and this is also maybe where China is a bit ahead of us with the decision what it should do and they let it do more we could say. So in the western tradition we are maybe thinking a bit more about what it should do and that connection maybe let's hear from another expert from the historical domain. Dominique, what is your working relationship with AI? How would you say you collaborate with AI because we've just heard from AI can be a tool or should be a tool? What do you think, should it be a tool or should it also be part of the conversation? So should it get its own voice or should it really be just a tool? Is it switched on? I feel a little bit like an impostor to be among you because I'm a user friendly of Transribus and by the fact that I just arrived in front of very difficult handwriting to this effort and I realized that Transribus could help me to teach this paleography very hard and even for me to be able to recognize this handwriting and everything. So I'm not sure that AI is the key to interpretation of history but AI is a tool and it serves a real revolution that is going on in the treatment of the archives and that's why we need to think about this revolution and all the little details that you were talking about. So yes, it's changing profoundly the way we are reconstructing the past. Of course, I don't need to spend five years of my life writing my thesis about ten years of the political thought of Napoleon just before he invaded Russia because the amount of paper to read is huge and to find something new also. So now I can go on 300 years of manuscript and find... It's no more a sample, how would we say that? We're not working with samples, we're working exhaustively with the search and with our knowledge of historians, we'll always need paleographers, historians to be able to interpret what the data are offering us. What I find fantastic is also the sharing of collection, to be able to vanquish the aloneness of our work and to be able to be ubiquitous, to be in Paris or Rome archives and being in my own office and this has changed completely the work that we are doing and especially this collaborative way of doing things for example, you could come in my collection and just... One function very important in Transcubus is the unclear. So you can check the unclear and you can add and like a wiki thing, you can work together so this is very, very important. So if I understand this correctly, you see it more as a tool. It's a tool but it's a techno-social tool. Okay, so it combines the social potential of the members of the scientific community, is what you're saying. Good, let's maybe move a little bit between the poles of the spectrum that we have here today. Günther, a question for you that is a bit in the vein of what we've just heard. You have managed digitization and digitalization group for more than 20 years now and for about half that time now AI has played an important role for this. How realistic do you think that in the future the credo in the humanities will be to play on a theme by President John F. Kennedy that the credo will be ask not what AI can do for you but what you can do for AI and I know this is a question that's very interesting for you because your answer is maybe not that clear-cut as what we've heard so far. My spontaneous answer would be... Use the microphone please. Excuse me. My spontaneous answer would be I have done enough for the AI but it pays back so that's definitely. I think for me it's as you said you need to solve a problem and that's the mission and I just can say the mission to solve the problem of text recognition that's 20 years old when I discovered it in my PhD thesis. I wanted not to read 4,000 letters of Johann Wolfgang of Goethe. I just wanted to search for a Nazism. That's it. And this was really to solve the problem of searching I'm sure that the AI will play a big role there but we have to understand that there is a problem with searching and that it is not enough to say well yes you can search. Of course it's an improvement but we want to have more and we need this kind of vision to say what do we want and in the ideal case I would expect that AI can do a lot. It can be a tool but it can also be a dialogue partner. I'm sure if I don't read too many books I have to confess but I remember very well reading a real standard book about the biography of Johann Wolfgang of Goethe. And it had I don't know 1,000 footnotes and my thinking was if all the footnotes are digitized you would get a kind of network with the research literature and connect this with the sources of what they are talking about. I mean that's a kind of knowledge I never could reach in my life but the computer can do and the AI can do and then to get some suggestions as we saw the nice image with the chairs from you. I had a closer look to some of the chairs and I thought oh many of them are really ugly. They look nice at the beginning if you look at it but if you look from the view would you buy it? Well I'm not sure. But there are also probably 1 out of 10 or 1 out of 50 which is great which is a new idea which is a nice image and the same would happen with a personal research assistant or whatever you would call it. It would give you new ideas and you would say it makes sense or not. So you would see a somewhat more independent role for AI in the conversation as opposed to just being a better search solution? Yeah I think there are no limits. There are no limits and it must be clear that there is no difference actually between the learning of machines and our learning. It's just a different physical basis but the principles are very similar. And yeah, continuing this line of thought if we allow AI to become part of the conversation I think there may be new dangers arising because I think when you let AI be a part of the conversation about history there are some things that might be problematic. For example, think of deepfakes. That's AI too, so AI can write history as well as interpreted already. Or yeah, history could be a source of knowledge to learn about humanity and to defeat humanity. What do you think about these dangers, Professor Glauner, how far away are you from a terminator scenario that Elon Musk often evokes? Well I think the fear of an out of control AI is exaggerated. There's a lot of sci-fi around it and I also like Star Trek and Terminator. I know all of that is really cool. But a very famous AI researcher named Andrew Eng he was the founder of Coursera and started all these online courses about AI. He was a professor at Stanford. He was asked to assess how far we are in AI and if that scenario is anywhere near. And I think he gave a really good explanation so I want to tell you that. He said, well in AI we have made progress and we have some next steps we want to do and he compares this to space. So there are researchers and NASA thinking about sending humans to the Mars at some point in the next few decades. There may be some researchers already thinking about how to colonize Mars in the distant future. So in AI we also have some plans what we could do sometime in the future. But he said he cannot really prevent AI from turning evil because we are too far away. Like in space no one is so far thinking about how to combat or handle overpopulation on Mars because it's just too far away. We first need to make a lot of steps there and that's what Andrew said that he cannot productively avoid it. He's working on AI, shipping code, adding value. Obviously those scenarios at some point they may become relevant but we're just too far away. Okay, thank you. That's a bit reassuring, I'm not sure I believe it but let's see what the next five years even will bring. Yeah, maybe let's have some questions from the audience if there are any. Lucia, is there someone who could lend Lucia a microphone? Yeah, you'll get the microphone in a second. I really... Is it switched on? I really don't like the microphone but anyway, Günther, allow me to disagree with you about you say we are nearly close to the end of the technology for tax recognition because I agree with, may I call you Patrick? Or because... Glauna. I am more on his opinion and that's what I would like to ask. Can the tax recognition also help us with lost information and then here I'm thinking that's a problem I'm dealing with is a document that the time or the whatever I ate some words, documents that survive poorly the teeth of time or like the Dutch would say the van tijt and so some documents are pretty much damaged but we could read still so now they are completely destroyed so for example when you miss words because some bug ate and then we train enough data so that we can fill in the gaps that's my question because then the information is lost or do I need to express myself in another way? Do I make any sense? You understood the gist of the question? Günther? We all have seen the example from Google where they added some text to manuscripts from Israel so the idea is rather clear to predict what could be there or maybe if it's a very faint and hard to read thing to also somehow predict what could. Yeah, I think it's project Ithaca that you're talking about. Yes, I'm sure there are... I didn't want to say that there are no improvements but if you look at some cycles you see that sometimes you have really steep progress and then it's becoming a bit lower. Okay, let's maybe have one more question from the audience if there is any if you would like to ask the experts they're here for you now but you'll probably get the chance later on as well so if there are no more questions from the audience then we can conclude with a very short round of the one question that we came here to discuss and an answer by each of you maybe try to limit it to two sentences I'll be counting. Dominique, is AI the key to history? AI is not replacing the historian but the human is at the basis of AI. Okay, Professor Gerne. Well, I absolutely agree AI is not about replacing people it's about supporting people make sure we have a better life and we don't want to be out of business eventually and there are huge potentials of AI that are yet to be unlocked. Günther? Yeah, I'm sure that the journey just began. Okay, Jordan Sweet. That was one sentence you would have one to go if you'd like, okay? Great, Annemiekke? Well, I certainly believe that it can be supportive of any field and let's keep the people in the loop. Okay, great. And what it is sure to be helping us in the future is unlocking history and that's what the motto of our cooperative is so let's unlock history together. Good, thank you, dear panelists and I'll hand back to Matthias. Thank you very much.