Leipzig AI and Harvard DH

Leipzig Artificial Intelligence Summer School

Last week (June 25-28), I attended the International ScaDS.AI Summer School in Leipzig, Germany with graduate students, post-doctoral researchers, and practitioners to discuss topics in the field of artificial intelligence and big data. Although I’m sure I was the only humanist in the room, it was an enriching experience to be surrounded by a diverse group of passionate individuals in various aspects of STEM. The summer school delved into cutting-edge topics in areas such as machine learning (ML), large language models (LLMs), and neural networks. The series of talks alternated from theoretical knowledge (e.g. mathematical foundations) to practical/real-world application (e.g. earth and environmental sciences).

Here’s a list of the talks over the four-day summer school (also here):

• Martin Potthast, “Evaluating Generative Retrieval Systems”
• Rainer Mühlhoff, “The Risk of Secondary Use: Challenges in the Ethics and Governance of AI”
• Robert Haase, “Large Language Models for Quantitative Bio-Image Analysis”
• Patrick Ebel, “AI-Assisted UI Design – Using Computational Models of Human Behavior to Create Better Interfaces”
• Pavlo Bazilinskyy, “AI-Based Traffic Safety Research with Public Datasets”
• Lily-belle Sweet, “Advancing Agricultural modelling with machine learning through interdisciplinary coordination”
• Josefine Umlauft, “Leveraging Earth System Data Cubes for Enhanced Environmental Insights through Machine Learning”
• Jonathan Bedford, “Applying Machine Learning to Geodetic and Scaled Fault Zone Laboratory Data to Understand Earthquake Processes”
• Christopher William Johnson, “Applications of AI to Earthquake Physics: Learning Fault Slip and Precursors to failure to Advance Earthquake Predication”
• Christoph Lehmann, “Data, Assumptions, Models and Uncertainties: A Statistical View on Responsible AI”
• Guido Montúfar, “Biases of Gradient Optimization in Neural Networks”
• Andrea Agazzi, “Correcting SGD for Scalable Bayesian Inference”
• Paweł Dłotko, “Invitation to Methods of Topological Data Analysis”
• Rudolf Seising, “Fashions of Artificial Intelligence” {my favorite talk, essentially a history of the approaches to AI from ENIAC in 1946 to the present day}
• Søren Hauberg, “Reparametrization Invariance, Identifiability and Representation Learning”
• Sayan Mukherjee, “A Sheaf-Theoretic Construction of Shape Space”
• Ingo Steinwart, “Density-Based Cluster Analysis”
• Mariam Hassib, “Exploring Human-AI collaboration: Decision Making, Reliance, and Trust”
• Daniel Buschek, “Human-AI Interaction in Writing Tools”
• Arthur Fleig, “Applied Mathematics in Human-Computer Interaction”
• Birte Platow, Hermann Diebel-Fischer, and Roderich Barth, “Panel Discussion”
• João Belo, “Optimization Methods for Interface Design and Adaptation”
• Adrian Lindenmeyer and Daniel Schneider, “Utility of Knowledge Uncertainty Estimation in Human-Computer Collaborative Decision-Making”

And here’s a few of my reflections:

1. Housed at the Leipzig Zoo, the venue (see here) was exquisite, especially the non-stop catering. Of course, STEM fields have much more money than the humanities, but it would be nice if in the future if we could see similar resources and support extended to the humanities. The experiences and opportunities provided by well-funded programs can greatly enhance learning and collaboration across all fields of study. Imagine the advancements we could make if the same level of investment were given to the humanities!
2. Hearing presentations on topics that I sometimes quite literally knew nothing about, I was exposed to a plethora of new material, concepts, programs, and research areas that has much potential for advancing humanistic studies. Learning about advanced topics such as machine learning algorithms, data visualization techniques, and data frameworks significantly expanded my understanding and appreciation for what these digital tools will be able to do in the future. This exposure, furthermore, has opened up exciting new avenues for future research in my own academic trajectory, including ideas for digital humanities projects, which might eventually allow me to explore innovative solutions to complex problems in the application of AI to humanistic research.
3. A fascinating trajectory of generative artificial intelligence is going to be improvements in transformer-based deep neural networks that allow AI-algorithms to undergo a sort of self-reflection and self-correction that will enable it to produce truly new research insights on its own accord. Stated differently: at the present, generative AI can, by definition, take learned patterns and structures of input and, when given a human prompt, apply similar characteristics of the patterns to new questions. The thought is that eventually no human input/guidance (except for the data set) will be needed for AI to conduct and produce original research. Given a data set (e.g. a database of Ugaritic texts) AI will be able to do self-directed original research and produce associated content (e.g. article-length papers, pedagogical videos, etc.), although human prompts are still possible. Of course, concerns have already been raised about the misuse of generative AI, such as deepfakes for the sake of deception and manipulation as well as the classic fear/critique of automation resulting in job loss.
4. Another trajectory of AI is in the area of spatial computing, specifically eye-tracking and object detection technologies. Eye-tracking technology analyzes where and how long a person looks at certain areas, which has applications in user experience research and psychology; object detection, on the other hand, allows AI systems to recognize and locate objects in an image, video, or real life. Taken together and applied to the humanities, one can imagine a world in which the user, wearing augmented reality glasses, can seamlessly interact with digital information overlaid on historical artifacts or texts in a museum or classroom, enhancing both educational experiences and research capabilities.

The obvious question for the reader of this blog is “Why?” Why would I, a Semitic philologist studying ancient Near Eastern languages and texts, attend a conference on artificial intelligence and big data? In addition to point two above on exposure to new concepts and ways of thinking, the digital humanities is the future; artificial intelligence offers advanced tools for analyzing large corpora of texts, detecting patterns at a quantitative/statistical level, and can uncover insights that are impossible to achieve manually. Additionally, new technologies can guide field archaeologists where to dig, facilitate digital preservation of texts and objects from the ancient world, and create interactive and educational experiences that makes our work more accessible to a broader audience.

Harvard Digital Humanities Certification

In addition to the Leipzig summer school, I also recently completed a certification course on the Digital Humanities (DH) offered by Harvard University faculty and students through the edX platform (see all certificates here). This course orients and trains the student in a variety of software tools and techniques that allow humanists to ask new questions about their data. There are three major objectives of the course: (1) provide an orientation to the DH and what that means in different disciplines; (2) introduce a variety of DH projects and tools; and (3) teach skills related to command line functions and text analysis. While this course was only a first step into the exciting field of the DH, it will provide a firm foundation as I continue to find new ways of integrating computational and digital methodologies into my research with the goal of leveraging cutting-edge technologies for a more sophisticated analysis of ancient texts, ultimately enhancing the scope and depth of my scholarly work.