The course will use digital resources and software to explore current AI applications in archaeology. Students will gain hands-on experience with established and emerging applications of AI in archaeology. Practical applications include automatic remote prospecting using drone, lidar and satellite imagery, as well as the extraction of large amounts of data for statistical analysis, such as large scale burial and artefact extraction, orientation, and analysis.
AI has a very wide set of possible use cases in archaeology, and is already significantly increasing the efficiency of repetitive and tedious tasks via automation. For instance, collecting data to create databases of features like graves, artefacts and buildings from legacy resources such as print books, can be a very time consuming task. AI applications such as AutArch can help to significantly speed up this task. Similarly, scouring large swathes of satellite imagery to find archaeological features is a common way to find new sites or monitor existing ones. AI can help to significantly reduce the amount of human input required to process this imagery. This not only speeds up the process but can also help reduce mental fatigue and thereby the error rate of the expert for this repetitive task. Similar approaches have been created to work with pictograms, ceramics, script, projectile points and stone tools.
The course will explore practical, ethical and legal issues involved in AI research and communication. It will also cover the possible future impact of AI in archaeology, both positive and negative. Regarding generative AI, it will show that ChatGPT and other Large Language Models (LLMs) are an opportunity for archaeology, as well as a challenge.
Target audience
This course will help demystify AI for students so they can confidently apply it to their own work. They will learn when to use AI, its current prospects and limitations, and see concrete examples. No prior experience in computer science, statistics or programming are needed.
Composition of the organizing team
The course is organized by two archaeologists working with quantitative data (MB, AM), a software developer (KK) and a data scientist with experience in remote sensing and GIS (AJ). We can provide practical experience in machine learning, data mining, the AI analysis of legacy data, including text and images, and the detection of archaeological sites using computer vision.
Structure of the workshop
Prior to the workshop
Around one month before the course, students will be asked to send a working paper (approx. 5 pages) for pre-circulation covering topics such as “How can AI help to detect new archaeological sites?” Or, “How can AI help in artefact analyses?” A preliminary reading list is included below. The organisers will provide feedback in writing and orally on Day 1 as a starting point for the exploration of AI.
Day 1
Welcome
In the introduction, all attendees will be asked to introduce themselves and their research interests, as well as provide some expectations they have for the course. This will serve to familiarise the group with each other and allow the lecturers to tailor the focus of their content based on the experience and expectations of the group.
Introduction to AI and oral summaries
This lecture will provide a general overview of different types of AI, covering topics like deep learning and machine learning. This lecture will also cover standard research tasks which are used in AI research to give students a broad overview of how to adapt them to their own archaeological research. Students will be asked to deliver a short two minute oral summary of their working paper.
Introduction to Machine Learning and archaeological applications
This session will provide a more detailed overview of machine learning applications commonly used in archaeology:
image based tasks such as classification / segmentation / object detection in artefact / ecofact analysis, remote sensing, etc.
tabular data classification in predictive modelling, artefact classification, etc.
Introduction to generative AI and its applications
Generative AI has gained widespread attention due to the mainstream popularity of several LLMs (large language models) such as ChatGPT and Gemini. Their ability to generate text provides some obvious use cases. Despite their seemingly powerful abilities and confident tone, they still suffer from AI hallucinations and lack specialised details for more archaeology-specific use cases. Besides generating text and images based on prompts, generative AI can not only reconstruct artefacts like script or ceramics, but can also perform tasks like predicting the morphology of lithic flakes removed from cores. This lecture provides the attendees with a broad overview of the general concepts behind generative AI to facilitate further discussions about reliability, explainability and ethics.
Ethical and legal issues
The usage of AI in archaeology presents both ethical and legal challenges. Ethically, there are concerns about the potential loss of human expertise, as reliance on AI might devalue traditional archaeological skills and insights. Additionally, AI interpretations of data could reflect biases in the datasets it was trained on, leading to skewed conclusions. Legally, issues arise around the ownership and copyright of training datasets to train AI. The course will give a brief overview of the legal framework for data mining scientific publications in the EU and beyond.
Presentations and respective discussions, short paper
Attendees will select a publication from a list related to the usage of AI in archaeology and give a short five minute presentation summarising how AI was used to achieve an archaeological task. Students are encouraged to identify the training and test datasets, the AI methods used, and any benefits and/or shortcomings of the use of AI in their example.
Day 2
Introduction to AI Image Processing and archaeological applications
This lecture will cover the usage and best practices of deep learning to perform classification, object detection for archaeological use cases. It prepares the students for the practical sessions using these methods to detect, analyze and classify objects from archaeological drawings and also to detect sites on satellite imagery.
AI for archaeological illustration
Students will use generative AI to create archaeological illustrations based on prompts. For instance, they will generate photorealistic faces of prehistoric individuals, based on the phenotypic predictions published in recent ancient DNA studies. The prospects and limitations of this approach will be discussed.
AI for archaeological prospection, remote sensing and other image processing tasks
In this interactive session students will learn a practical prospecting method using commonly used GIS software to efficiently analyze large amounts of satellite or drone images. By annotating archaeologically relevant sites on a small piece of the whole dataset, AI models that are integrated in a very user-friendly way can be trained to detect features on larger datasets.
AI for document data extraction (AutArch)
In this hands-on seminar, students have the opportunity to use AutArch to extract data from digital documents. AutArch integrates several types of AI into an existing and ready to use application. By the end of the seminar, students will have practical experience in using AI assisted tools to extract data and present a possible AI integration into archaeological research.
Day 3
Presentations
Students will give a short five minute presentation about their work and how AI could help to enhance it. Feedback will be delivered by the organisers, who will try with the students to identify a relevant AI solution.
Future applications and integration to archaeological workflows (pointer to course 2)
The future of AI in archaeology has large potential to revolutionise the field. As AI tools become more sophisticated, they will likely become integral to archaeological research and complement human expertise. On one side existing advances into AI research have to be integrated into archaeology, but also new advances in AI research will continually have to be adopted. This lecture will give an outlook into the possible future.
Final discussion and concluding remarks
Organizers: The Nordic Graduate School in Archaeology – Dialogues with the Past, Antoine Muller (UiB), Maxime Brami (JGU Mainz), Kevin Klein (software development, Mainz), Amina Jambajantsan (MPI Jena)