CAA International Recap: AI & ML in Archaeology

Technology is changing how we care for, share, and shape cultural heritage. AI is already part of how we catalogue collections, transcribe records, recommend content, restore artefacts, and interact with visitors. But it also raises deeper questions about ethics, authority, access, and care.

GLAMorous AI (GLAM = Galleries, Libraries, Archives and Museums) is a short, occasional digest for those working in culture, technology, and heritage. It brings you clear, critical, and accessible updates on:

🔍 Tools and projects using AI in creative, critical, or useful ways
📚 Reflections on data, digitisation, and preservation
🧠 Ethical provocations about memory, automation, and trust
🛠️ Practical resources for GLAM professionals
❓ One big question to take back to your team

Whether you work with metadata, exhibitions, public engagement, or digitisation pipelines, this blog is for people making heritage work in a digital world.

CAA International

CAA International (Computer Applications and Quantitative Methods in Archaeology) brings together archaeologists, technologists, and methodologists from around the globe to explore cutting-edge computational approaches. This year’s AI and ML CAA sessions, twelve in total, spanned topics from robotics, standing buildings, remote sensing to artefact analysis, emphasising how AI and machine learning are redefining archaeological research.

🔍 Other Tools & Projects in archaeology

• DeepAndes: A Self-Supervised Vision Foundation Model for Multi-Spectral Remote Sensing Imagery of the Andes
  Introduces DeepAndes, the first transformer-based vision foundation model pre-trained on three million 8-band satellite image patches, specifically for Andean archaeological remote sensing. It excels in few-shot classification, retrieval, and segmentation of landscape features with minimal annotation.
  📄 Read more here

  Advanced Deep Learning Approaches for Automated Recognition of Cuneiform Symbols
  Presents a pipeline of five deep-learning architectures (VGG16, EfficientNet, MobileNet, InceptionResNetV2, bespoke 2D CNN) trained on a custom cuneiform dataset. Two top models achieve >98 % accuracy on Hammurabi’s Law 1, automating translation from Akkadian to English and revealing linguistic links to Arabic.
  📄 Read more here

  Angkorian Reservoir Mapping
  A deep-learning pipeline applied to LiDAR and satellite data revealed dozens of lost reservoirs in the Angkorian heartland, offering new insights into Khmer hydraulic engineering.
  📄 Read more here

  Physical Monitoring of Neolithic Restorations
  Digital Applications in Archaeology & Cultural Heritage (online 24 April 2025)
  Presents an AI-driven workflow for automated detection of structural changes in restored Neolithic architecture—streamlining conservation oversight and risk assessment.
  📄 Read more here

  AI-Enhanced Protein Sequencing in Archaeological Bones
  WarpNews, 22 April 2025
  New AI methods identify 42 % more peptide fragments in archaeological bone samples, vastly improving biomolecular analyses of past diets and environments.
  📄 Read more here

📚 Reflections on Preservation & Data

These studies showcase how large, well-annotated datasets and foundation-model approaches can scale remote-sensing surveys and artefact classification—but they also underscore ongoing challenges around annotation cost, model bias, and interpretability across diverse cultural contexts.

🧠 Ethical Provocation

As we increasingly outsource “seeing” sites and categorizing objects to AI, who gets to define what counts as a valid find? How do we ensure that AI-driven narratives respect source communities and don’t reproduce colonial biases?

🛠️ Practical Resource

Tool: GeoPACHA: Geospatial Platform for Andean Culture, History and Archaeology
A web-based platform that fuses high-res satellite/airborne imagery with machine-learning-driven feature extraction to semi-automate site-detection across vast Andean landscapes. Early tests cut manual survey time by over 70 %.
📄 Project Website

❓ Question of the Issue

With more GLAM projects hinging on foundation models and large labelled datasets…
How can we build sustainable annotation pipelines that center community knowledge and safeguard against algorithmic bias?

📖 This Month’s Featured Read

Managing Artificial Intelligence in Archaeology (MAIA)
MAIA (Managing Artificial Intelligence in Archaeology) is a COST Action (CA23141) creating a pan-European network for multimodal AI tools, shared datasets, and best-practice guidelines in archaeology. A vital hub for anyone working at the intersection of AI and cultural heritage.
👉 Learn more about MAIA

💬 About Me
Hi! I’m Alfie—a researcher, writer, and archaeologist working on the digital side of heritage. I’m especially interested in the ethics of AI, open and reusable data, and what happens when modern tech meets ancient stories.

I started GLAMorous AI as a space to explore this without jargon, hype, or 100-page reports. Just thoughtful, practical insights—one post at a time.

✉️ Stay Curious
If this sounds like your thing, I’d love for you to share it or get in touch with a project or idea.
👉 Subscribe or read more at glamorousai.beehiiv.com
👉 Get in touch