At leading universities abroad, artificial intelligence is now helping to read two-thousand-year-old scrolls and to bring the teaching of philosophy and history to life. We present a selection of examples from research and teaching; including tools you can try yourself, and mentions of what doesn't work yet.
What is happening with AI here at our faculty was on display at the January conference, and further projects have been running since July – we will return to them in our news later this year. This time, however, we look beyond our borders: what exactly are foreign universities doing, and how might it inspire your own teaching or research?
At Princeton, Barbara Graziosi and Johannes Haubold lead the Logion project – a language model trained on premodern Greek. It uncovers scribal errors accumulated over centuries of copying and suggests reconstructions of passages gnawed away by mice, damaged by moisture, or obscured by stains. The model does not offer the "correct" solution but several possibilities, thereby setting the scholar's thinking in motion
This project by the British Library and the Alan Turing Institute (2018–2023) analysed enormous volumes of newspapers, books, and maps, asking how the Industrial Revolution affected the lives of ordinary people. It produced open tools such as MapReader (computer vision for historical maps) and T-Res (recognition of place names in newspapers). An explicit part of the project was also the study of bias in historical data – that is, who is represented in digitised collections and who is not.
Elizabeth Margulis (Princeton) uses language models to analyse thousands of free-form descriptions of what music evokes in people. The same excerpt from Webern evoked a sense of impending doom in English-speaking listeners from the American Midwest, while speakers of the Dong language from China's Guizhou province imagined playing joyfully outside with friends.
The scrolls from Herculaneum, carbonised in the eruption of Vesuvius in 79 CE, cannot be physically unrolled – every attempt destroyed them. On 25 June 2026, a breakthrough was announced: scroll PHerc. 1667 became the first to be virtually unrolled and read from beginning to end, including the discovery of authors previously unknown to history. The path there led through X-ray microtomography at the synchrotron in Grenoble (tens of terabytes of data per scroll) and through machine learning that recognises carbon ink, which is almost indistinguishable from the papyrus in the scans. Notably, archivists had once assigned this scroll a legibility score of zero – it was essentially written off.
At Arizona State University, students in a philosophy of science course train their argumentation skills with AI. The student takes on the role of Karl Popper, the AI plays his opponent Rudolf Carnap, and the scene is set in the 1930s on the steps of the hall where the Vienna Circle is about to meet. Another point of interest is a new philosophy programme focused on AI, consciousness, and ethics, planned to open in autumn 2027.
At Yale, a student created LuFlot, a chatbot built on the complete published work of Luciano Floridi, with which students can converse. The chatbot also links back to the texts, so users can trace every claim to the original.
Historian Benjamin Breen (UC Santa Cruz) had his students live through a day in the life of a person at the height of the plague epidemic of 1348. It is not a finished application but a set of prompts you copy into a chatbot, which then launch a simulation in the form of a conversation. The students then wrote a paper on where the simulation got things wrong.
Three scenarios to try (the link opens a document with the prompt; copy it into a chatbot):
Breen also offers other settings – the fall of the Ming dynasty, the voyage of the Beagle, the Nixon–Khrushchev kitchen debate, or a simulation of your own hometown. You can also try these other intriguing simulations:
Has one of these projects caught your attention, and would you like to try something similar in your course or research? We are happy to help!
From August 2, 2026, the transparency rules set out in Article 50 of the European Artificial Intelligence Regulation will apply throughout the European Union. Here is a summary of what has changed and how this affects the faculty.
Google Gemini has launched this year’s free plan for students.
