Schmidt Sciences Gives $10 Million to 19 AI and Humanities Projects
Schmidt Sciences will give $10 million to 19 research teams that apply artificial intelligence to problems in the humanities, from a shattered marble map of Rome to ancient famine and climate records. The grants come from its Humanities and AI Virtual Institute (HAVI), and the teams will openly share code, model weights and benchmark datasets.
- Schmidt Sciences is awarding $10 million to 19 projects through its Humanities and AI Virtual Institute (HAVI).
- The awardees were chosen from 501 proposals spanning 42 countries.
- HAVI's total funding commitment over the last two years now exceeds $21 million.
- Teams will openly share code, model weights and benchmark datasets.
- Projects range from reconstructing the Forma Urbis Romae to reading cuneiform, Latin and Jibo manuscripts.
What the grants cover
According to Schmidt Sciences, the 19 projects are run by global interdisciplinary teams and aim to tackle some of the humanities' most stubborn puzzles. They also help researchers develop and test new AI capabilities. The work spans archaeology, ancient languages, the history of science, global-scale shifts in knowledge, and the making of art, music and performance over time.
An expert advisory board chose this year's awardees from 501 proposals spanning 42 countries. With this round, HAVI's total funding commitment over the last two years exceeds $21 million, the organization said.
Among the examples given are using AI to reconstruct a marble map of Rome that shattered in 226 AD, analyzing sperm whale calls for evidence of vowels and regional accents, and drawing on ancient records to understand why some societies endured climate stress while others faced famine.
Selected projects
The announcement lists the funded projects in detail. Several focus on texts and languages:
- Comparative method: a Carnegie Mellon project aims to deliver open-source software that automatically reconstructs proto-languages across the Romance, Uralic, Sinitic and Austronesian language families.
- CuneiAI: researchers at George Mason University and the University of Helsinki want to process and analyze hundreds of thousands of cuneiform tablets.
- Latinitas: the project aims to let scholars search, analyze and reconstruct fragmented Latin texts spanning three millennia, many never translated or studied.
- Jibo manuscripts: a team from the University of Washington and Tohoku University proposes to teach AI to read the Japanese script called Jibo, so scholars can study the hundreds of thousands of texts in this canon.
Others deal with objects, images and science history. The Great Marble Map of Rome project, led by Indiana University Indianapolis, will use AI to help reconstruct the Forma Urbis Romae, which survives in roughly 1,200 pieces. The team plans to extract data from a scan of each fragment, including luminescence signatures, geometry, carving directions and tool marks, and develop new AI-driven techniques to fit pieces together.
A New York University project will train AI to assemble and generate three-dimensional renderings of Neolithic objects such as ceramics, bones and eggshells, addressing the archaeological problem of "refitting" fragments. A University of Chicago project seeks to build a database of every astronomical observation in the Babylonian Astronomical Diaries, which also record weather, commodity prices and historical events over six centuries.
One project on the history of chemistry will train AI to rebuild and translate alchemical recipe collections, with wet-lab validation of particularly significant reactions. Another, co-funded with the Schmidt Sciences Climate Center, asks when and why climate stress tips into famine and civilizational crisis.
Earlier work and what comes next
Schmidt Sciences says the program has already backed projects that would have been difficult to imagine a decade ago. They include an AI reconstruction of Delacroix murals destroyed by fire in 1871, an open tool that helps small museums digitally restore deteriorating artifacts using smartphone photos, and systems that make Black digital archives overlooked by traditional collections easier to discover.
Wendy Schmidt, co-founder of Schmidt Sciences, said: "At a time when AI is reshaping how we learn and understand the world, we have a responsibility to put human knowledge at the heart of its development."
For researchers, the open-source commitment is the practical point. Code, model weights and benchmark datasets are to be shared, so other scholars in fields such as linguistics, archaeology and art history can reuse and test the tools. The source text provided here is cut off before the final project descriptions, so the full list is not covered.
