Apollo, the world’s first advanced large language model specifically built for Ancient Greek, is launching to help scholars decode tattered and torn historical papyrus fragments. Developed by the Austrian Academy of Science in partnership with French AI lab Mistral and technology services firm Sail Reply, the AI model is trained on roughly 600 million historical Greek words drawn from manuscripts, inscriptions, and papyri.
How Apollo Restores Ancient Greek Papyri
Academic libraries worldwide hold hundreds of thousands of Ancient Greek papyrus fragments that are too damaged to read easily. Apollo automates this heavy lifting by acting through a free chatbot interface available to academics, where it reviews tattered documents and proposes the most statistically likely words or passages.
“When it sees Homer, it supplements Homeric Greek. When it sees an inscription in Doric dialect, it uses Doric dialect,” says Anna Dolganov, a historian and papyrologist at the Austrian Academy of Science, as reported by WIRED. Sail Reply partner Dimitris Vlitas notes that unlocking historical knowledge at this scale “was unthinkable a year ago.” Stephen Colvin, a professor of classics and historical linguistics at University College London, explains that very few people possess the required depth of expertise in Greek history, meaning Apollo bakes that rare specialized knowledge directly into its interface.
Evaluating the Impact on Classical Scholarship
Scholars bogged down by painstaking reconstruction work expect Apollo to accelerate research timelines significantly, allowing them to focus on the broader historical implications of texts rather than basic decoding. Armand D’Angour, a professor of classical languages and literature at the University of Oxford—home to the world’s largest ancient papyrus collection—states that having a machine suggest viable words for a gap would speed up matters considerably.
Experts emphasize that the model will not unearth lost literary masterpieces like new plays by Sophocles. Most unrestored papyri are mundane everyday documents such as personal letters, marital contracts, and civil service papers. Instead of altering major historical timelines, Apollo aims to uncover new details about daily life in antiquity and substantiate existing scholarly assumptions by adding tiny elements of knowledge about the ancient world with every successfully restored document.
Addressing Risks and Future Applications
Relying on a probabilistic language model to fill gaps in ancient texts carries the inherent risk of introducing errors into the historical record. To mitigate this, developers built Apollo to generate a selection of word options rather than a single definitive answer, ensuring that human scholars retain ultimate control over historical reconstruction. Vlitas notes that if Apollo proves successful, the underlying technique can be applied readily to other ancient languages like Latin or Egyptian, or to any academic discipline that benefits from indexing and distilling massive text corpora.
