Hundreds of thousands of Greek fragments await restoration

Academic libraries worldwide hold vast collections of damaged Ancient Greek papyri. Scholars can restore many by filling in missing words, but the work is slow and requires rare expertise.

Meet Apollo, the ancient Greek AI

The Austrian Academy of Science is releasing Apollo on Wednesday, trained on 600 million historical Greek words from manuscripts, papyri, and inscriptions. It's the first advanced language model built specifically for Ancient Greek.

Apollo fills gaps with statistical precision

The model is trained to fill in missing words in damaged papyri by predicting the most statistically likely text. It understands different Greek dialects and literary styles, from Homer to inscriptions.

“When it sees Homer, it supplements Homeric Greek. When it sees an inscription in Doric dialect, it uses Doric dialect.”

Scholars can focus on meaning, not reconstruction

Restoring papyri traditionally required identifying word divisions, dating documents, and consulting reference materials. Apollo automates this, letting academics spend time on interpretation instead.

“If I had a machine telling me, 'Here are the three possible words that could fit into that gap,' it would speed up matters considerably.”

Expect incremental gains, not major discoveries

Most unrestored papyri are mundane documents like letters and contracts, not lost plays. Apollo will likely uncover small details about daily life rather than revolutionary historical finds.

“If you were a layperson, you might think suddenly we'll get a few new plays by Sophocles, but that's not going to happen.”

Human judgment remains the final check

Apollo proposes multiple word options for scholars to choose between, not final answers. This keeps human expertise central and prevents errors from corrupting the historical record.

“The crucial point is that human competence needs to remain.”