Austrian Academy releases Apollo, an AI model for Ancient Greek texts

In short: The Austrian Academy of Sciences is releasing Apollo, a specialized AI model designed to help academics restore and search Ancient Greek texts.

What happened

Academic libraries hold hundreds of thousands of Ancient Greek papyrus fragments, many of them torn, faded, or incomplete. Restoring them is slow work, since scholars often need to figure out where one word ends and the next begins, then guess missing parts based on context.

On Wednesday, the Austrian Academy of Sciences will release Apollo, which it describes as the first advanced large language model for Ancient Greek. A large language model is an AI system trained on lots of text so it can predict what words are likely to come next, like a very powerful autocomplete. Apollo was developed with French AI lab Mistral and tech services firm Sail Reply, and trained on about 600 million historical Greek words from manuscripts, papyri, and inscriptions.

Apollo will be offered to academics through a chatbot. It is meant to help researchers find fragments related to their area of study and to suggest possible words or passages where a document has gaps. Researchers say it can also adjust its suggestions based on the style of Greek it is seeing, such as Homeric Greek or a regional dialect.

Why it matters

This could speed up a task that currently depends on a small number of highly skilled specialists. It may not uncover famous lost works, since many papyri are everyday documents like letters and contracts, but it could add small, useful details about daily life in the ancient world. Scholars also warn about mistakes, so Apollo is designed to suggest multiple options, with humans making the final call.

Source: Wired