The Austrian Academy of Science released Apollo, the first advanced AI model for Ancient Greek, on Wednesday, trained on 600 million historical Greek words. The model is designed to assist scholars in restoring damaged papyrus fragments by filling in missing text with the most statistically likely words or passages.
Apollo is built to help academics more rapidly identify papyrus fragments relevant to their specific sub-disciplines, as well as promising new avenues of research. It is freely available through a chatbot interface, allowing scholars to focus on the implications of historical documents rather than the reconstruction process.
"Unlocking knowledge in this way was unthinkable a year ago," said Dimitris Vlitas, partner at Sail Reply. The model is trained on manuscripts, papyri, and inscriptions, and can supplement Homeric Greek or use Doric dialect when encountering inscriptions in that dialect.
The announcement follows recent advancements in AI applications for historical research, including OpenAI’s solving of a 200-year-old math problem and Google DeepMind’s genetic mutation dataset. Scholars expect Apollo to accelerate the restoration process, though it is unlikely to change broad historical understanding.
"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," said Stephen Colvin. However, the model could help uncover new details about life in antiquity and substantiate existing scholarly assumptions.
The Austrian Academy did not say how Apollo will be used beyond its current scope, and raised concerns about the risks of relying on AI for historical interpretation. "The crucial point is that human competence needs to remain," said Anna Dolganov. "If we become totally reliant on AI transcriptions and interpretations of historical material, that’s when the problems start."
Source: wired
