The Austrian Academy of Science, Mistral, and Sail Reply are releasing Apollo, a large language model trained on approximately 600 million words of ancient Greek text. The free tool aims to help researchers reconstruct damaged historical documents.
Apollo represents a specialized approach to AI development, focusing on a single historical language rather than modern multilingual datasets. The model was trained on ~600M ancient Greek words, enabling it to fill gaps in fragmentary papyrus texts and assist scholars in recovering lost historical details.
The collaboration brings together institutional expertise from the Austrian Academy of Science, computational resources from Mistral (an open-source AI company), and development support from Sail Reply. By making the model freely available, the partners intend to democratize access to advanced language processing tools for classical studies researchers worldwide.
The project highlights growing interest in applying machine learning to humanities research. Similar specialized models have emerged for Latin and other historical languages, suggesting a broader trend toward AI tools tailored for specific academic domains rather than general-purpose applications.
As animal behaviorists deploy AI to interpret animal communication, researchers are wrestling with whether these technologies benefit the animals themselves or merely satisfy human curiosity.
UK Prime Minister Andy Burnham plans to leverage the country's upcoming G20 presidency to broker an international AI governance agreement before the UN General Assembly.
Intrinsic, Alphabet's robotics software company, has open-sourced Intrinsic Core under the Apache 2.0 license. The move provides developers with foundational tools for building physical AI systems.
Anthropic CEO Dario Amodei will address the UN Security Council on Wednesday to discuss artificial intelligence, appearing alongside OpenAI CEO Sam Altman at a dedicated meeting on AI's future.