EuroLLM: a multilingual large language model that supports all 24 official EU languages
Head of Aveni Labs Dr Alexandra Birch was part of The University of Edinburgh team working with Unbabel, Instituto Superior Técnico and Université Paris-Saclay, to develop EuroLLM, which was trained on MareNostrum at the Barcelona Supercomputer.
It delivers state of the art performance for medium sized models, with far less compute due to its high quality training data and recipe. Plus, because it’s completely open source, more people will have the power to experiment, test and build from it. Some of the real-life applications and implications of EuroLLM include:
- Improved Multilingual Communication: help to break down language barriers in everyday communication. For example, it could be used in customer service to translate between customers and advisers where language barriers exist. This would be especially helpful for businesses operating in multiple countries, or call centres based internationally, ensuring better accessibility and more seamless interactions across languages.
- Enhanced education and training: Online learning or training can use EuroLLM to provide more accurate translations, explanations, and content across languages. This would make educational or training materials more accessible, enabling better cross-border collaboration and identification of training content and applications.
- Better Access to Legal, Governance and Regulatory Information: Many legal, governance and regulatory documents are available in officially recognised languages of the EU, but that presents translation difficulties. EuroLLM could help translate these documents in real-time, making them more accessible to all, especially those who are not fluent in their country’s official language.
- Support for Businesses and eCommerce: Companies operating in multiple European countries can use EuroLLM for product descriptions, customer support, and content marketing across languages. This could greatly reduce translation costs and improve customer experience by ensuring that communication is clear and culturally appropriate.
- AI Research and Development: As EuroLLM is open-source, it provides a valuable resource for researchers and developers working on AI projects. By using this multilingual model, they can test and improve their own AI systems, advancing the field of natural language processing (NLP) and machine learning. This is particularly interesting for more vertically-aligned or specific industry-based applications.
This is a hugely exciting collaborative project which Aveni is delighted to be associated with. Our foundations as a business have always been based in providing real industry-focused solutions, that are as accurate and transparent as possible. We are committed to the development of specific applications at a sector level – in our case Financial Services – as we think this is the only way to deliver the results needed.
With innovations such as EuroLLM and its open-source approach, is a further step towards the true power of efficiency, accuracy and enhanced productivity where it is really needed.Have a look for yourself: explore EuroLLM, and see what it can do.
