Science and Technology

Técnico researchers are part of the team that will create the Portuguese large language model AMALIA

The artificial intelligence system announced by the Prime Minister at the opening of the Web Summit should have its first version ready by the end of the first quarter of 2025.

Researchers from Instituto Superior Técnico are part of the team that will develop AMALIA, the Portuguese large language model (LLM) that was announced by Prime Minister Luís Montenegro on 11 November during the Web Summit in Lisbon. This large-scale language model designed in Portugal will be trained using data in the Portuguese language on national and European supercomputers.

An LLM is a model that uses artificial intelligence to process, understand, and generate text in natural language. It can be used as a component in various systems, such as dialogue systems and chatbots, search systems, and automatic question-answering systems.

AMALIA (Automatic Multimodal Language AI Assistant) will start from a model with around 9 billion parameters (pre-trained on 4 billion words) and fine-tuned on Portuguese data extracted and filtered from Arquivo.PT.

The project will take 18 months to complete and the first version of the model is expected to be ready by the end of the first quarter of 2025.

At the opening of the Web Summit, the Prime Minister listed future applications of this LLM in education (with ‘an artificial intelligence educational tutor’ for each student), access to public administration services (‘simpler, more direct and more personalised’), and business growth (companies will be able to ‘design their services in an era of artificial intelligence also in Portuguese’).

The team responsible for developing AMALIA includes members of Instituto de Telecomunicações, a research centre affiliated with Técnico, Unbabel, a member of the IST spin-off community, NOVA University Lisbon and Fundação para a Ciência e a Tecnologia.

In the media: