Virginia Dignum, AI professor at Umeå University, believes that developing large language models (LLMs) in specific languages, like Portuguese, allows AI systems to reflect cultural context. Dignum, part of the UN’s High-Level Body on AI, emphasizes that a Portuguese LLM will preserve the country’s cultural identity, unlike models like OpenAI’s ChatGPT, which are primarily based on English data.
The Portuguese LLM, named Amália, will debut in the first quarter of 2025, with a final version set for 2026. Paulo Dimas, CEO of the Center for Responsible AI, highlights that Amália’s core features are linguistic accuracy, cultural representation, and data protection.
This project is vital for preserving the Portuguese language in AI and ensuring control over applications developed in the language. The initiative is supported by research centers like Nova FCT and Instituto Superior Técnico.
Related News