Latam-GPT Latin America AI model for sovereignty
A new era of technological independence is dawning with the Latam-GPT Latin America AI model, an initiative led by Chile’s National Center for Artificial Intelligence (CENIA). This project aims to create a free, open-source, and collaborative large language model specifically trained on the region’s unique languages, history, and cultural contexts.
A Collaborative and Culturally Aware AI
Unlike global competitors, the project’s spirit is not competition but collaboration. Álvaro Soto, director of CENIA, emphasizes that the goal is a model aware of Latin American cultural requirements and challenges. Therefore, it seeks to understand different dialects and regional nuances that large commercial models might overlook. The project already involves 33 strategic partnerships with institutions across the region.
The foundation of this initiative is a massive dataset of over eight terabytes of text, equivalent to millions of books. This information, gathered from 20 Latin American countries and Spain, powers a language model with 50 billion parameters, putting its scale on par with models like GPT-3.5. This allows the Latam-GPT Latin America AI model to perform complex tasks with a deep regional understanding.
Infrastructure and Future Impact
Developing such a powerful tool requires immense computing power. For this purpose, the project relies on the supercomputing infrastructure at the University of Tarapacá in Chile, equipped with state-of-the-art Nvidia H200 GPUs. According to experts at CENIA, this infrastructure is fundamental for training the model and ensuring regional technological sovereignty.
The first version of the Latam-GPT Latin America AI model is expected to launch this year. However, this is just the beginning. The open-source nature of the model will allow organizations across the continent to adapt it for specific sectors, such as education, healthcare, agriculture, and culture, fostering a new ecosystem of localized AI applications and preparing new generations for a technologically advanced future.