A coalition of twelve Latin American nations is collaborating to launch Latam-GPT, the first major AI language model specifically trained to understand the region’s linguistic diversity and cultural richness. Led by Chile’s National Center for Artificial Intelligence, the open-source initiative aims to broaden access to AI technologies across Latin America.
Chilean Science Minister Aisén Etcheverry noted that the development of Latam-GPT could enhance the accuracy and performance of global models that are predominantly trained in English.
The project, which began in January 2023, is designed to serve as a foundational technology for building chatbots and safeguarding Indigenous languages. A translator has already been developed for Rapa Nui, the native language of Easter Island.
Beyond powering chatbots and other applications, Latam-GPT is being developed as a tool to preserve Indigenous languages. The initiative plans to extend this capability to more Indigenous tongues, supporting customized education programs and virtual public service assistants, among other uses.
Built on Llama 3 technology, the model is being trained on a regional computing network that includes cloud-based systems and infrastructure at Chile’s University of Tarapacá. According to Reuters, the project has received support from Amazon Web Services and CAF, a regional development bank.
The launch of Latam-GPT marks a significant stride in creating AI technology tailored to local needs and cultural contexts, addressing long-standing gaps in AI representation across Latin America.
