SANTiago, June 17 (Reuters) – Latin American countries are working together to launch Latin Americans in September. This is the first large-scale artificial intelligence language model trained to understand the diverse cultural and linguistic nuances of the region, Chilean officials said Tuesday.
The open source project, piloted by Chile's national national AI Centre (CENIA), along with more than 30 regional agencies, is seeking to significantly improve AI ingestion and accessibility across Latin America.
Chilean Minister of Science Aisen Etcheverry said the project “could be a democratic component of AI,” and uses models that reflect local culture and language to envisage applications in schools and hospitals.
Developed since January 2023, LATAM-GPT seeks to overcome the inaccuracy and performance limitations of global AI models, primarily trained in English.
Officials said it was intended to be a core technology for developing applications like chatbots, rather than a direct competitor to consumer products like ChatGPT.
The key goal is to preserve the indigenous languages developed by the first translators already developed for Easter Island's native language.
The project plans to extend this to other Indigenous languages for applications such as virtual public service assistants and personalized education systems.
The model is based on Llama 3 AI technology and is trained using a regional network of computers, including Tallapacca University in Chile and facilities for cloud-based systems.
Supported by Regional Development Bank CAF and Amazon Web Services.
Currently, there is a lack of dedicated budgets, but Cenia Head Alvaro Soto hopes to attract more funds by demonstrating the capabilities of the system. (Reporting by Fabian Cambero, Editing by Lisa Shumaker)
