Training AI from the Global South: Challenges and Lessons Learned from LatamGPT

The Instituto Cervantes in Los Angeles (ICLA) invites you to this talk by systems engineer Omar Flórez, who will present his experience designing and training LatamGPT, the first foundational language model developed in Latin America, with an emphasis on language, culture, and technological sovereignty in the age of artificial intelligence. Based on this project, the technical, linguistic, and cultural challenges of building AI from the region are analyzed. The diversity of Spanish and Portuguese, the collection of culturally representative data, the efficient use of computational infrastructure, and the role of open models in strengthening scientific and technological autonomy are discussed. The talk proposes a reflection on how AI can learn from local contexts without losing scientific rigor or global ambition.