Latam-GPT: Latin America’s Bold Bid for AI Sovereignty
Latam-GPT Launch: Open Source AI, Regional Bias & Digital Sovereignty
"We’re at the table—we’re not on the menu." With these words, Chilean President Gabriel Boric introduced Latam-GPT, the first open-source large language model (LLM) designed specifically for Latin America. Developed by Chile’s CENIA with a coalition of 60+ institutions across 15 countries, this model aims to shatter the US-centric biases found in systems like ChatGPT.
In this episode, we explore how Latam-GPT—built on Meta’s Llama 3.1 architecture with 70 billion parameters—was trained on 8 terabytes of regional data to understand local slang, history, and social nuances. We discuss why having only 4% of global AI data in Spanish is a threat to cultural heritage and how this $550,000 "David" plans to stand up to the "Goliaths" of Silicon Valley. Plus, a look at future applications in health, retail, and the upcoming supercomputer in northern Chile.
Latam-GPT, AI Sovereignty, Open Source, Chile, Gabriel Boric, Llama 3.1, CENIA, Machine Learning, Digital Diplomacy.
This episode includes AI-generated content.
