Chile launches Latam-GPT as Latin American AI to combat US-centric bias

Chile has unveiled Latam-GPT, an open-source artificial intelligence (AI) model tailored specifically for Latin America to counter the US-centric bias which is common in global AI systems.

Latam-GPT is developed by the Chilean National Centre for Artificial Intelligence (Cenia) and is trained on millions of data points from across the region, showing a rich cultural diversity.

Addressing bias and promoting regional sovereignty

President Gabriel Boric spoke about the importance of Latam-GPT in positioning Latin America as an active and sovereign player in the future digital economy. “We're at the table — we're not on the menu,” he said.

Chile's Science Minister Aldo Valle also reflected that the model was created to counter stereotypes and generalisations about Latin American people and countries. He also warned that without regional involvement, Latin America might substantially lose its cultural heritage in the AI ​​revolution.

What sets Latam-GPT apart from commercial AI models such as ChatGPT or Google's Gemini is that it is open-source, allowing programmers across Latin America to customise it as per their needs.

Universities, libraries, government bodies, and civil society organisations in countries including Argentina, Brazil, Colombia, Mexico, Peru, and Uruguay have contributed immensely to the development of Latam-GPT.

It is mainly trained on Spanish and Portuguese content, and Latam-GPT's developers plan to incorporate Indigenous Latin American languages.

The model is capable of recognising regional slang, idioms, and speech patterns, and is therefore more relevant to local users.

With a modest budget of $550,000 and plans to train future versions on a supercomputer in northern Chile, it is believed that Latam-GPT would empower Latin America with AI that respects its distinct identity and needs.