Does Artificial Intelligence work in English?

(Image: Pixabay CC0) Researchers have shown that large language models primarily trained on English text seem to use English internally, even when they are prompted in another language. As AI increasingly runs our lives, this may have important consequences regarding linguistic and cultural bias. Large language models (LLMs) including Open AI's ChatGPT and Google's Gemini have taken the world by storm, surprising with their ability to understand and respond to users with seemingly natural speech. While it's possible to interact with these LLMs in any language, they are trained on hundreds of billions of text parameters mainly in English, and it has been hypothesized by some that they do most of their internal processing in English and then translate to the target language at the very last moment. Yet, there has been little evidence of this - until now. Testing Llama. researchers from the Data Science Laboratory (DLAB) in the School of Computer and Communication Sciences studied the Llama-2 (Large Language Model Meta AI) open source LLM to try to determine which languages were being used at what stages along the computational chain.