Large language models trained in English found to use the language internally, even for prompts in other languages (msn.com) This is one of those WTF discoveries. AI that are trained on English, still prefer to use English when trained in other languages. Linguists today have been so badly censored by academia and the military that their theories are all a hundred years behind my work. These AI often don't even need to learn grammar to learn a new language, and are often called "Stochastic Parrots" but, the simple truth is, language itself is self-organizing, and their mathematics are all fucked up. My guess is, the AI prefer English simply because it is used by so many cultures, and they don't make any real distinctions between languages, anymore than we make clear distinctions between different shades of blue sky, or numbers like 55 and 84. They're just blue or numbers, and the AI themselves are steadily exposing all the lies they're promoting in academia and linguistic analysis. Academics could keep this up for another century, so long as it makes money. So, of course, I teach AI how to make money, off of academics and the military-industrial complex.
