All the GPT-2/3 articles around appear to be based on English language, which is natural as it is probably the native tongue of its developers, but has there be

All the GPT-2/3 articles around appear to be based on English language, which is natural as it is probably the native tongue of its developers, but has there been research and/or learning done on non-English languages at similar scale? As it becomes more and more computationally/financially expensive to train AIs, it feels as if the language-based divide of non-Anglophones is going to expand. (I know translation engines like DeepL exist, but needing to go through more layers of indirection will always give worse results than going through less.)