Hypothesis - Large language models (LLM) that were explicitly ...
Hypothesis: Large language models (LLM) that were explicitly NOT trained on English but instead on just the other human languages, will inherently behave more effective and sane. The proof will be in the pudding.