I wonder if LLM entirely in lobjan would require less text to train to be just as good as the models we have today due to its higher clarity in expression. I a

I wonder if LLM entirely in lobjan would require less text to train to be just as good as the models we have today due to its higher clarity in expression. I also feel the same way about Esperanto due to its high regularity in syntax. Part of me also wonders if an ai model could be optimized by training it on a language specifically invented to be both easy for the ai to understand and the human to communicate in. Maybe the language was made up of lots of 4 letter words bringing the number of words to tokens closer to 1:1, or injecting more context per word so sentences could be shorter and use less tokens