Maxime Riché comments on Large language models aren’t trained enough