Comparing lexical and syntactic complexities in academic writings generated by AI and those produced by humans