Educators and trustworthy ML researchers are concerned about the "black box" nature of AI solving, where students may get the correct answer without understanding the underlying concepts.