When can we trust the results we get from AI, and when is learning impossible? Researchers have shown that there are some ...