The machine-learning programs that underpin their ability to “see” still have blind spots—but not for much longer ...

Artificial intelligence (AI) systems can be fooled by certain image inputs. Called adversarial examples, they incorporate ...