“Christian Szegedy predicts that within the next few months, the AI field will face threats from both "subversive" (inadvertently evolved) and "adversarial" (explicitly malicious) AI systems.”