Researcher Fool a Google Image Recognition AI Into Thinking a Rifle Was a Helicopter

To safeguard AI, we’re going to need to solve the problem of ‘adversarial examples.’
http://bit.ly/2oXdUZH