Deep Learning Adversarial Examples – Clarifying Misconceptions

The 6th point is very interesting:<p>&quot;Adversarial examples generalize across models trained to perform the same task, even if those models have different architectures and were trained on a different training set.&quot;