@goodfellow_ian 's Adversarial examples that fool image classification algorithms by carefully adding noise https://t.co/B4FnIqbQSS pic.twitter.com/iuZey1n1Q9

— Albert Pinto (@70sBachchan) June 28, 2017