The generally accepted idea is that deep learning neural networks are as as good as the data you feed them. In this blog, we review recent achievements of deep learning and address whether such systems are able to augment and transcend the original learning material through self-learning. And ultimately can become --even better-- than the data you feed them.