Google applies machine learning to cryptography
 

ViewBase - Cryptocurrency Market Insights

Interesting paper published by Google - Learning to Protect Communications with Adversarial Neural Cryptography (https://arxiv.org/pdf/1610.06918v1.pdf)

An interesting situation they did not considered in the paper:
Would an attacker that had access to Alice and Bob's neural nets, but not the private key, be able to formulate an algorithm to de-crypt their messages? I'd guess not, but I don't believe they explored this in the paper.

Someone had a go at implementing it in Tensorflow:
https://github.com/ankeshanand/neura...phy-tensorflow