The interesting thing about that RNN that you linked that writes code, is that it shouldn't work at all. It was just given text files of code and told to predict the next character. It wasn't taught how to program, it never got to see an interpreter, it doesn't know any English yet has to work with English variable names, and it only has a few hundred neurons to represent its entire knowledge state.
The fact that it is even able to produce legible code is amazing, and suggests that we might not be that far of from NNs that can write actually usable code. Still some ways away, but not multiple decades.
The fact that it is even able to produce legible code is amazing
Somewhat. Look at what happens when you generate code from a simple character-level Markov language model (that's just a look up table that gives the probability of the next character conditioned on the last n characters, estimated by frequency counts on the training corpus).
An order-20 language model generates fairly legible code, with sensible use of keywords, identifier names and even comments. The main difference with the RNN language model is that the RNN learns to do proper identation...
DeepMind's go AI, called AlphaGo, has beaten the European champion with a score of 5-0. A match against top ranked human, Lee Se-dol, is scheduled for March.