This is a nice (and freely available) article by Oliver Coleman:
Models of the brain: What Should We Borrow From Biology?
It was presented on the AAAI Symposium How Should Intelligence Be Abstracted in AI Research and reviews several basic properties of neural networks that are possibly of critical importance for their functioning but mostly neglected in artificial neural networks. This is a great article for every computer scientist interested in neural networks and in the neuroscientific perspective on neural networks. It is hopefully a source of inspiration for new and more complete models.