Deep neural networks trained on big data have revolutionized information
processing applications, in particular in vision and now also in language
processing. I would like to contrast these networks with models of brain
functions, specifically with recurrent networks that describe reaction
time dynamics. I will discuss an example of such a recurrent network
called a neural field model, and I will show its ability to describe eye
movements and an internal models for a robot arm. While these are studies
in cognitive neuroscience, it will allow us to discuss new directions in
machine learning such as Turing machines and small data applications.