No, because the distributedness of the representation means that exponentially large Markov chains would be needed to replace linearly growing deep models. See this paper on the recurrent temporal RBM (http://www.cs.utoronto.ca/~ilya/pubs/2008/rtrbm.pdf).
It's like saying, "aren't binary trees just linked lists?"
It's like saying, "aren't binary trees just linked lists?"