Status and/or thoughts on ELMs? (Extreme Learning Machines)

I've been looking into various NN training options and stumbled across Extreme Learning Machines which, the creators purport, supposedly train much faster than deep networks using more traditional back prop training methods.

I notice this paper from 2004: http://www.ntu.edu.sg/home/egbhuang/pdf/ELM_IJCNN2004.PDF

And an article in IEEE IntelligentSystems from 2013: http://www.ntu.edu.sg/home/egbhuang/pdf/IEEE-IS-ELM.pdf

Does anyone have any experience with ELMs? Impressions, opinions?