IARPA posts RFI for energy efficient computing architectures
On June 8, the Intelligence Advanced Research Projects Activity (IARPA) released a request for information (RFI) for energy efficient computer architectures (IARPA-RFI-16-05). Responses to this RFI are due no later than 4:00 p.m., Eastern Time, on July 15, 2016.
The Intelligence Advanced Research Projects Activity (IARPA) is seeking information on developing novel computer architectures to achieve energy efficient calculations beyond the limitations of current CMOS technologies, and in particular for applications involving Artificial Neural Networks (ANN). This request for information (RFI) is issued solely for information gathering and planning purposes; this RFI does not constitute a formal solicitation for proposals.
Artificial neural networks show promise for making sense out of patterns in data, such as facial recognition or object detection, however, executing software emulations of neural networks is currently much less efficient than the brains from which they were derived. This RFI seeks ideas on more effective computer technologies for this class of algorithm. Existing applications of interest to IARPA include:
The Go game playing program (a 280 GPU cluster beats a very good human player)
Unsupervised learning programs that scan videos to discover common features
Various software packages that include code libraries for common ANN building blocks (convolutional network layers, backpropagation, etc.).
Full information is available here.