Monday, April 21, 2008

US Patent 7359888 - Nanowire crossbar neural net

Conventionally artificial intelligence is based on software configured for general purpose microprocessors or as application specific integrated circuitry. However, since the underlying materials used for microprocessors and ASICs are based on silicon or other semiconducting crystalline structures there are limits to the flexibility of A.I. designs in the conventional approaches. Electronics based on molecular or amorphous nanomaterials which are more capable of reconfigurability may not be subject to such constraints. This is a fairly interesting patent from Hewlett Packard which teaches one way of implementing such a configurable A.I. using a neural net formed from nanowire crossbars. Claim 1 reads:

1. A nanoscale neural network circuit comprising: input nanowire signal lines; output nanowire signal lines; and at least one layer of neural network nodes programmed into a molecular-junction-nanowire crossbar.