On June 13, 2007, at Pillsbury Winthrop in Palo Alto, SDForumâ€™s Emerging Tech SIG hosted Dileep Georgeâ€™s presentation â€œHow to Make Computers Work Like the Brainâ€.
Dileep started out with a Bachelor’s degree in Electrical Engineering from the Indian Institute of Technology in Bombay and went on to a Masterâ€™s degree in Electrical Engineering from Stanford University. His interest in the brain led to the Redwood Neuroscience Institute where he worked with Jeff Hawkins to express neuroscience in mathematical terms. Proving these concepts, in 2005 he founded Numenta with Hawkins and Donna Dubinsky. By 2006, they had their first working prototype of Hierarchical Temporal Memory (HTM).
HTMs try to replicate the structure and function of the human neocortex. It attempts to solve problems normally easy for humans but hard for machines, like maneuvering cars through traffic. HTMs are not programmed they are trained. They look for patterns. Sensory data is sent from the bottom of a hierarchy through a massively parallel network that eventually reaches a threshold or tipping point decision. A visual recognition HTM would be able to tell the difference between a dog and a cat by collecting enough sensory pattern data from a hundred sensors. Those sensors pass the decisions made at their level up to the next and so on until it reaches the top. The votes are counted and if it is more than fifty percent one or another, the decision is made.
Numenta is opening up its environment to encourage a community to develop HTM systems. Anyone can download a copy from their site. Maybe someday an HTM will recognize this pattern of development and carry on the work.
Copyright 2007 DJ Cline All rights reserved.