Just finished reading the book “On Intelligence” by Jeff Hawkins(Palm Pilot fame) last night. It is a good read. The book is about understanding brain. Jeff’s fascination and interest in understanding the brain is awesome, the book maintains the curiosity though the end. I enjoyed reading it. Especially reading about invariant representations. How brain remembers patterns independent of which senses they were collected from. He gives a structure/model for explaining the invariant representation based on the findings on neocortex’s inter-connections .It also dwells into questions like consciousness, intelligence, which are a always fascination to most humans and argues that everything could be explained by the learning(pattern matching) algorithms of neocortex. The book is a nice read, will be great to follow Numenta, the product based on hierarchical temporal memory. I have seen the applications in computer vision to identify dogs on their website.
The book doesn’t assume anything about reader’s background. All concepts are explained in simple terms, except for the part where he introduces the layers in the neocortex where it gets a bit into deeper into the biology of human brain/brain-cells. The best part for me were the parts with invariant representation. I have thought about brain myself in the terms of psychology. But how would we make machines understand things is the basic question, but On Intelligence gives a clearer picture of how the information is routed and stored and predicted and recognized when similar things happen. Of course, implementing and building a machine is the target of Numenta.
Of course, there is a lot in the book than what i mention here. I am very glad i read it. Definitely recommend it!