Thursday, August 08, 2013

neuroscience just changed its mind.

I studied a bit of neuroscience in grad school as a way to understand human thought - particularly why it is not easily explained and definitely not similar to the way computers work.  Even when computers seem to think like people (e.g. Watson on Jeopardy), it is usually just an illusion of anthropomorphism.  In grad school, we used a basic Hebbian model of the way the brain works.  I was always amazed that a theory developed in the 1940s worked better than anything that had been proposed since.  

Of course, he didn't have the fMRI and other scanning technology so his models were all conceptual.  Now that we have sciences of neurobiology, neurochemistry, and neurophysiology, we have a lot more insights into the details.

I am in the middle of reading Ray Kurzweil's new book "How to Create a Mind" and he talks about how the cortical column creates a hierarchical structure that is very familiar to him.  He has spent his lifetime in AI and his pattern recognition model of the mind fits this perfectly.  Except perhaps it doesn't.  The latest research suggests that it is not so simple.  Even a brilliant thought leader like Kurzweil can be wrong.  I guess there is hope for the rest of us.