My wife and I just got back from a very restful trip out to the Shepherd of the Ozarks Camp in Arkansas.  It was really wonderful.  I scheduled next to nothing, made some new friends and caught up on sleep.  I finished two books while I was there – The Design of Everyday Things, which I’m reading for a class this semester, and On Intelligence, which I read just for fun.  I felt as though The Design of Everyday Things would be best rewritten as a brochure about design – there were some really great points, but the book did not need to be 272 pages.

On Intelligence, however, was wonderful.  The book is by Jeff Hawkins and Sandra Blakeslee.  Jeff Hawkins started Palm, among other things, and is a neuroscientist along with his many other technology ventures.  He talks about a general idea of brain function, specifically dealing with the neo-cortex, but casts the whole thing in a technological light.  His discoveries have serious implications for intelligence in computer systems.  He outlines a new system, better than neural networks and has a company called Nuementa which is trying to implement that system.  I won’t be using his work this semester in grad school, but I am using neural networks to do some basic pattern recognition.  I wish I had more time to research his technology.  If you want to know more about Jeff Hawkins, you might check out his talk at Stanford.

I also met a wonder retired professor that has become a friend and I look forward to talking more.  He’s been very helpful so far in providing some much needed guidance to a career path that seems a little ambiguous at times.  All in all, it was a really wonderful trip.

So, over the past year and a half I’ve been looking into Neural Networks in Flash.  Neural Networks are essentially used in “Artificial Intelligence” and Pattern Recognition.  So I consulted with one of my best friends who is an embedded systems engineer and had him explain to me how they work (as he went to school for programming/etc and took a class of AI).  Well, about a week or two ago I finally got everything working and in a format that I can show to people.

So the video you’re about to see, should you choose to watch it, is just showing the program “training” on some Xs and Os and figuring out the distinguishing features between the two, then when it’s done training I draw some new (never before seen) Xs and Os and it tries to tell me which one is which.  As you’ll see, it was successful, but it isn’t always successful in both training correctly or the recognition afterwards.  I still have some tweaking to do.

Neural Network in Flash by Philip Likens

Best,
Philip Likens
Texas based Interactive Artist and Web Design Instructor