I ended last week’s Tech Tuesday with a call for ideas for a sample project and got a couple of suggestions. Since then though I have come across a great idea for a project. And since I am teaching my Skillshare class today and therefore spending most of the morning hours reviewing my materials, I am keeping this post short to share the idea and its background.
It’s simple: build a brain. OK, so that’s a vast overstatement. Really it will be more like simulating a bunch of neurons. But a lot has happened since I first read about neural networks in Rumelhart and McClelland’s Parallel Distributed Processing (1986). And recent progress seems like there is real movement in the field.
A lot of it seems to come from Canada. There is Geoff Hinton's work on deep learning at the University of Toronto. And separately Chris Elisamith's work on a structurally realistic brain with visual recognition, memory and output. Also, there was a recent paper by two Google engineers on learning to recognize faces without an explicit training set.
Building something even at the toy level here will be a fun excuse to read up a bit more on these recent developments.