I was fairly unimpressed with Jeff Hawkins' talk about Hierarchial Temporal Memory. I got the feeling that he wasn't using the word "hierarchy" quite the way I understand it, either overstating its applicability or using it as a synonym for "structure", and he had some ideas about the brain that are, as far as I understand, wrong, like that the motor/sensor areas of the brain actually don't exclusively control the brain (a lot of imaging data disagrees with him, at least), that object recognition is entirely a bottom-up process (which is laughable), and that information always converges at higher levels as per what I will assume is what he means when he talks about hierarchy - the branching of the vision system is also relatively well known. The mix of this, that he has published little in reputable journals, that his models arn't head-and-shoulders above competing cognitive models, that there is little-to-no imaging research, as far as I can tell, supporting his ideas, and that whatever he was before, he's a businessman who's aiming to patent his ideas and sell his technology makes me think that he's pretty fringe, at least on the science side of things. He still may make a decent amount of money because while his programs are not expert systems, the attitude underlying his research is that practical results are more important than "doing it the same way the brain does". I suspect there are many good reasons he didn't go to the psych department to talk nor was this widely publicised there.
But... it was something to do. My stomach is full of angry for my having made today's first meal a large quantity of sweets.
