The Grok prediction engine from Numenta announced

Hawkins' reasoning struck me as very misguided. I'm not an expert in the field and I'm not up to speed on what Numenta's current form is, but his concept of creating half an intelligence has always offended my sensibilities.

I wanted to say something useful here but I don't know enough about what Numenta's actually doing. I know much of what is not being done, and it basically amounts to willful ignorance in my opinion.

Edit: let me revise this, since I'm being voted down with prejudice anyway. Hawkins has stated very clearly that the only valuable part of the brain to ML/AI research is the neocortex. He's plainly wrong, it should be easy enough to understand this is the final touch on the evolution of our brains—the least important in a survival sense. More directly, without a motivating force there is no opportunity for an intelligent system.

My gut reaction is that the whole thing is a fraud, but I have up to now refrained from using such terms because I'd rather not argue from a position of ignorance (there are things I don't know about Numenta.) On the other hand, it's obvious this subject brings about the ignorant, so I'd like to take the opportunity to address these people: you are deluding yourselves, just because you want to believe something doesn't make it true. (That goes double for you Jeff.)

Have a nice day.

> ... people seem to give Numenta a very dogmatic approval and anyone who seems to question it on discussion boards gets flamed in a religious manner.

I wonder why that is? Is it that Numenta is doing good nerd-marketing. It appeals well to the technical crowd (not AI experts but to people who love technology / programming ). I think among that crowd (and I am one of them) there is the hope that maybe the AI revolution is finally here. This one approach will take us "over the edge" into some kind of singularity. Over the years there have always been what seemed to be very promising "revolutionary" approaches that always made it seem a machine passing the Turing Test is "just around the corner". Wanting to have this hope and wanting to dream is what keeps these kind of stories going.

I am no AI expert but just judging from what I know and if I had to make a guess I would say this is mostly marketing. It seems to me AI is not a field where an "outsider" comes, gives TED talk, writes a book and now we have a revolution. In a way, to me this smells like http://en.wikipedia.org/wiki/A_New_Kind_of_Science type situation. Maybe I am just jaded and skeptical and good thing I am not involved in any such research (I would have given up and nothing would have been discovered) so I am glad people are researching and keep working on the problem, I just don't think this is "it" yet.

I don't think Hawkins is necessarily saying the neocortex is the only part of the brain that's interesting to understand.

I think the main argument is that there is reason to think that there is a single "neo-cortical algorithm", that it is somewhat simple compared to the massive complexity of the brain as a whole and that if you can understand this algorithm, there will be massive payoffs.

That argument too might be misguided but it seem like a pretty focused idea for going from brain to artificial intelligence (a problem where one would expect that simplifying approaching would be quite necessary).

I'm more doubtful of his idea that the neocortex is primarily about prediction. I suspect that's only one aspect.

I breezed through the PDF on the HTM Cortical Learning Algorithms and there's nothing new since the 80s -- over-fitting is still a problem. Perhaps there's an implicit assumption that because learning is online, over-fitting doesn't happen or isn't important.

On a related note what are some good algorithms for anomaly detection?

I want to run my credit card transactions through it to flag things for review.