Palm founder Jeff Hawkins on neurology, big data, and the future of AI

Numenta/Grok/Hawkins harps on this idea of "streams" of data and how traditional NNs don't work well because they lack recurrent connections. The whole "no recurrent connections" thing was not true in the first place because recurrent neural networks most certainly exist and did in 2005, they just weren't very effective. However, fast forward to 2013 and nearly a decade of new research, recurrent neural nets are more powerful than ever. I suspect they could go head to head with anything Hawkins is selling.

The problem more generally ends up as a philosophical one: if we're trying to create a machine capable of intelligence, how much inspiration should we take from the brain? Perceptron based NNs loosely approximate the brain while Hawkins believes a more literal approach is the way forward. Still, even Hawkins' approach stops way short of something like the blue brain project.

Personally, I suspect it's much like achieving flight: We need to build an airplane (a machine that takes advantage of the same phenomena as a bird does in order to achieve flight). Just like the bird is not the only way to fly, I don't think the human brain is the only way to generate intelligence. But until someone can prove it one way or another, I have a feeling it's a debate we'll be having for a while.

Academia is really pushing the boundaries of neural networks and getting machines to perceive visual and auditory input. Great to see commercial ventures step in to create value from those advances. Google is hiring in that field because [0] voice and sight would seem to be the two most natural inputs (vs typing and clicking).

Interesting that Grok is not trying to push the boundaries of neural network size in the same way that new Google employee Geoffrey Hinton has stated as his goal. Grok's approach would likely create huge value to the Honeywell's and GE's of the world. There are lots of turbines, thermostats, generators, and robots to improve.

Will the day ever come that your coffee maker has a Nest-like capability to determine when you're up, a Google Now-like function laity to know if you have time for coffee, and a Grok-like capability to know when to have your coffee ready?

Seems ridiculous, but there's a chip in just about everything. So, why not?

[0] My own speculation.

Here's an related talk by Jeff Hawkins at Google, where he gets into his ideas about the structure of the brain, and what he's done to reconstruct it in software/hardware.

http://www.youtube.com/watch?v=4y43qwS8fl4&list=FLgNLujlOCEu...

Numenta was founded in March 2005, more than 8 years ago. What has it accomplished?