Hierarchal Temporal Memory... Rduke???

[Zygote]

Anyone here know anything about this stuff... opinions, comments or critques?

Basically, its a theory and project regarding intelligence and intelligent machines that attempts to model our neocortex and proposes new theoretical foundations for fully understanding the neocortex's function.

Info:


http://www.numenta.com/Numenta_HTM_Concepts.pdf
http://www.stanford.edu/~dil/RNI/DilJeffTechReport.pdf

[Rduke55]

I've worked a pretty good deal on the neocortex but I'm afraid I don't know much about this.

They're playing pretty fast and loose using "the neocortex" alone and in general. But neocortex sounds cool and smart.

And statements like this make me wonder what they're talking about:

"It has been known for over twenty-five years that the neocortex works on a common algorithm; vision, hearing, touch, language, behavior, and most everything else the neocortex does are manifestations of a single algorithm applied to different modalities of sensory input."

But, they're modelers trying to sell their model so what can you do?

It seems to me like they are using general principles of the nervous system to make a better computer but not to understand the nervous system better.

[Zygote]

[ QUOTE ]
I've worked a pretty good deal on the neocortex but I'm afraid I don't know much about this.

They're playing pretty fast and loose using "the neocortex" alone and in general. But neocortex sounds cool and smart.

And statements like this make me wonder what they're talking about:

"It has been known for over twenty-five years that the neocortex works on a common algorithm; vision, hearing, touch, language, behavior, and most everything else the neocortex does are manifestations of a single algorithm applied to different modalities of sensory input."

But, they're modelers trying to sell their model so what can you do?

It seems to me like they are using general principles of the nervous system to make a better computer but not to understand the nervous system better.

[/ QUOTE ]


The idea is based on a theory by Vernon Mountcastle (http://en.wikipedia.org/wiki/Mountcastle). Hawkins claims that this explains why all areas of the neocortex look the same, and why our brains are flexible enough, so that wiring visual information to auditory centers wouldn't hinder our ability to process the visual information.

[Rduke55]

[ QUOTE ]
[ QUOTE ]
I've worked a pretty good deal on the neocortex but I'm afraid I don't know much about this.

They're playing pretty fast and loose using "the neocortex" alone and in general. But neocortex sounds cool and smart.

And statements like this make me wonder what they're talking about:

"It has been known for over twenty-five years that the neocortex works on a common algorithm; vision, hearing, touch, language, behavior, and most everything else the neocortex does are manifestations of a single algorithm applied to different modalities of sensory input."

But, they're modelers trying to sell their model so what can you do?

It seems to me like they are using general principles of the nervous system to make a better computer but not to understand the nervous system better.

[/ QUOTE ]


The idea is based on a theory by Vernon Mountcastle (http://en.wikipedia.org/wiki/Mountcastle). Hawkins claims that this explains why all areas of the neocortex look the same, and why our brains are flexible enough, so that wiring visual information to auditory centers wouldn't hinder our ability to process the visual information.

[/ QUOTE ]

Now Mountcastle I'm familiar with. It's the algorhitm things I'm not clear on.
And there's currently a vigorous debate on how similar the different cortical areas really are.

Also, for some really cool rewiring stuff, check out Mriganka Sur's ferret stuff.

I have to run but I'll check more out on it tomorrow.

[aeest400]

Hawkin's book On Inelligence (http://www.amazon.com/gp/product/0805074562/ref=ase_bookstorenow57-20/103-0961680-7021430?s=books&v=glance&n=283155&tagActionCode=bo okstorenow57-20) (link) is one of the most interesting works discussing how the brain actually computes information. Even if it is not actually correct, Hopkins strikes me as one of the few researches whose ideas hold significant promise for advancing our understanding how the brain functions. There has been so much bad, wasteful work on lame artificial intelligence and pedestrian approaches to cognition (e.g., this box is for grammar) that's it's nice to see someone trying to tackle more difficult issues with a serious grounding in biology.