Proposes there is a single neocortical algorithm, which is termed the memory-prediction framework. We know that the cortex is hierarchical from studies of the visual cortex and processing. And we know it has six layers and is organized in columns. So, it is a very large thin pancake organized into (millions?) of columns.
Each column is a pattern matcher with a vocabulary of patterns. It see a pattern and sends the "name" of that pattern up the hierarchy. Or, it says HELP!!!. Higher levels of the tree may be able to reconcile novelty, but, if it is new, it will break thru to our consciousness.
recognition travels up the tree. Training travels down the tree. It pushes memory down the hierarchy. Think hand eye, or any skill you have that is "second nature".
Nerve bundles or "tracts" travel between columns/region to other columns or regions in the same hemisphere, or (2) through the corpus callosum (mostly) to the other hemisphere, or (3) out of the brain like the corticospinal tract (CST). Need to learn more about pyramidal cells (which are the only cells? to send out these "projections").
Check these out:
Image from Visual Physiology Pages
Showing Six Layers 2D Map of "pancake" of the Macaque Cortex, from from Felleman, D. J. and Van Essen, D. C. (1991) Cerebral Cortex 1:1-47.
Brain Overview ( randomly chosen?,
fMRI Viz, <> YON's Volume Rendering Page <> Emissive Clipping Planes (SIGGRAPH 2003) <> YON - Jan C. Hardenbergh
I finally finished. midnight, I guess this message is a draft, but never to be finished. I'll send a note when this gets finished - https://www.jch.com/jch/hawkinsi.htmlThe book proposes there is a single cortical algorithm, which is termed the memory-prediction framework. We know that the cortex is hierarchical from studies of the visual cortex and processing. Hawkins has the same algorithm running at all levels of the hierarchy. Essentially (to me) the lower levels recognize a pattern of signals and produce a name from region (perhaps even cell?) specific vocabulary. That name is passed up to become one of the signals to the next layer. Since it is a name, it is a very pure signal allowing invariance.
Good example, walking thru the same door day after day without thought unless something changes. You do this with thinking - no consciousness, unless something is different.
Hawkins spend the first three chapters dumping on AI and neural networks and basic brain anatomy. We have the reptilian brain with the limbic system and not much new. The neocortex is a six layer deep sheet wrapped around the cortex? or it is equivalent to the cortex? Anyway, the neocortex has special memory which is (p.70)
Mountcastle in 1978 and Otto Creutzfeld in 1977 noticed the the neocortex is totally 2D. Where is a 2D sheet? What a nightmare. Bad enough to flatten out a sphere, but, how about flattening out an encephalized sphere!? As an aside, the white matter is very fine and dense material. Can't make out anything at current resolutions, but, you can use DTI - Diffusion Tensor Imaging to follow the Tracts. Optic Radiation are tracts.
Three circuits (p. 147) - a review midway thru the 70 page "how the cortex works" chapter.
So, Forrester, you have probably paged out all of this stuff.
I liked the section on creativity and consciousness - in the end the only reasons we are smarter than the chimps is that our cortex is bigger and we have a more advanced language - just a matter of degrees.
The notion that it takes a tremendous amount of creativity to perceive correctly is good.
All this stuff about sequences makes Gardenfors explanation of the power of metaphor - the same geometric shape in a concept space transfered to a new domain seem apt. The same pattern of patterns applied to a new domain.
The predictions wears a little thin - "We believe.. creative when our memory prediction system operates at a higher level of abstraction." Making the connections is NOT predicting them. "Oh, I recognize this" - TOO Simplistic. Hawkins goes on ... "Creative work violates prediction..." So does noise!
[humor short circuits expectation]
Gyrencephalization is a magnificent word. page with highest rank for it is on The Pyramidal Cell in Cognition: A Comparative Study in Human and Monkey: http://www.jneurosci.org/cgi/content/full/21/17/RC163
YON - Jan C. Hardenbergh - https://www.jch.com/jch <> Voxelographer no evolution. sometimes it depresses me. <> Garbage @ Avalon On Apr 3, 2005, at 11:14 AM, Forrester H Cole wrote:
I finished on intelligence yesterday. Definitely an interesting read, although I think the philosophical aspects of it are kind of weak. I don't think Searle would agree, for instance, that having the Chinese Room make predictions about what symbols it will see next would make it understand chinese. The point (which I don't agree with) is that an automaton and a set of rules can never understand something the way a human can.
Also, I find the definition of intelligence as something that has memories and makes predictions somewhat unsatisfying. Towards the end of the book, Hawkins talks about the applications of intelligent machines to various things, such as cars and proving mathematical theorems. If you have two cars, one with a cortical algorithm driving it and one with a conventional AI system, but they both behave in the same way, I don't really see how you are justified in calling the cortical one intelligent and the conventional one not. Hawkins is right that behaving like a human being shouldn't be the definition of intelligence, but neither should thinking (in part) like a human being.
I guess I think that we should just avoid the word intelligence when talking about computers, the way Hawkins suggests we avoid the word consciousness when talking about brains. It seems that by intelligent we mean either "can perform a set of complex tasks," or "acts like a human being," and in either of those cases we could just say "smart" (as in smart bomb) or "human-like."
I thought the best parts of the book were the parts where he talks about how the cortex works. Of course, those are also the parts about which I know the least, so maybe a neuroscientist would say "the parts about the cortex were garbage, but I thought the philosophical parts were interesting."
F YON - Jan C.Hardenbergh wrote: If that's the
, then I think you will like Hawkin's book. Heck, I can guarantee you will find it worthwhile, refundable in mindless debugging or some such thing. But there is quite a bit about the Chinese Room, etc. Hawkin's goal is to build "intelligent machines" or something. John Sosoka is plowing thru it, too. Another random tidbit hot off my cortex: how we make corporations defend their right of personhood - take them away by town ordinance. http://www.oriononline.org/pages/om/03-6om/Kaplan.html YON - Jan C. Hardenbergh - https://www.jch.com/jch <> Voxelographer Predictability is the very definition of Reality -Jeff Hawkins