Re: Jeff Hawkins: Why Can't A Computer Be More Like A Brain?

On 15 Apr 2007 16:47:31 -0700, "feedbackdroid"
<feedbackdroid@xxxxxxxxx> wrote:

On Apr 15, 1:47 pm, Traveler <trave...@xxxxxxxxxx> wrote:
On 14 Apr 2007 08:48:29 -0700, "feedbackdroid"





<feedbackdr...@xxxxxxxxx> wrote:
On Apr 14, 7:27 am, Traveler <trave...@xxxxxxxxxx> wrote:
http://www.spectrum.ieee.org/apr07/4982

Here's an excerpt that I like, only because I've been saying the same
thing about the importance of time for many years:

[begin quote]
So how does a node know what to learn? This is where time plays a
critical role and is one of the unique aspects of HTM. Patterns that
occur close together in time generally have a common cause. For
instance, when we hear a sequence of notes over and over, we learn to
recognize them as a single thing, a melody. We do the same with visual
and tactile patterns. Seeing a dog moving in front of us, for example,
is what teaches us that a left-facing dog is actually the same as a
right-facing dog, in spite of the fact that the actual information on
the retina is different from moment to moment. HTM nodes learn
similarly; they use time as a teacher. In fact, the only way to train
an HTM is with input that changes over time. How that is done is the
most challenging part of HTM theory and practice.

Because HTMs, like humans, can recognize spatial patterns such as a
static picture, you might think that time is not essential. Not so.
Strange though it may seem, we cannot learn to recognize pictures
without first training on moving images. You can see why in your own
behavior. When you are confronted with a new and confusing object, you
pick it up and move it about in front of your eyes. You look at it
from different directions and top and bottom. As the object moves and
the patterns on your retina change, your brain assumes that the
unknown object is not changing. Nodes in an HTM assemble differing
input patterns together under the assumption that two patterns that
repeatedly occur close in time are likely to share a common cause.
Time is the teacher.
[end quote]

IOW, it's 100% temporal. So called spatio-temporal intelligence is
nonsense. I have been saying this long before Hawkins came out with
"On Intelligence".

100% temporal. Come on, Louis, that's not what it said.

That's exactly what Hawkins is saying. You just don't get it. The
visual object must move (that's the temporal part) in the visual
field. If it does not move, you can't see it, let alone learn anything
from it.

I don't know about you, but if I hold my eyes as perfectly still as I
can, then I actually see a full colored visual field about 150-
degrees wide. This is not due to micro-saccades either. It's due to
"sustained" responses in retinal bipolar and ganglion cells. It's been
known since day-one there are 2 basic classes of retinal response,
sustained and transient. The transient cell responses require
continual movement.

Also, if I fix the view for longer than a few **seconds**, then it
does fade, as the sustained cell responses gradually diminish.
However, in the meantime, I get to see and recognize quite a lot of
the outside world.

Let me tell you a secret that you probably will not find anywhere
else. Just remember where you read it, ahahaha... Sustained sensory
signals, including proprioceptive signals, are not processed by the
sensory cortices. They go straight to the cerebellum. You want to know
something else, cerebellar processes are 100% unconscious. The
cerebellum is an automaton that receives its training from the motor
cortex via the basal ganglia. Its primary purpose is to help the rest
of the brain handle routine motor tasks, especially when the brain is
busy thinking about other things. So I doubt very much that you can be
conscious of those signals. Do you have a reference to your claim that
you can consciously see anything if objects in the visual field are
kept stationary, even for a few seconds?

Having said that, does that make sustained sensory signals
non-temporal? I don't think so. Every sustained signal has a beginning
and an end, does it not?


[cut]
In biological vision, both temporal and spatial have equal status.

The only thing that has anything spatial about it is the design of the
eye/retina. But that is not intelligence. In thelligence is what is
done with the signals after they reach the brain. That part is
temporal. Hawkins' HTM does not care where the sensory signals are
coming from. He says so in the article.

I know that's what Hawkins says. But he also thinks all computations
in cortical columns are identical too, but this is clearly wrong. As
reference for this, I've recently read a couple of Semir Zeki's books,
and he's been recording from cortical cells for 35+ years.

I think Hawkins is referring only to the sensory cortex. If not, he's
wrong. Still, the brain cell assemblies (e.g., the cerebellum, the
motor cortex, the temporal lobe, the basal ganglia, etc...) are
remarkably homogeneous within their boundaries.

Question: what is spatial about a discrete signal? Answer: nothing.


What is spatial is the retinotopic mapping into the cortex. Cells in
nearby columns respond to visual stimuli at nearby points in visual
space. Without this "side by side" correlation, the temporal
correlation would mean nothing. If the retinal cells mapped randomly
into locations in the cortex, the visual field would just look like
noise.

IMO, the retinotopic mapping is there for physical and biological
reasons only. It has nothing to do with intelligence (signal
processing and neural learning) per se. I've written about this
before. If you could scramble the location of every neuron in your
brain while maintaining the connections intact, there would be no
change in your behavior.

Louis Savain

Physics From the Bible!
Shaking the Foundations of Physics:
http://www.rebelscience.org/Seraphim/Physics.htm
.

Relevant Pages

  • Re: Jeff Hawkins: Why Cant A Computer Be More Like A Brain?
    ... critical role and is one of the unique aspects of HTM. ... and tactile patterns. ... "sustained" responses in retinal bipolar and ganglion cells. ... signals with other sensory data such as music and speech signals. ...
    (comp.ai.philosophy)
  • Re: Seeing is behavior
    ... This tendency for nerves to sort out signals is a known fact ... the brain develops. ... a scaffolding of glia cells as they migrate to their appointed ... it is perhaps a group connection rather than an individual ...
    (comp.ai.philosophy)
  • Re: synesthesia Re: Doing The Narrative Drains
    ... tetrachromacy is how the brain wiring must work. ... splicing a human cone gene. ... then signals from cone 1 cells to group 2 would fail to ... adjacent cells would tend to receive signals from ...
    (rec.arts.sf.composition)
  • Possibility of telepathy information theory (theory of)
    ... Telepathy has been considered one of the phenomena for which science ... On the other hand we know that electric fields ... So it is just a matter of asking: has the brain enough computing power ... signals of other brains? ...
    (sci.econ)
  • Possibility of telepathy information theory (theory of)
    ... Telepathy has been considered one of the phenomena for which science ... On the other hand we know that electric fields ... So it is just a matter of asking: has the brain enough computing power ... signals of other brains? ...
    (sci.physics)