Thanks guys. It is relieving to get your insights.
Hawkins' second set of statements, in my original thread, seemed to
not "feel" right.
Granted, scientists do not know how our grey matter works - not from
an in depth / confirmed stance at least - so its probably fairly easy
to come up with theories. The problem with theories is that the
person who proposed it, usually ends up finding information / reasons
only to back it up. This is blind sided.
As a complete novice, one with a clean slate, and with an open mind,
I can certainly see that the "solution" can go either way (once we
get beyond the oversimplifications - such as Hawkins' "neocortex
with the hippocampus sitting on top" being absurdly simplistic - as
Bo pointed out. At least these simplifications may do the job of
getting us started - towards a direction - to later be proved correct
or not during testing).
As for the "feel" not being right with Hawkins' statement that the
answer lies in the hardware, I felt this way because intuitively, we
should be able to model reality with software. As Josh pointed out,
"once any particular function in the nervous system has been worked
out, it's straightforward to simulate it in software" - this does
"feel" better than Hawkins' claim.
Here's an oversimplification: If we take virtual machine technology
(I'm talking about VMs - like the CLR, and the JVM) and we apply our
learnings about the nervous system, we should get a Virtual Brain.
Right? I'm NOT saying this is the answer of course - nor am I trying
to trivialize the difficult problems. It was just an
oversimplification statement.
Well, just hearing your comments, and thinking about them, actually
makes me feel much better - and hopefully, if any one else on this
thread ever questioned "Will we get there with software on current
CPU architectures?", the answer may very well be a resounding YES!
I can't wait until someone on this thread has a series of mini
breakthroughs that will prove Hawkins was incorrect :-)
~Aki
On 21-Feb-07, at 1:04 PM, J. Storrs Hall, PhD. wrote:
On Wednesday 21 February 2007 11:52, Aki Iskandar wrote:
I'd be interested in getting some feedback on the book "On
Intelligence" (author: Jeff Hawkins).
...
The basic premise of the book, if I can even attempt to summarize it
in two statements (I wouldn't be doing it justice though) is:
1 - Intelligence is the ability to make predictions on memory.
2 - Artificial Intelligence will not be achieved by todays computer
chips and smart software. What is needed is a new type of computer -
one that is physically wired differently.
1 is a piece of the truth, to what extent varies in the opinions of
people
here and has been discussed extensively. It isn't particularly
original with
Hawkins.
2 is silly. Once any particular function in the nervous system has
been worked
out, it's straightforward to simulate it in software. You can vary the
hardware to optimize, if the algorithm is wierd enough (see graphics
accelerators) but that should come *after* you know what you're doing.
"Premature optimization is the root of all evil."
-- C.A.R. Hoare
The other idea in OI worth noting is Mountcastle's Principle, that
all of the
cortex seems to be doing the same thing. Hawkins gets credit for
pointing it
out, but of course it was a published observation of Mountcastle in
the first
place. My AI architecture is influenced by the observation,
although it's not
quite as useful as it might seem. (E.g. it would be satisfied by
saying each
column is a general-purpose processor.)
Josh
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303
