> But that ignores another point worth mentioning, which is most sensory
> data the human brain works with is one-dimensional even though we don't
> think of it that way.  Audio, for example, is perceived as a
> one-dimensional signal that is analyzed for spatial cues at a higher
> level.

It's true.  Sound processing is simpler, which is why I advocate starting
off a sensing AGI with hearing and touch as its two senses.


> Vision is more complicated, but even that is more like a 1.5-dimensional
> data stream when you get down to it.

Really, human vision is best understood mathematically as a set of 2D
datastreams, 3 from each eye (3 because of 3 colors).

> I think it is unnecessary to
> fully map low-dimensionality data into a sophisticated and resource
> consuming 3D space to be able to effectively behave as though you are
> fully aware of your 3D surroundings, particularly since humans don't do
> this in the sense I think is commonly believed.  The space is inferred
> from a relatively small collection of vectors automatically being
> stripped from a low-dimensionality data stream and simple processing
> that happens on those vectors.

Well, what really matters is whether the system has easy access to efficient
*geometric operators on its internal virtual dimensional space.*  Can it
rotate its internal images?  Can it translate them?  can it filter them to
deal with poor lighting conditions?  etc.

The human brain is notoriously wasteful of space resources.  For example, it
can afford to keep a whole sheet of neurons representing a whole bunch of
pixels, even if most of these are inactive nearly all the time.  In a
contemporary digital computer one is best of taking a different approach,
and using sparse data structures (like nodes & links) to represent pretty
much everything.  The important question is what operators are efficiently
executable over these sparse data structures....

In Novamente, we can represent spatial info as nodes & links just like other
info, but, to make spatial operators efficient, we'll need to implement a
special index of the nodes/links representing spatial info... so that e.g.
it's very very fast to find the set of nodes representing a bunch of pixels
that are along the line inbetween two given pixels, etc.

-- Ben G



-------
To unsubscribe, change your address, or temporarily deactivate your subscription, 
please go to http://v2.listbox.com/member/

Reply via email to