On Wed, Jan 14, 2015 at 5:52 PM, Colin Geoffrey Hales via AGI
<[email protected]> wrote:
> https://www.youtube.com/watch?v=6i9kE3Ne7as

Christof Koch seems to give 4 different definitions of consciousness
at different points in the video. In order:

1. "What it's like", distinct from the mental state of being awake.
2. Ability to respond to sensory input in a way that you would expect
a human to react, independent of brain defects or injury such as the
lack of a hippocampus (long term memory), emotions, language
(aphasia), ability to move (locked in syndrome), etc.
3. Tononi's integrated information (phi), which Koch interprets as the
maximum increase in complexity (in bits) of the system over all
possible partitions. (The interpretation is important because Tononi
does not rigorously define phi. Maguire et. al. show that under a
reasonable information-theoretic interpretation that phi is not
computable. http://arxiv.org/abs/1405.0126 ).
4. Ability to pass a "what is wrong with this picture" test, in which
pictures depict common objects that don't belong together (for
example, a man standing sideways in mid air).

I think that by P-consciousness you mean something like (1). But since
we can't test this directly, we normally use something like (2) (which
Koch seems to imply without directly stating it). That is,
consciousness means acting like a human. This is why we devise tests
like (4), by picking something that is easy for humans but hard for
machines. You seem to have this approach, because "doing science"
(identifying the problem, researching related work, designing the
experiment, analyzing the results, and publishing a peer reviewed
paper) seems to require a lot of skills that only humans possess.

Likewise, Koch's test requires vision, language, and a common sense
model of the world, which are all hard problems for machines. Google
seems to have nearly solved the problem, although the error rate is
still higher than human level. http://arxiv.org/pdf/1411.4555.pdf

What I don't understand is how you express so much confidence that
consciousness is the magic key to solving AGI, and yet the conclusion
of your book is that you are totally befuddled by it. If I showed you
an optical illusion like
http://miltonious.com/wp-content/uploads/2012/06/Optical-illusions-miltonious-lines-parallel.jpg
you could easily accept that the lines are really parallel by
measuring them with a ruler. You could accept a neural explanation for
the illusion, that the neurons in your visual cortex that detect lines
at various orientations mutually inhibit each other to make the
intersection angles appear larger. You would not insist on some new
unknown kind of physics that is distorting the ruler because your eyes
never lie. Yet when I offer a neural explanation for the sensation of
"what it's like" (reinforcement of sensory perception), that is
exactly what you are doing.

-- 
-- Matt Mahoney, [email protected]


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to