I've dropped this poisoned pill (along with many others) a few times
but nobody picked it up, so here is a more explicit version.

First, let's start by saying that we are not going to deny General
Intelligence to people who are bound to hospital beds and wheelchairs
and can communicate only by batting an eyelid or something like that.
But for those of us that are building AGI babies/children etc, and I
would include myself here, we will probably need more than a one-bit
channel to grok the world. Much more interestingly, if we were to
"survive" on our own, meeting all kinds of real-time i/o requirements,
we might need more or less bandwidth to the world, with admittedly
enormous variations depending on a more or less malformed "universe",
ranging from one where death is instant to one that survival is
unconditional and therefore any i/o is unnecessary.

Are there at all any proofs/metrics/guesstimates regarding the visual,
auditory or whatever bandwidth required to "decode" a real or virtual
world? Let me just note here that, for all I know, we seem to have
"decoded" physics purely by vision, though it is not unimaginable that
a tribe of blind Einsteins could create opto-auditory experiments that
would expose the underlying quantum world. I have to note as well that
the arrogance of modern physics may be completely unwarranted, there
is no proof that our world is not so malformed that we are shielded
from the true "patterns". We could also take the view that our physics
experiments would not have been possible without the other senses that
participated in the creation of scientific infrastructure, for example
touch and vibration while building apparatus and smell while
distinguishing compounds and elements that are necessary in
experiments. Again, we could "rig" spectrographs to give us audio
clues regarding the purity of substances, but could we build a
spectrograph just by ... looking?

So, back to the troubling question: could it be that world-decoding
requires 2 "retina-cams"? Will VGA do? Could it be that VGA has to be
supplemented by dog-like smell? Can a small virtual world be
decoded/survived with a fixed, small bandwidth? A world as small as
chess or poker needs nearly complete information for full-on
engagement, how can it be we are decoding a humongous universe with
two retina-cams? In Conway's Game of Life there is no place for
sensors and AGIs because of the determinism,right? I mean, how can you
create a pattern that creates an internal representation and
expectations of the surroundings without being able to choose a time
and place for experimentation, repeat under the same conditions etc.
Determinism aside, if there were any structures that looked like
environmental probes they would probably be huge.  But how to create
two and more modalities, a Life "hearing" and a Life "smell"? I'd
suggest we keep "cheating" with virtual domains, ie we do not try to
implement AGI with Life's components, rather we dualistically maintain
the "logic" outside the virtual world but the sensors have to be in
the world, we should not make available to the AGI global and complete
data as if a god looking from above.

As I am writing this I can't help thinking that you have to bring a
full fledged AGI to the decoding/survival problem, bootstrapping your
environment and "invironment" to the point of having an unlimited
intelligence ecosystem probably requires more intelligence than the
intelligence you need with the ecosystem completed and in full effect
with all kinds of machines and computers and all kinds of entropic
impossibilities around you. It could also be argued that you could
come up with string theory and all of mathematics before you ever got
your first bit from the environment (though interaction with the
environment has been vital for all human concepts and achievements).

AT


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com

Reply via email to