It seems to me that some properties of large systems of neurons (brains)
are fundamental, while others are emergent. Much of the controversies in
many threads appear to reflect people's differing opinions as to what is
fundamental, and what is emergent.

To illustrate, in the smallest multi celled animals with neurons, like
hydras, the neurons perform primarily process control functions. Even in
us, our hypothalamus that performs process control functions seems to
function largely independently of the rest of the brain. Hence, there are
pretty good arguments that process control is fundamental.

However, things that only man does well and some primates do poorly, like
preplanning complex activities, is almost certainly an emergent property of
some particular development in their nervous systems. Hence, there is
probably nothing in our genomes or construction that is "designed" to do
this, but rather, components that have been honed to do other things just
happen to be able to do these things. Given the wide variation in things
like "intelligence tests", we are apparently just barely able to do these
things. These are the emergent properties.

Of course, when a property emerges, it then gets refined by evolution so
that the "line" between fundamental and emergent soon becomes blurred.

It seems (to me) that the entire thrust of AGI is to DESIGN systems to be
able to mimic the EMERGENT properties of neural systems, rather than
"designing" systems to do other things, one of whose emergent properties
happens to be GI.

The world from this particular POV looks interesting, because the entire
"AGI debate" looks a bit different:

If AGI can be made to work, it would probably be MUCH more efficient than
neural systems (which is one of AGI's selling points). However, there is
absolutely NO evidence that this is even possible, because GI is an
emergent property. Further, we have good evidence in that we took ~150
million years to evolve, that there is NO biologically-realizable direct
solution to GI. Of course, this casually casts aside the central
presumption of AGI, that we ARE a direct solution to GI.

OTOH, Trying to design hyper-complex systems to have particular emergent
properties would seem to be extremely difficult, probably requiring the
same sort of experimentation as has been going on for the last 150 million
years - or some sort of "direct readout" like neurological diagramming,
etc. Of course, this casually casts aside the dreams of generations of NN
developers.

I suspect that a debate about which properties are emergent and how to
design systems that do such things could quickly sort through the masses of
AGI-related arguments.

Steve



-------------------------------------------
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