Ed Porter wrote:
Richard,
In your blog you said:
"- Memory. Does the mechanism use stored information about what it was
doing fifteen minutes ago, when it is making a decision about what to do
now? An hour ago? A million years ago? Whatever: if it remembers,
then it has memory.
"- Development. Does the mechanism change its character in some way
over time? Does it adapt?
"- Identity. Do individuals of a certain type have their own unique
identities, so that the result of an interaction depends on more than
the type of the object, but also the particular individuals involved?
"- Nonlinearity. Are the functions describing the behavior deeply
nonlinear?
These four characteristics are enough. Go take a look at a natural
system in physics, or an engineering system, and find one in which the
components of the system interact with memory, development, identity and
nonlinearity. You will not find any that are understood.
“…
“Notice, above all, that no engineer has ever tried to persuade one of
these artificial systems to conform to a pre-chosen overall behavior….”
I am quite sure there have been many AI system that have had all four of
these features and that have worked pretty much as planned and whose
behavior is reasonably well understood (although not totally understood,
as is nothing that is truly complex in the non-Richard sense), and whose
overall behavior has been as chosen by design (with a little
experimentation thrown in) . To be fair I can't remember any off the
top of my head, because I have read about many AI systems over the
years. But recording episodes is very common in many prior AI systems.
So is adaptation. Nonlinearity is almost universal, and Identity as you
define it would be pretty common.
So, please --- other people on this list help me out --- but I am quite
sure system have been built that prove the above quoted statement to be
false.
Ed,
You have put words into my mouth: I have never tried to argue that a
narrow-AI system cannot work at all.
(Narrow AI is what you are referring to above: it must be narrow AI,
because there have not been any fully functioning *AGI* systems
delivered yet, and you refr to systems that have been built).
The point of my argument is to claim that such narrow AI systems CANNOT
BE EXTENDED TO BECOME AGI SYSTEMS. The complex systems problem predicts
that when people allow those four factors listed above to operate in a
full AGI context, where the system is on its own for a lifetime, the
complexity effects will then dominate.
In effect, what I am claiming is that people have been masking the
complexity effects by mollycoddling their systems in various ways, and
by not allowing them to run for long periods of time, or in general
environments, or to ground their own symbols.
I would predict that when people do this "mollycoddling" of their AI
systems, the complex systems effects would not become apparent very soon.
Guess what? That exactly fits the observed history of AI. When people
try to make these AI systems operate in ways that brings out the
complexity, the systems fail.
Richard Loosemore
P.S. Please don't call it "Richard-complexity" .... it has nothing to
do with me: this is "complexity" the way that lots of people understand
the term. If you need to talk about the concept that is the opposite of
simple, it would be better to use "complicated". Personalizing it just
creates confusion.
-------------------------------------------
agi
Archives: http://www.listbox.com/member/archive/303/=now
RSS Feed: http://www.listbox.com/member/archive/rss/303/
Modify Your Subscription:
http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4
Powered by Listbox: http://www.listbox.com