Mike,

On 7/9/08, Mike Tintner <[EMAIL PROTECTED]> wrote:
>
>  What surprises me is that Steve especially - who does unusually here have
> an eye to the real world as opposed to the artificial, textbook world, of
> problems that an AGI and humans must face - seems like nearly everyone else
> to have a massive overvaluation of the powers of logic.
>

I think there is a confusion here in the meaning of "logic". Specifically,
consider...
1,  Logic- that body of boolean and probabilistic methods that would be
unavoidable on the way to a degree in Physics, Math, Electronics, etc. (your
apparent meaning)
2.  Logic - the body of all possible methods, with research continuing into
new methods as problems are discovered that old methods fail to resolve.
There are many minor disciplines that have developed their own special
methods, e.g. police creating the "prisoner's dilemma"; negotiators always
looking for the "win-win solution" (which is a LOT like Reverse Reductio ad
Absurdum); game theory, where the final step is to choose AT RANDOM between
the various courses of action, according to carefully calculated
probabilities, etc. Taken together these provide a qualitative leap in
problem solving abilities that has yet to be appreciated by any but the few
who are fully up to speed on these methods.

I have successfully applied these methods in the medical domain to cure a
number of people with apparently "incurable" illnesses WITHOUT mounting any
sort of major medical research effort. I see no great shortage of knowledge,
but I see a VAST inability to apply what already exists.

>
> If you look around at the vast universe of problems that we deal with -
> scientific, artistic, technological, historical, business, marketing, what
> am I going to cook for supper?, what am I going to wear?, how am I going to
> write my reply to this post? - logic is very rarely used, and only in
> limited ways. This glaring fact seems to have escaped everyone's attention.
>

However, our present "logical" processes have some traps in them, which
result in running into "problems" Once something has risen to the level of
being a "problem", the ONLY methods we need consider are those the people
now do NOT know. I have concentrated ONLY on difficult problems and NOT on
the simpler ones that you are referring to. I claim no expertise at all in
that actually MUCH more difficult domain.

Right now in history, the really valuable low-hanging fruit is in difficult
problems.

>
> Logic is fine if you know all the options for a given problem and all the
> consequences/properties of each option - if the problem.is self-contained,
> perfectly specified and closed-ended.
>

WRONG! This ONLY applies to Type 1 logic explained above. My efforts are in
Type 2 logic, where very little need be known.

>
> But with real-world problems,  you usually only know a few options, if any,
>  you don't know all the consequences even of those, and there are a vast set
> of still-to-be-discovered-and-explored options -  the problem is fluid, very
> imperfectly specified, and open-beginned, open-middled and open-ended. You
> show the nous part of intelligence (which AGI-ers completely ignore) in
> thinking about and exploring those imperfectly-charted parts of the real
> world to find new and improved options..
>
> As Steve says: "I am NOT claiming that Dr. Eliza can solve all problems,
> only those that are constructed with known vulnerable cause-and-effect chain
> links." - i.e only self-contained problems.
>

WRONG AGAIN! I only need to know of two links in most cause-and-effect
chains to be able to cure the problem. The remaining ~80% can remain
completely UNknown.

 And he doesn't give a single example of any real world, major dispute that
> has been resolved by logic or his particular method - because,
> I'm confident, none exists.
>

I have already responded to this challenge in another posting. Many of these
methods are only a few years old and are NOT known to those who are
supposedly in charge of resolving political disputes. That a new technique
(that is not generally known) has not yet succeed is no mark against the
technique, unless and until it has had at least one failure.

However, the medical domain is quite different, e.g. I easily cured my first
case of COPD (Chronic Obstructive Pulmonary Disease, commonly called
Emphysema and believed to be incurable) in Ted with NO specialized COPD
sub-domain knowledge. I simply treated it as a research project, educated
the patient, eliminated obvious contributing factors to simplify the
situation, searched for and eliminated remaining factors, etc. In short, I
simply engaged in a repair process rather than looking for some sort of
semi-magical cure.

>
> Real AGI is reasoning fluidly about real-world, open, ill-structured
> problems.
>
> Narrow AI (and AGI-as-currently-practised) is reasoning rationally
> (logico-mathematically) in structured ways about artificial world, closed,
> neatly-structured problems, (which the programmer has specially prepared
> beforehand). Or you could say - to address Valentina's point - it's
> problem-solved, simplified-world lab science as opposed to
> problem-to-be-defined, complicated-world, in-the-field science.
>

All of the above is in the domain of Type 1 logic.

>
> When will *artificial* intelligence get real?
>

Here, we are in *COMPLETE* agreement. Also, I hereby stipulate that Dr.
Eliza's methods are narrowly applicable to the "fall-out" from simplistic
Type 1 logical processes, and I see little opportunity to extend them to
create anything like an AGI, with the following narrow exception:

One form of "Singularity" is when the Internet becomes smarter than we are.
Presumably we will solve the simple everyday problems ourselves and only go
the the Internet with it is worth the typing. I believe that Dr. Eliza, in
very nearly its present form, would with sufficient backing become that
particular form of singularity, though still not on the path to becoming any
sort of AGI.

Steve Richfield



-------------------------------------------
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=8660244&id_secret=106510220-47b225
Powered by Listbox: http://www.listbox.com

Reply via email to