Re: [Robot-for-President] Re: Why would AI want to be friendly?

From: L Misek-Falkoff (include@worldnet.att.net)
Date: Fri Oct 06 2000 - 08:17:25 MDT


[Non-member submission]

Greetings:

On this question, I think the Lenat 'Context' work is quite pertinent. I
followed a link you provided (thank you) and looked at a paper on Context.
If it's ok to say so, Lenat & Misek-Falkoff are in accord, if not ( or
though not of course), in practicality, similarly situated <g>.

In 1971, part of my Doctoral Dissertation "Computing a Context' dealt with a
theory of "Structural Thematics" (later developed into my "Claim Structure
Grammar." I used 2 sets of 12 context codes as exemplary I think and took
many years to represent same (using punch cards, those nice little
fixed-field stroage devices later mappable to relational data bases). An
English major in disrepute for using computers, and maybe just "because," I
wrote at great length, but a Technical Report I have I could send u FWP.
"Automated Contextual Analysis of Thematic Structure in Natural Language,
A.R. Jennings Computing Center, under ARPA grant and in conjunction with
M.I.T. Project MAC, and how does that date me!).

Lenat's work takes cognizance of human abilities to contextualize their
behavior and that of others. I must read the paper in detail later in the
month, but I believe they found it too expensive and arduos to build up
enumerative knowledge bases (in AI context) and developed a conceptual
framework, which on brief review I myself conceptuatized as largely but not
totally spatial & temporal, reminding me of Patrick Winston's early work
and some aspects of Minsky.

Sentences presented by Lenat show presuppositional and other context
constructive encoding and decoding capacities of humans. This resonated with
my approach of communication diagramming, and I was pleased to print out a
nice degram from Lenat I will use with some of my ancient renditions. I
thank you for the link in your prior post. Very interesting work.

This now brings us (me) to Professor Ludkowski's succinct and lyrical but
concrete rendering.
----------
> The transhuman can beat the living daylights out of you at chess or Go or
> > poker, and do the same to Deep Blue and Kasparov with scarcely more
effort.
> > Ve can hack source code, prove the Riemann Hypothesis, win a debate,
offer
> > psychiatric counseling, author a scientific paper, design experimental
> > procedures, write a poem, paint a picture, and create new technologies.
Any
> > other questions.
-----------

My sense is that contextual concepts are responsive to the Professor's aptly
poised contrasts, and also that canonical representations will probably
(always?) have to characterize mans' efforts to achieve machine
intelligence, if not to affect it, as previously discussed here. Whereas -
who knows how human intelligence works? Not back to punch cards perhaps. A
perhaps orthogonal issue you have broached, prior, is whether that matters.

What think you. FWP?

:) ldmf. At your service abidingly.

<=email from: ldmf = L.D.Misek-Falkoff, Ph.D., J.D..
include@worldnet.att.net NYState
http://www.egroups.com/group/DisabilitiesParty.com (Chair, Chronic Pain
Delegation)
http://data.ole.net/news/details3.hbs?myrec=213 NPD 2000 (Sept) Online
Convention/Coverage
http://www.egroups.com/group/CYBERLIBEL-AND-EMAIL
http://www.egroups.com/group/Poems_of_Pain_and_Promise
http://www.egroups.com/group/ChronicPainCooperate
http://www.egroups.com/group/Trigeminal_Neuralgia_and_LAW

----- Original Message -----
From: Franklin Wayne Poley <culturex@vcn.bc.ca>
To: <extropians@extropy.org>
Cc: <robot-for-president@egroups.com>
Sent: Monday, September 25, 2000 6:28 PM
Subject: [Robot-for-President] Re: Why would AI want to be friendly?

> -------------------------- eGroups Sponsor -------------------------~-~>
> eGroups eLerts
> It's Easy. It's Fun. Best of All, it's Free!
> http://click.egroups.com/1/9067/18/_/433155/_/969920932/
> ---------------------------------------------------------------------_->
>
> On Mon, 25 Sep 2000, Eliezer S. Yudkowsky wrote:
>
> > Franklin Wayne Poley wrote:
> > >
> > > I have given hundreds of IQ tests over the course of my career and
> > > participated in the development of one of them (Cattell's CAB). If I
were
> > > to measure transhuman-machine intelligence and human intelligence; and
> > > compare the profiles, how would they differ?
> >
> > The transhuman would max out every single IQ test. It is just barely
possible
> > that a mildly transhuman AI running on sufficiently limited hardware
might
> > perform badly on a test of visual intelligence, or - if isolated from
the
> > Internet - of cultural knowledge. A true superintelligence would max
those
> > out as well.
> >
> > The transhuman can beat the living daylights out of you at chess or Go
or
> > poker, and do the same to Deep Blue and Kasparov with scarcely more
effort.
> > Ve can hack source code, prove the Riemann Hypothesis, win a debate,
offer
> > psychiatric counseling, author a scientific paper, design experimental
> > procedures, write a poem, paint a picture, and create new technologies.
Any
> > other questions?
> >
> > -- -- -- -- --
> > Eliezer S. Yudkowsky http://singinst.org/
> > Research Fellow, Singularity Institute for Artificial Intelligence
>
> Yes. According to "Futurama" which is where I get my better ideas on
> machine psychology, John Quincy Adding Machine was the first
> Robot-for-President. According to your web site, "final-stage AI" will
> reach "transhumanity...probably around 2008 or 2010". In which election
> will you first run John Quincy Adding Machine?
> FWP
>
>
> *** The Era of Total Automation is Now ***
>
>



This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:50:15 MDT