Newsgroups: comp.theory.cell-automata,comp.ai.alife,comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!bloom-beacon.mit.edu!gatech!howland.reston.ans.net!swrinde!sdd.hp.com!hpscit.sc.hp.com!lf.hp.com!apollo.hp.com!hpax!news1.boi.hp.com!hp-pcd!hpcvsnz!news
From: 5LG3 (Terrill Snyder)
Subject: Re: Help for new student of ALife, GA, and CA programming
Sender: news@hpcvsnz.cv.hp.com (News )
Message-ID: <D45F2o.2Ds@hpcvsnz.cv.hp.com>
Date: Fri, 17 Feb 1995 14:32:47 GMT
References: <3fkk97$av7@spool.mu.edu> <3fmd3h$djo@network.ucsd.edu>  <3g3psu$cnf@scratchy.reed.edu> <3gna9j$nlp@hptemp1.cc.umr.edu> <3goke5$cos@hptemp1.cc.umr.e <792323884snz@nezumi.demon.co.uk>
Nntp-Posting-Host: seanm.cv.hp.com
Organization: InkJet Business Unit (IJBU)
X-Newsreader: WinVN version 0.82
Lines: 84
Xref: glinda.oz.cs.cmu.edu comp.theory.cell-automata:3332 comp.ai.alife:2473 comp.ai.genetic:4986

In article <792323884snz@nezumi.demon.co.uk>, Martin@nezumi.demon.co.uk (Martin Tom Brown) says:
>
>In article <D3nG09.AL2@murdoch.acc.Virginia.EDU>
>           mjd4c@uvacs.cs.Virginia.EDU "Michael J. Daniel" writes:
>
>[snip]
>> It is by no means clear that if we placed one monkey for each atom
>> in the universe in front of a typewriter and had them type for
>> 1,000,000,000,000,000,000 years, that we would create any interesting works
>> of literature.
>> You may say, "You have not waited long enough!"
>> And I would reply, "Your concept of 'long enough' has no meaning
>> in this universe, but you expect the results from it to live in this universe."
>
>It is fairly easy to calculate how long you might expect to wait, for a single
>configuration to occur by chance - given an alphabet (eg. states for a CA)
>and the number of tokens to be specified. So for example:
>
>Using a keyboard 100 with 100 keys and
>a target 10000 characters of Midsummer Night's Dream (Shakespeare).
>There are               10000
>                     100                ( 1.0E+20000 ) possible states
>Never the less this means that for each random trial there *is* an 
>independent probability of 1 in 1.0E+20000 of getting Shakespeare.
>It is clear that using a reallisable number of superfast typing monkeys
>will not shorten the expected time to wait by much. For example I recall
>the total number of particles in the universe is about 1.0E+100 or so.
>Working out the expected waiting time is left as an excercise to the student.
>This concept of "how long you have to wait is *perfectly* valid" it is just
>that the answer is rather longer than the *age* of the universe.
>
>When evolutionary pressures exist, the choices are not random, and very 
>quicky bunch around the target region as survival of the fittest rules.
> 
>> Similarily with the chinese room problem.
>> I submit that there is no, and never will be any such, book
>> that a non-chinese speaking person can sit in a room,
>> receive intelligent, relevant chinese text, (ie a chapters
>> from a book of philosophy, along with questions on the chapter)
>> and using just the book, translate the questions into intelligent 
>> answers. You can not capture intelligence on a piece of paper.
>> Or even 1,000,000,000,000,000,000,000 pieces of paper.
>> You still only have a stack of paper.
>
>Pure speculation!! I'd like to see your proof.
>I don't think you can prove this one way or the other (yet).
>
>The smart answer is that I could find a *Japanese* person who could read
>the *meaning* of Chinese characters, and learn the rest from Chinese books
>without *ever* knowing what the Chinese sounds like (ie speaking it).
>I also expect that the test subject would learn written Chinese as a side
>effect in the event that this experiment were ever carried out. 
>
>I am inclined to believe that it will be possible to capture intelligence
>on a non-living medium, although I think CDROM or holograms may be needed
>to get a practical system. After all a brain is just a complex configuration
>of interlinked cells made of ordinary atoms (mostly C,N & O).
>A *big* CA might just do it. It's all down to configurational entropy.
>
>Regards,
>-- 
>Martin Brown  <martin@nezumi.demon.co.uk>     __                CIS: 71651,470
>Scientific Software Consultancy             /^,,)__/


It would be a snap to write a procedure that would generate all stories of
(arbitrary length ) fifty thousand words or less, the trick would be to know which ones
 to read.

Even if the monkeys did type all of the books ever written, so what.  A formula
can express the same thing, much neater, and the fact that one possibility exists
as an individual expression on paper, and the other as a potential within a formula
do not make them all that unequal.  The fact that they are all typed makes no 
difference at all, because you still have to FIND the ones that you want.  It would be
no easier than sorting through the permutations of a brute force formula that 
generated all of the books that could ever be written.

The trick to AI for some time to come is going to be ways to find the ones
that we want, in a system that can generate all of the possibilities.



Terrill Snyder

