Newsgroups: comp.ai.philosophy,sci.logic
Path: cantaloupe.srv.cs.cmu.edu!nntp.club.cc.cmu.edu!miner.usbm.gov!rsg1.er.usgs.gov!stc06.ctd.ornl.gov!fnnews.fnal.gov!uwm.edu!spool.mu.edu!bloom-beacon.mit.edu!news!minsky
From: minsky@media.mit.edu (Marvin Minsky)
Subject: Predictable Turing Machines
Message-ID: <1995Jul5.204427.20915@media.mit.edu>
Sender: news@media.mit.edu (USENET News System)
Cc: minsky
Organization: MIT Media Laboratory
References: <3tdo39$4f7@wsinis10.win.tue.nl> <3teg8l$nb5@cismsun.univ-lyon1.fr> <3teh4p$n03@netnews.upenn.edu>
Date: Wed, 5 Jul 1995 20:44:27 GMT
Lines: 42
Xref: glinda.oz.cs.cmu.edu comp.ai.philosophy:29630 sci.logic:11917

In article <3teh4p$n03@netnews.upenn.edu> weemba@sagi.wistar.upenn.edu (Matthew P Wiener) writes:
>In article <3teg8l$nb5@cismsun.univ-lyon1.fr>, Claude Chaunier <cchaunie@bat710 writes:
>>Interactivity means also that when someone asks Neil something, he is
>>already increasing his knowledge, in particular when someone asks him
>>twice the same thing, he does not give twice exactly the same answer.
>
>Sure he does.  Two times zero is zero.
>
>>And I surely want a computer like Neil.
>
>Yes.  A suffusion of yellow is just about your speed.
>-- 
>-Matthew P Wiener (weemba@sagi.wistar.upenn.edu)

Really, why don't the two of you get off the net until you've got
something worth the larger group's attention.  In particular the
discussion of Turing machines isn't worth the rancor:


If we have a Turing machine with inputs from a computable world, then
we can make a Turing machine that does the same thing without the
world, by simulating it the world in the machine.  

If the world has a noncomputable aspect, then the Turing machine with
inputs can output a non-r.e. set.  However, there's no way, so far as
I can see, to ever know if this is the case.

If the *timing* of the world's inputs is noncomputable with respect to
the Turing machine's speed, then again you get noncomputability, but
still cannot know it.

If the world also includes a *random* aspect, then the great theorem
of Shannon, Shapiro, DeLeeuw, and Moore applies: the computablity of
the output depends, with probability 1 in a somewhat complicated
sense, on whether the "p" of the coin-toss is itself a computable real
number.

I doubt that any of this relates to cognition or philosophy much,
because the noise in any big world will make all computations subject
to practical unpredictabilities, regardless of whether they are, in
principle, recursively predictable.

