X-Message-Number: 8640
Date: Mon, 29 Sep 1997 20:52:21 -0700 (PDT)
From: John K Clark <>
Subject: Turing Tests and Machines
-----BEGIN PGP SIGNED MESSAGE-----
In #8632 On Sun, 28 Sep 1997 Peter Merel <> Wrote:
>you might determine that something is not intelligent, but there's
>no way to test that it actually is intelligent.
I think the opposite is true. If something passed my Turing Test I would not
hesitate in calling it intelligent, but if it failed the test I would
hesitate in calling it unintelligent.
>This is entirely reasonable: since any problem may be transformed
>into the question of whether or not a particular program will halt,
>intelligence is always necessarily limited or else it could solve
>the general halting problem - which is demonstrably impossible.
There are some true things that can not be proved to be true in a finite
number of steps, not by intelligence and not by anything else, so I don't see
how that helps us distinguish between intelligence and its lack.
>One process is less intelligent than another with respect to some
>particular game iff it loses that game more often than the other.
Sounds reasonable to me.
>intelligence as an absolute criterion appears to mean nothing at all.
I have a hunch there is more than a little truth in what you say.
In #8633 Thomas Donaldson <> On Sun, 28 Sep 1997 Wrote:
>Some time ago, when I mentioned that some people in computer science
>had discussed variations of Turing machines which used real numbers
>(I DON'T mean floating point numbers, I mean real numbers ie.
>sequences of integers potentially going off to infinity to the right
>but not to the left) [...] The reference is by HT Siegelmann, ED
>Sontag, SCIENCE (268(1995) 545-548). In the particular article given,
>they discussed neural nets and how their results with such real
>number computers led to conclusions about neural nets also (their
>article naturally refers to previous work on such computers). It
>turns out that a neural net with real-number weights on each
>connection cannot be imitated by a Turing machine. FULL STOP.
It's been a while but I think I remember that article, even though it wasn't
very memorable. Sure their gadget would leave a Turing Machine in the dust,
that's not surprising, just one of those real number weights would contain,
not an astronomical, but an infinite amount of information. What would have
been surprising, absolutely flabbergasting in fact, is if they had given some
idea about how such a machine could actually be built, but of course they did
not. I'm not whining about technological difficulties and I don't demand they
produce a working device, all I want is a hint about how the thing could
possibly operate without violating the known laws of physics. If you consider
mathematics only and are willing to invoke infinity then it's easy to make a
paper machine do anything you want, only one little thing stopping you from
actually building it, physical reality.
>In each case the infinite real number isn't used directly; instead,
>as part of the computation it is calculated out to the required
>number of decimal places for the computation the machine is
>performing.
Right, and guess how many decimal places it would need to do something a
Turing Machine could not. Give me a number, any number at all, I don't care
how big it is. Sorry, that's too small.
John K Clark
-----BEGIN PGP SIGNATURE-----
Version: 2.6.i
iQCyAgUBNDByv303wfSpid95AQEWMQTlHLKBuUF89bucjRXL/dbfhN6vw6TrbOsV
hxVl/UVFCAw8ErgvOUoRm0zIkr74nWJ3M7wcz3nPDihNYVJQIB7dachaezqoVAK4
frY/tOYreTcDzTxWzyx8Zp/WhuEDt/RhwJLHwdQ9oo4oz24H50IF9GXbdj8S7FK8
qwsGL78hrWWxsLumprcE5qf0vftmLCCZl7xrnTsXSeVtkd267w==WMpj
-----END PGP SIGNATURE-----
Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8640