Path: utzoo!utgpu!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!rutgers!rochester!pt.cs.cmu.edu!andrew.cmu.edu!ap1i+
From: ap1i+@andrew.cmu.edu (Andrew C. Plotkin)
Newsgroups: comp.ai
Subject: Re: Artificial Intelligence and Intelligence
Message-ID: 
Date: 4 Dec 88 17:30:58 GMT
References: <484@soleil.UUCP> <0XTukNy00Xol41W1Ui@andrew.cmu.edu> <42328@linus.UUCP>,
	<1069@microsoft.UUCP>
Organization: Carnegie Mellon
Lines: 34
In-Reply-To: <1069@microsoft.UUCP>

/In article <42328@linus.UUCP> bwk@mbunix (Kort) writes:
/>Andrew C. Plotkin writes:
/> >I maintain that a human can be simulated by a Turing machine.  Comments?
/>
/> ... I use a quantum amplifier in my coin flipper.
/>Correct me if I'm wrong.  But a Turing Machine is obliged to follow
/>a deterministic program.  Hence a Turing machine cannot simulate my
/>dice-tossing method.

chrispi@microsoft.UUCP (Chris Pirih) replies...

/Nothing prevents the Turing machine from flipping a coin and acting
/[deterministically] on the [random[?]] result.  (What exactly are we
/trying to simulate here, Barry?  Is the coin, with its quantum
/randomness, a part of the human who consults it?)  Besides, is it
/necessary that a simulated coin-flip be "truly" random, or just
/effectively unpredictable (i.e., that the Turing android eschew
/foreknowledge of its pseudo-random pseudo-coin-flip)?

I agree with Chris here. When I flip a coin in my head, I have serious doubts
that the results are truly random (based on an amplified quantum randomness.) It
might just as well be a complex pseudo-random generator. It would work just as
well for any practical purpose, such as the balanced ethical dilemma (sp?) you
mentioned.

(Practical purpose meaning anything where you need a single random bit. If you
try to spit out a long string of random bits, the non-randomness of the process
becomes painfully clear -- there is a strong correlation between each bit and
the several-bit sequence that precedes it. This can be checked with a simple
computer program (on each bit, using the previous four or five bits to predict
the next). I've found it a 60% to 70% accurate predictor; no doubt a more
complex program could do better.)

--Z