X-Message-Number: 14965
From: <>
Date: Mon, 20 Nov 2000 12:24:51 +0100 (CET)
Subject: Turing and Pi

 > Message #14956
 > From: 
 > Date: Sun, 19 Nov 2000 13:26:55 EST
 > Subject: Turing and Pi
 > 
 > If YEGO you will skip this--it's yet another effort to open the eyes of the 
 > uploaders.
 
This better be good.
 
 > Can a Turing computer (Universal Turing Machine) simulate or emulate a 
 > sentient (feeling) being?
 
The real question is, of course, can a specific discrete physical
system, a computer, emulate relevant aspects of another physical
systems, including biological ones?


 > The Turing computer is digital, sequential, algorithmic, and finite (although
  > not necessarily bounded). It has been shown--and in fact I think it is 
 > self-evident to those with the background--that a Turing computer can "do 
 > anything" that any digital computer can do--i.e., it could (eventually, in 
 > principle) produce any desired number, or set of numbers, to any desired 
 > degree of accuracy.
 
There is no such thing as a Turing computer, there is only a Turing
machine. You can read it up here:

	 http://plato.stanford.edu/entries/turing-machine/
	 http://www.turing.org.uk/turing/scrapbook/machine.html

You're probably thinking about the von Neumann computer
architecture. Which is irrelevant in this context, since adding
unnecessary contraints on the shape of the emulation device. In fact
we can rule out use of a von Neumann (or Harvard) architecture for
practical reasons.

 > The memory store can grow without limit, and the program can change also, 

 > either by self-modification or by allowing random influences such as nuclear
  > radiation or cosmic rays. Or, if preferred, one could rule out random 
 > influences or/and some types of self modification.
 
You're wrong. I recommend you reread the definition of Turing machine.

 > The "emulation" supposedly consists of eventual production, from time to 

 > time, of sets of numbers corresponding (within uncertainty principle limits)
 
 > to the quantum states of a person and (part of) his environment and history,
  > past or/and future.
 
Once again you're choosing an arbitrary paradigm from the huge set of
possible ones. Simulating a biological system at quantum level of
theory is a ridiculous overkill. I also very much doubt the gating of
an individual ion channel somewhere bears much relevance to your
thinking process. We're somewhat nearer the required detail level when
we talk about compartmental modelling.

 > The uploaders believe that if the algorithm or program, with the initial 
 > store, corresponds closely enough to the laws of nature and some set of 
 > initial conditions, then there will be (closely enough) a one-to-one 
 > correspondence or isomorphism between the physical actuality and the 

 > computer-generated sets of numbers, and that "therefore" there will be a live
  > emulation. 
 
You seem to be on a vis vitalis trip. Life is a just another physical
process. We can simulate physical processes at atomic level today,
it's called computational physics and computational chemistry. We can
also simulate the electrochemical behaviour of biological neurons on a
number of scales, from molecular to mesoscale. This is called neuronal
model(l)ing.

This is not about "belief". This is the state of the art in
computational sciences. You seem to have skipped a couple of decades
of progress in the field.

 > The built-in assumption is that nothing matters except patterns of 

 > information and the processing of information (or time evolution of patterns
  > of information). This assumption has a mild degree of plausibility, but is 

 > not self-evident and not proven true. But problems remain even if one accept
  > some version of the assumption.
 
Sure there are problems: data acquision as well as software and
hardware engineering problems. The question of scanning a critter at
sufficient resolution, process and and store the data set. The problem
of bootstrapping a hybrid learning emulator, which derives knowledge
not from manual input of magic fudge factors, but automatically learns
by reaching down to a deeper level of theory, and can build higher
order layers by abstraction. The problem of creating this
functionality, and scaling up to larger organisms.

Notice that none of this requires any new physics. It's "just" technology.

 > As I have pointed out before, ALMOST ALL of the numbers generated in the 
 > computer represent intermediate calculations or operations of one sort or 
 > another. The sets corresponding to sequential quantum states of the system 
 > are distinguished in two ways, at most. One way is by arbitrary labels. 
 > (Searle might claim that is the only way.) The other is by necessary and 
 > sufficient relations implied by the laws of nature, and I think this is 

 > possible--i.e., I think it is possible that one can find semantics inside the
  > symbols, without any Rosetta Stone. But that still leaves--at minimum--two 
 > problems.
 
I don't understand what you're saying here, and I have a strong
suspicion that you don't, either.
 

 > The first problem, once again, is binding of space and time. FEELING may (and
 
 > I think probably does) demand a REGION of space and time (possibly some kind
  > of standing wave, the self-circuit), not a point or quantum cell. In such 
 > case, isomorphism will NOT do the trick.
 
1) Please restate this formally. As is, it means everything, and hence
   nothing.

2) Please provide evidence (facts, like) for your position.


 > Second is the problem Mike Perry has mentioned, the presumptive existence of
  > hidden sets either in the world of physics or the world of Plato or of 

 > mathematics. If you expand pi in some base, for example, whether base two or
 
Uh, evidence?

 > some prime number, you may eventually get a segment of digits corresponding 

 > nearly enough to some person and a segment of his environment and history. If
 
 > that segment is written down in some fashion, does or did or will the person
  > exist and feel? Will or does he exist or feel even if the numbers are never 
 > written down? Never discovered? 
 
Ignoring the obvious mysticism of objective existance of a formal
system in absence of observers (slabs of basalt Plato do not read), I
presume you're talking about the difference of creating a trajectory
of examining an existing, previously computed trajectory. Is a
simulated critter alive in the frames dumped to the storage device?

That's an interesting question, for a change. I have some thoughts but
no ready answer for that one.


 > At minimum, I think, these considerations cast serious doubt on the uploading
 
 > thesis. In my opinion, they put it heavily on the defensive, making it only a
  > long shot.

Look, you'r objections are already nuked by some of state of the
art. Right now people are writing cell simulators and neuronal
codes. In 15-20 years we should be able to make an individually
accurate computational model of primitive critters, e.g. a
nematode, maybe even something as smart as aplysia or even a lamprey.

I wish you would live to see such work, and I would love to hear you
explaining away a virtual critter showing chemotaxis or previously
learned labyrinth navigation.

We can make urea without requiring a kidney, or animal, whether man or
dog, these days. Your objections run along the same vein.

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=14965