|
Hi Steve,
I think the situation here may simply be a misunderstanding
between the possibility of "creating life" in a computer program and "simulating
life" in a computer program. His view was that it was not possible to create a
life form within that limited realm, but he would never say simulation of a
life form was impossible (quite the contrary). His caution would be to
always remember the difference between a simulation and the "real
thing".
Ultimately, he was saying: If we don't understand the real thing,
but find a way to simulate it... how is that simulation supposed to teach
us what is really going on in the real thing? This was the crux of
his concern on the subject of using computer simulation in this area
of science. So often, science wants to replace the natural system with something
easier to comprehend, which is all well and good, that is-- until it is
forgotten that the replacement is NOT the same as the natural system. The
"machine metaphor" is one example of just this kind of
"forgetting".
On the subject of where all entailment in a computer comes from, I
wonder if he wouldn't say that it all comes from human beings. The CPU is a
human creation and is based on part of our own mental thought capability. The
entailment in a computer, therefore, originates outside of the
CPU.
Judith
----- Original Message -----
Sent: Thursday, August 19, 2004 1:54
PM
Subject: [ROSEN] Question on
Non-simulability
One of the recurring themes in Rosen's argument is that organisms possess
non-simulable models. He thus argues that computer simulation is not a
fruitful way to investigate the "what is life" questions.
His argument about non-simulability centers around the impossibility to
simulate causal impredicative loops by a computer because a computer program
is ultimately a finite state machine.
Could someone on the list point out what is wrong with the following
objection to Rosen's argument? I'm aware that my understanding is limited so
I'm just asking for may education.
Question.
It seems that by program Rosen mostly means a fixed set of
instructions operating on some inputs and generating some outputs (turing
machine style). In this sense I would agree that simulations are limited
because a program cannot modify itself. However, that is too narrow a
view of what a computer can do. Many modern computer languages
incorporate meta-language facilities that allow the program to
dynamically modify itself and even dynamically modify the language (e.g. LISP
meta object protocol).
Rosen's answer to that seems to be that, regardless of what the
software does, it has no entailment - all entailment is in the CPU. I do not
understand this argument. We can certainly take a CPU that supposedly has
entailment and completely simulate it in a software of totally different
system. The resulting simulated CPU is indistinguishable from the original,
yet, according to Rosen it has no more entailment. Where did it go? It seems
like it is now embodied in the simulator software.
Secondly, it seems like Rosen basically argues that since CPU
is a simple state machine that anything that can be simulated in
software is ultimately a state-machine. This argument seems to be
similar to reductionist reasoning that Rosen often crititicizes.
It's somewhat analogous to saying that since organism consists of molecules
all it's propreties derive from the molecule. We know that it's not the
case for the organism because a lot of its properties are embodied in its
organization. Why must this not be the case for the computer where the
software embodies organization? For example, I do not understand why
certain software structures cannot represent impredicativites. Seems like it's
done on a regular basis. For instance, various control systems incorporate
impredicativities of all kinds in the form of feedback systems.
Thanks,
- Steve
Do you Yahoo!?
Yahoo! Mail is new and improved - Check
it out!
|