A possible definition of life.

James Foster foster at skink.cs.uidaho.edu
Mon Apr 21 11:04:12 EST 1997


In article <335B3215.45597B97 at house.med.und.ac.za> Rob Miller <rmiller at house.med.und.ac.za> writes:


   James Foster wrote:
   > 
   > In article <m0wHUt6-0004OBC at uctmail2.uct.ac.za> ed at MOLBIOL.UCT.AC.ZA ("Ed Rybicki") writes:
   > 
   >     "Life (anywhere) is the phenomenon associated with th replication of
   >    self-coding informational systems".
   > 
   > This is too imprecise.  To which phenomena do you refer?  Replication?
   > Self-coding?  Decoding?  Translation?  Does your definition make my
   > copy program alive?  How about my photocopier?
   > 

   Seems pretty precise to me -- Ed's statement implies to me the
   phenomen*on* (singular) which is specifically the existence and
   function of `self-coding informational systems.'  Your photocopier
   is clearly not self-coding, and your copy program is not
   sufficiently self-coding to replicate on its own (without more
   assistance than the `natural phenomena' occurring on your hard
   drive - like a computer virus could get by with).

I'm beginning to be a devil's advocate here, using absurd arguments
only for illustrative purposes, but...

Suppose you put the wiring diagram into the copier.  Now the diagram
and the copier together are a "self-coding informational system" and
the phenomenon associated with what the machine is doing is life.  

I meant phenomena (plural) because there are several things happening
during replication, and Ed's definition fails to distinguish any of
them.

   > I remember a philosophy paper in college arguing that thermostats were
   > alive: they are self regulating, have internal representations, must
   > be in a host, etc.

   But they are not self-coding.

But they make the thermal properties of the house "self-coding".

   > Personally, I think defining life is a fool's errand.  We're not

   Or perhaps a hobby for net-nerds like us :-)

Ouch, and touche.

   > really interested in "what is life".  Ed is interested in how viruses
   > do what they do.  Alife people are interested in algorithms which
   > display interesting behaviors.  You don't have to define the term to
   > justify what you do, or even as a prerequisite to doing it.

   I think Ed's interested in the question on a lot of levels, as am
   I.  Alife gives me insights into models of protein folding, and I
   believe a lot of Alife `experts' would be at least annoyed by your
   justification of their field as an opportunity to `display
   interesting behaviors.'  Just as many AI folk believe they are
   gaining insight into human intelligence with their work, defining
   (and modeling) life helps us understand what constitutes life, how
   it works, and what we really mean by the term.

Anything that gives you insight is good.  But that doesn't make it
correct.

And I've annoyed the alife "experts" before.  I think that most of
what the alife folks do is exactly what I describe.  They claim to be
trying to gain insight into other phenomena (and they are trying to do
that) but their methodology is to write algorithms which display
interesting behaviors.  "Hey, look, that looks like a swarm." 

You could also gain insight into phenomena by a careful study of the
Talmud, or by shuffling "idea cards" and looking for patterns in the
arrangements you find.  You could get genuine, and even correct,
insights.  But the process is only a tool for creative thinking, since
it's association with the subject matter is tenuous.  

Again, the insights may be genuine, and the conclusions correct.

You need to get away from your creative tools to test your hypothesis
with real systems in order for your work to be of genuine scientific
merit.  Most alife work fails to do this.

...I remain unconvinced that you need to define life in order to
understand it.  The same is true in other disciplines.  You don't need
to define matter in order to do physics.  You don't need to define
energy in order to do chemistry.  The mode of operation where you
begin with definitions seems to work best (and perhaps only) in
disciplines like CS and mathematics (and philosophy)...where we're
investigating consequences of definitions.
-- 
James A. Foster			email: foster at cs.uidaho.edu
Laboratory for Applied Logic	Dept. of Computer Science
University of Idaho		http://www.cs.uidaho.edu/~foster

pgp key available at: ftp://ftp.cs.uidaho.edu/pub/foster/pgp-key.asc



More information about the Mol-evol mailing list