Newsgroups: comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!gatech!swrinde!emory!mind.org!runic!thantos
From: thantos@runic.mind.org (Alexander Williams)
Subject: Re: ALife Park
Organization: Runic Writings UUCP Link: Convoco Hasturam
Message-ID: <D8qMBK.LH@runic.mind.org>
References: <3o61j0$cbp@ns.cityscape.co.uk> <3otcdi$abc@gap.cco.caltech.edu> <D8IyHB.4pA@runic.mind.org> <3p5kug$o1q@gap.cco.caltech.edu>
Date: Wed, 17 May 1995 19:33:18 GMT
Lines: 162

In an arcane scroll, Alexander Williams quotes the holy scripturist
C. Titus Brown, replying to the mystic words as written, saying:

>>   I think I have more faith in the "physics" of the Tierra
>>simulation code than in RL physics.  The latter tends to be fractal
>>in complexity, confusiong, and contradictory at different scales.
>
>(Just to let you know, tierran physics may produce a fractal fitness
>landscape; I can give you the one reference on this, but no-ones really
>done much work on it (yet)).

   The fitness landscape itself isn't what I'm addressing, but the
actual underlying "physics" that the Tierran organisms exist within,
their assembly language, in short.  I'm sure, nay, certain that
given a sufficently open definition of fractal, the fitness
landscape of Tierra is fractal.  I've seen discussions and
mathmatical proofs that the edges of oak tree leaves describe a
fractal, which then go on to demonstrate, etc.  "Fractal" has become
almost a throwaway term in the sense.
   We have full knowledge of the Tierran architecture, instruction
set and how they interact with the actual physical architecture
which underlie them.

>Obviously you've never done network programming ;).

   Actually, I /have/, but I leave the messy internals to the real
system buffs.  :)  Of course, to work on a serious MUD, you /have/
to be able to grovel through some of the system network stuff, at
least in general.

><sigh> No, not really.  In theory, we possess the full knowledge.  However,
>net.tierra runs compiled for a particular platform, which runs a particular
>OS, a particular compiler, a particular type of CPU & assembly language,
>across a particular type of Ethernet using a particular brand of the IP
>protocol.  You figure it out...

   In actuality, we /do/ possess full knowledge.  We have (or will
have) a technical spec that lists the hooks into the generalized
Ethernet drivers, etc etc etc.  We have a precise and concise
listing of the instruction itself, which the Tierran organisms will
be running.
   Only the knowledge of that /last/ is really necessary to state
that the Tierrans won't be a problem, in fact.  If part of the
network code pretends to be Doom ][ and sprays broadcast packets all
over the net, that's a problem with the networking code, not
something the Tierrans can access directly.  If you've questions
about the networkability of Tierra, the resources it consumes and
limiting security penetrations, those are concerns about the Tierra
security/networking code and limitations, /not/ the Tierran
organisms evolved within a precisely delimited architecture.
   Despite the number of generations the Tierrans and kin have been
running on non-networked systems, they've never possessed the
screen or taken over a hard drive of themselves.  Why does the
addition of another facet to the program cause so much FUD?

>You're confusing the issue again: can a virus that purports to do what the
>Good Times virus does exist?  No.  However, people, in their overconfidence,
>have now stated that NO e-mail virus can exist.  This is clearly not true,
>at least to me, and could become an issue eventually.

   /I've/ never stated NO email virus can exist; one most certainly
did for a while in the heyday of BBSs, an ANSI-bomb.  No experts
have said NO email virus can exist, what they /have/ said is that
it'd be such a limited target/limited application program to
construct that it wouldn't be worth the effort to create, much less
propogate through the highly-limited vectors it could use to travel.
   Part of the reason its never /going/ to be an issue is
"diversity," the same thing that keeps entire ALife populations from
dying out/converging prematurely/what-have-you.  The number of email
systems out there using variants that'd break a finely tuned virus
is non-trivial.  The virus, thus, has to be tuned to attack a
trivial number of targets, thus, dies under its own mass before it
starts.
   [You'll notice I hedged on the question of MIME-systems earlier,
and for a good reason: if it becomes a global standard for execution
and encapsulation, it becomes a widely-displaced vector, like air
for the common cold.  I'm reasonably sure the MIME developers
recognize this, though, and don't allow outside execute permissions
without jumping through hoops locally, etc...]

>Spreading disinformation is bad.  Spreading too much information is also
>bad.  If you simply say "no, the good times virus cannot exist", that's
>enough.  If you then add on "And no e-mail virus can exist either", that's
>wrong, and can lead to dangerous overconfidence (as well as people ignoring
>future reports of an e-mail virus that COULD affect them, simply because
>computer experts told them it couldn't).  Need I continue?

   Of course not, but you're building a straw man.  Neither I nor
the "experts," (pick your fave) have stated catagorically that one
/cannot/ exist, but the odds are vanishingly small.  I /can/
estimate those odds to within a hair's breadth of surety, mainly
because I'm pretty familliar with email /reader/ systems, etc, and
know what'd be necessary for it to work across a group of terminals,
much less readers.
   Spreading the disinformation that overestimates the liklihood of
the existance of such a virus is /worse/, in my opinion, because of
the "Cry Wolf Syndrome."  You start getting so many scares that if
one really /is/ found in the wild, you won't be able to see the
signal for the noise.  If its commonly held to be "vanishingly
unlikely" in the general mind, then a single report that hinges on
"what remains, however improbable, must be true," it matters and
draws attention.

>I'm assuming a bit of hyperbole on your part (yes, I've done my best to
>avoid configuring sendmail; try smail instead ;), but you're still just

   I needed sendmail's power...  Go fig.  I'm a Tim Allen fan,
too...

>plain wrong.  You're confusing the two issues: how complex is it to 
>understand the foundation vs. how complex it is to understand the eventual
>process it undertakes/undergoes.  When I say that no-one understand tierra
>completely, I'm talking about the latter; you're talking about the former
>when you say that it IS possible to understand.

   You don't /need/ to know the details of the process if you know,
a priori, that the system is incapable of reaching a certain state,
constrained by the underlying implimentation.  Tierra is a
reasonably complex system, but not so much so its completely beyond
understanding.  In fact, its pretty comprehensible within its domain
once you wrap your mind around a few alien concepts.  Sendmail, on
the other hand...
   [OK, slight hyperbole.  Very slight.]

>The whole point of complex systems is that you get a rich set of interactions
>from a very simple foundation.  The whole point of researching them is to
>understand the general guidelines for the creation of this set of interactions.
>Almost by definition, tierra as a complex system is too complex to do anything
>but run in order to characterize the exact final state.

   Very true, but you can't start importing your creations mystical
abilities beyond their basic architecture just because they
incorporate "complex systems."  Then the whole meaning of the term
goes right out the window to join "fractal," "chaos theory" and a
host of other terms that were watered down beyond the point of
/having/ a meaning.
   The Tierran architecture and instruction set is /just/ complex
enough to put the whole system into an area of possibility where
complex behaviours can arise.  (Langton uses the lambda value as a
measure of that quality, as I recall.)  The system as a whole
describes a deterministic set of rules that are put together in
different ways by entities within it, blah blah blah as has been
sais a hundred times.  HOWEVER, that does not invalidate our a
priori knowledge of the system and instruction set and using that
knowledge to know where the system is constrained.
   We may not can say with exact precision what it /will/ do, but we
can say with complete certainty what it /won't/ and /can't/ do.

>I realize I'm starting to sound like a broken record, and, at this point
>the thread isn't really contributing anything.  I'd be happy to continue
>discussing this via e-mail...

   Pift, its a welcome departure from the interminable discussions
of human consciousness and the continuous flow of posts announcing
conferences I couldn't attend even if I /was/ university funded...
:)

-- 
thantos@runic.mind.org (Alexander Williams)     | PGP 2.6 key avail
  Should we shed our mental pants and compare   | DF 22 16 CE CA 7F
  the size of our consciousnesses?              | 98 47 13 EE 8E EC
      Jan Sand to Marvin Minsky                 | 9C 2D 9B 9B
