[ExI] Next moment, everything around you will probably change

Lee Corbin lcorbin at rawbw.com
Fri Jun 22 07:08:21 UTC 2007

Jef writes

>> But as I said, the similarity metric says that they're the same person,
>> and a wife of one wouldn't tell the difference between the two, and
>> so on. In other words, they seem in all ways to be the same person.
>> They just hate each other is all.  (And that's hardly novel:  we often
>> wonder if a given individual "hates himself" in some way.)
> You appear here to neglect your own belief and arguments that physical
> substrate doesn't matter at all (I completely agree, of course), and
> that what matters is function, defined of course by the agent's
> physical structure (and by something else of which you appear
> consistently unaware.)

Here, yes, I did happen to use the "similarity metric" to deem two
biological machines (duplicates) to be the same individual. But of
course, were one uploaded and one not, the similarity metric would
still say that they're the same person, provided that they continued
to have the same memories (the most important thing), and (as a
result, usually) have the same behavior.

>  How correct can it be to refer to separate
> instances as being the same person on the basis of their almost exact
> physical similarity, when their functioning recognizes the separate
> existence of the other, so as to hate, compete with or even destroy
> the other?  (Or even to cooperate.)

Their "functioning" is much broader, I would say, than you are
suggesting. It includes all their behavior (which, importantly, is
greatly a reflection of their memories). The *only* bit of 
functioning that is peculiar is that they are fighting, e.g., trying
to eliminate each other.  Two modified Tit-For-Tat programs
are "out to get" each other, yet manifestly they are the same
program. Hey, that's just the way they're built. It's what they
do.  A human retard that was trained to do nothing but box
would lash out at any other remotely human figure, including,
of course, a duplicate.  So it seems to me that this antagonism
towards each other is a very tiny, tiny part of their behavior.

> There's a very practical point to this philosophizing (and it's not
> about personal survival.)  I've stated it twice now, even highlighted
> it with "---------------", and twice you've deleted it without
> comment:
> ------------------------
> <fnord>
>>>... any degree of selfishness will tend to put
>>> duplicates at odds with one another as they interact from within
>>> increasingly disparate contexts.
> </fnord>

Okay---True, any degree of selfishness will tend to put duplicates
at odds with one another.  So what?  What was I supposed to 
react to?  Hmm, well, maybe this:  recall that according to my
beliefs it is inherantly selfish for an instance of me to kill itself
immediately so that its duplicate gets $10M, given that one of
them has to die and if the instance protects itself, then it does
not get the $10M.  That may seem an unsual way to use the
word "selfish", but I mean it most literally.  Were "I" and my
duplicate in such a situation, "this instance" would gladly kill
itself because it would (rightly, I claim) anticipate awakening
the next morning $10M richer.  Now that's *selfish*, no?  :-)

> If we can get past the polemics we can consider the more interesting
> (in a practical sense) issues of systems of competition and
> cooperation, which is necessarily between **agents**.

Roles?  By agency, do you mean roles that different actors play?
A human being in a policeman's uniform plays the role of a policeman,
and it has occurred to me that maybe this is the kind of thing that
you mean by agency.  If so, I would retort that the real human being
behind the badge is vastly more complicated than the comparatively
simple role that it's playing.

>> Okay, since you have been so kind to cut and paste it again, I
>> will try to answer it as directly as I can.  I *don't* see those
>> as two separate individuals at all.  Neither would the people
>> who know them.
> People have no difficulty with biological twins being different
> persons, despite very high physical/functional similarity.  You would
> probably like to say this is because each twin has different memories,
> etc., but consider that other people can't see memories, etc., what
> they see is separate agency.

Actually, I do believe that identical twins are recognized by
their *differences* as seen by friends, parents, and so on, 
although I admit that a decade or so I was startled to learn
something about identical twins.  I conjecture that if you and
I conspired to create a duplicate of you, and then you 
introduced the duplicate to your family as your "long lost
twin brother", they'd soon come to be very suspicious,
especially if they've known other identical twins.  By Jove,
they simply would be UNABLE to tell any difference, and
would never get used to that.

>> But I say that they are simply mistaken:  it's as though
>> each has been programmed by nature to regard anything
>> outside its own skin as "the other" or as "alien".  I mean, we
>> could have to *totally* identical instances of the Tit-For-Tat
>> program playing each other (or rather a minor variation of
>> Tit-For-Tat that tried a random defection now and then),
>> and they naturally behave as though they are going up against
>> "the other", "the alien", the "other player".  Yet they are truly
>> identical, right down to the last statement of code.
> You're arguing circularly again, assuming your own conclusion.

But you are now going to claim that these two identical programs
are not the same program?  Or not the same..... what?  Sure, 
they're not the same instance OF THE SAME PROGRAM.

> This topic gets more interesting when we get past this (temporary?)
> impasse and consider the application of artificial agents of arbitrary
> physical/functional similarity fully dedicated to acting on behalf of
> a single entity -- variously enabled/limited instances of exactly the
> same self.

Yes.  Do go on.  You mean, again, as in two entities completely
alike in behavior except that each has an agenda that includes only
benefit for its own instance?   I guess that I just don't see the
centrality of this one aspect.  Admittedly, the "agenda" of an
organism to requisition all benefit to its own instance is pretty
common throughout the animal kingdom (including many humans).

A possible argument against your position goes like this. Suppose
that A and B are totally identical (structurally and so satisfy the
similarity criterion, same memories, etc.) but happen *not* to be
in each other's vicinity.  If they live on different planets, then they
cannot come into conflict.  At this point, doesn't your "conflict
criterion" fail to be applicable?   Yes, each is still requisitioning
benefit for its own instance, but now their agency seems the
same too.  Why are they different persons, unless all this time
you've merely meant by persons what I've been calling "instances"?

>> > Or maybe simpler for you, consider the two duplicates, each with
>> > identical intent to prevent the existence of the other. If, as you
>> > say, physical/functional similarity determines personal identity, then
>> > do you see the paradox entailed in a person trying to destroy himself
>> > so he can enjoy being himself?
>> I admit that there is irony in the situation of a person or program trying
>> to destroy instances that are identical to itself,
> Actually:
> Paradoxical if you insist that they **must be** the same person.

What paradox?  As I mentioned to Stathis (I think), it hardly
rises to the "paradoxical".  Two identical chess programs can
also fight it out,  Fritz 9.1 vs. Fritz 9.1.  So what's new? 

> Why should outside observers see two competing twins, no matter how
> physically/functionally similar, as the same person?

Because they act the same way, have the same memories, think
the same way, look the same way, and are indistinguishable
except for location?  It's not a fact that people will regard each
duplicate as the same person in terms of everything but location
(or perhaps recent number of parking tickets)?  In all the 
important ways that matter, people will regard them as the
same person.

> Should they
> treat the offensive software engineer Lee exactly the same as the
> defensive chess playing Lee? Were talking about separate agencies here
> and **for all** practical purposes, and **for all** observers
> (including these agents themselves), they are separate persons.

Well, okay, maybe I'm starting to see your point. No, they will not
treat them the same.  Indeed, we are treated differently every day
by people depending on what role we are happening to play. But
that hardly makes us different people during the day, or do you
think that it does?  

> Even to get them to cooperate, which should be the desired outcome,
> they must necessarily see themselves first as independent agents, and
> then as the same person only to the extent that they are seen as
> representing a single (abstract) entity known as Lee.

The same argument could be applied to me at different times of
the day?  Why should the agent Lee-8:30-am prepare a sandwich
for the Lee-12:30-pm, even though the former is not hungry?
Why do I  and my boss and everyone in the world regard
Lee-12:30-pm as really the same person as Lee-8:30-am
unless it's really so?   Our language and the terms it uses and
the concepts it refers to has come to *mean* precisely by
"they're the same person" all our usual assocations to the fact.

>> > Or back to the biological organism manifesting Disassociative Identity
>> > Disorder. In such a case we can agree that the
>> > physical/functional similarity is total since it's only a single
>> > organism, but we also agree that that any observer (including the
>> > observers manifested by that particular organism) will see different
>> > persons to the extent that they are perceived to act on behalf of
>> > different entities.
>> Hmm, well, we seem to have a hard disagreement here. Yes, let's
>> consider just the case we/I have been discussing:  indeed there
>> are many people who would hate their duplicates.
> Are you evading here the case of the biological organism manifesting
> DID, or are you conflating with the case of the duplicates?

Not sure.  I'll try to cover both bases.  The DID manifesting organism
will have differing memories and differing behavior, and that's what tells
people that it's really two people.  Notice that this is of a much deeper
and more substantial kind than mere different agency, e.g., the way
a policeman behaves in uniform and later at night the way he behaves
at a restaurant.  (Yet no one would claim that he was not the same

The duplicates, if made quite recently, are indisputably the same person
in the eyes of people who interact with them individually, and who are
not clued into the existence of a matter-duplication device.  All they'll
ever complain about are things like "Hmm, this morning you said the
same thing, remember?", or "Hmm, this morning you said you were 
going to be out of town all day!", etc.   It would *never* occur to
them that these were separate people.

>> So let's suppose
>> that A and A' are identical, and so---just as you say---they are
>> what you call "different persons" because they are perceived as
>> acting on behalf of different entities. Clearly here, they are acting
>> on behalf of different *instances* of a what was a single person.
>> You and I each beg the question in a different way.  You beg the
>> question by saying that they are clearly different entities, and so
>> are different people, and I say that (because of similarity) they
>> are clearly the same person (or program).  How may we resolve
>> this?
>> Well, as above, I suggest that we consult outside authorities of
>> higher reputation.
> What?  Appeal to authority -- on the Extropy list?!


>> If we send them into different rooms, can
>> someone who knows them well tell them apart?  (I say no.)
> Certainly they can be distinguished by someone who knows them well.
> One of them goes on and on about how "this shouldn't be happening,
> it's all just a deep mistake, how can I be so confused as to attack
> myself like this, I just wanted to keep my software job and also get
> to play more chess, there's no reason to be upset, I know beyond
> logical doubt that my other instance should actually be anticipating
> our increased pleasure, I know I'm a very reasonable person."  The
> other keeps saying "of course this was bound to happen, I see it
> clearly now even though I denied it when Jef tried to explain, I'm
> depressed and burnt out with writing software and if things don't
> change I'm going to do something...drastic."

You haven't accounted for the---to me---extremely bizarre and
unlikely eventuality that two duplicates of me would act so differently.
They'd never attack each other.  They both would be very "reasonable"
especially to each other.  Or did you introduce a brain lesion in one
of them or something?  (Sorry, I've forgotten.) 

And anyone who knows them well would *not* think that they were
different people, unless he was clued in on the amazing new duplication
machinery.  At very worst, people would consider me moody, being
in one mood at one time and another at another time.  Unless the mood
swings were so extreme as to constitute a form of DID (I presume),
then no notion that there were separate people here would occur to

>> What if we administer the best personality tests that have been
>> so far devised?  Will they show a difference? (Clearly no.)
> Are you saying that personality traits have some direct bearing on
> personal identity? We know that identical twins, separated at birth,
> have a very high correlation so you mean to say that to some extent
> they should be considered practically the same person?

Their correlation is not so high as you'd think.  I strongly suggest
"No Two Alike" by the author of "The Nurture Assumption", Judith
Rich Harris.  The title, it turns out, is to be taken very literally.

> Or are you
> saying that some extremely high correlation would indicate shared
> personal identity?  Please describe how high this would have to be,
> without any circular reference to your conclusion.

As I said earlier, even an advanced AI may present a bewildering
variety of possible metrics.  I will submit this as sufficient (though
no doubt not necessary):  if the two duplicates manage to
completely fool everyone into believing that there is only one
of them, and that no one no matter how intimately they know
him or her can tell the difference, then that's high enough. They're
the same person.

There is movie you may wish to see, "The Prestige" that explores
this very scenario.  Sorry to give away spoilers, but a woman
cannot tell the difference between two male duplicates that
she's in love with.  Can't ever happen, so far as I know, with
identical twins.

>> I still don't agree.  Are there other examples that can be offered?
>> How about some other examples where *agency* is clearly key?
>> Perhaps in daily life?
> I did that earlier, showing how over time a person (Aging Alice)
> changes and even spawns variants while maintaining the same agency.
> You chose to interpret it as a person continually or repeatedly
> dieing.
> <http://lists.extropy.org/pipermail/extropy-chat/2006-November/030188.html>

But isn't her agency changing, according to you?  She surely acts a
lot differently, (plays different roles and so on). Would you mind
saying again why her agency remains the same?  Just because people
use the same name for her, and are aware of her history?  Surely
if the 6 year old had a good friend, and that friend moved away,
it might be entirely possible for them to meet up later in a different
place, become friends, and because of a name change (say Alice
got married) never realize that it was their old chum from childhood.


More information about the extropy-chat mailing list