language, cloning and thought experiments

[Jack Mallah]

--- On Wed, 2/11/09, Stathis Papaioannou <stathisp@...> wrote:
> Well, this seems to be the real point of disagreement between you and the pro-QI people. If I am one of the extra versions and die overnight, but the original survives, then I have survived. This is why there can be a many to one relationship between earlier and later copies. If you don't agree with this then you should make explicit your theory of personal identity.

It is close to the point, but there is room for a misunderstanding so I have to be careful.  Here I am consolidating replies to some of the branched post threads and will present some thought experiments.

On personal identity:

As I explained, there are several possible definitions of personal identity, and the most useful ones are 1) All branched/fused people are the same person, 2) Causal chains determine identity, and 3) Observer - moments.

This can become confusing because it is not always clear which definition someone is using, especially if quickly typing out a reply to a tangentially related post.  This can lead to a kind of Hydra-whacking effect in which one point is dealt with, only another confusion is simultaneously created because (for example) I was not clear on what I did not spell out as it was not the main point at issue in the post I was responding to.

That was the case recently when some people misconstrued my use of the causal chain (in terms of "you might die, only the original will survive") as some kind of crucial point.  Causal differentiation applied to the question at hand, so I used that one.  If anyone has read my QI paper, you would have known that I accept that "teleportation is OK" and that measure is what matters, not so much the original vs. copy issue.  I will explain more below on these important points.

If I had to pick one definition and stick with it, I would go with the least misleading one, which is an observer-moment.

The important thing to realize is that _definitions don't matter_!  Predictions, decisions, appropriate emotions to a situation - these are completely independent of definitions of personal identity.  Personal identity is a useful concept in practice but not a fundamental thing, and therefore can have no fundamental relevance, unlike its misuse in QS thinking where it could supposedly affect a measure distribution.

On probability:

Bruno Marchal wrote:
> You say: "no randomness involved" but you seem to accept probabilities. Do I just miss something here?

Yes, Bruno, you did, though my quickness contributed.  In my QI paper I defined "effective probability" and carefully spelled out the roles it can play.  But again, in posting on tangentially related topics, it is much easier to just say "probability" and hope that people remember what I am really talking about.

Classically there are two kinds of probability: true randomness, and subjective uncertainty due to ignorance.  I do not believe that the former exists..  When I talk about probability it either involves some ignorance on the part of the subject (as in the Reflection argument), or the use of "effective probability" in theory confirmation.

I may get sloppy sometimes (and say probability) when talking about a situation after an experiment that is yet to be performed, but in thinking about such cases it is absolutely necessary to remember that in the MWI there is neither randomness nor subjective ignorance, and that one must use Caring Measure.

On the "first person" slogan:

Any observation is made by the person observing it.  In that sense, they are all "first person".

Truths do not depend on point of view.  We do not know the measure distribution, but we can guess about it, and can study a model for it.  Assuming the model is accurate, it is the distibution of these "first person" observations.

Calling it a "third person" view is a false charge; an accurate model is not a view, it is simply the truth.  Invoking "first person measure distributions" as an alternative is an empty slogan.

The real key point at which the QS fallacy appears seems to be that some people find it inconcievable that they will not have a future.  Thus, they assume that they will survive and only need to take into account effective probabilities that are conditional on survival.  This fallacy is undefined (in terms of personal identity which is required for the condition) and is false by definition of the measure distribution.

This can be seen using either causal chains (if a person is defined as a casual chain, then when the chain ends, so will he) or more generally just in terms of decreasing measure of observer-moments with age.  In the latter case increasing age is no different than, for example, increasing brightness of your visual field.  There is a sequence of observer-moments in which what you see is more and more bright, and after some point the measure distribution will decline as a function of increasing brightness.  You can define (or attempt to define) a conditional effective probability for what else you would experience as a function of increasing brightness, but I don't see people worrying about it in the same way they do for increasing age.  There is no logical reason to make such a distinction.

Thought experiments:

1) The fair trade

This is the teleportation or Star Trek transporter thought experiment.  A person is disintegrated, while a physically identical copy is created elsewhere.

Even on Star Trek, not everyone was comfortable with doing this.  The first question is: Is the original person killed, or is he merely moved from one place to another?  The second question is, should he be worried?

The answer to the first question depends on the definition of personal identity.  If it is a causal chain, then if the transporter is reliable, the causal chain will continue.  However, if the copy was only created due to extreme luck and its memory (though coincidentally identical to that of the original) is not determined by that of the original, then the chain was ended and a new one started.

The second question is more important.

Since we are considering the situation before the experiment, we have to use Caring Measure here.  The temptation is to skip such complications because there is no splitting and no change in measure, but skipping it here can lead to confusion in more complicated situations.

The utility function I'll use is oversimplified for most people in terms of being so utilitarian (as opposed to conservatively history-respecting, which might oppose 'teleportation') but will serve.

So if our utility function is U = M Q, where M is the guy's measure (which is constant here) and Q is his quality of life factor (which we can assume to be constant), we see that it does not depend on whether or not the teleportation is done.  (In practice, Q should be better afterwards, or there is no reason to do it.)  Therefore it is OK to do it.  It is a fair trade.

2) The unfair trade

Now we come to the situation where there are 2 ‘copies’ of a person in the evening, but one will be removed overnight, leaving just one from then on.  I’ll call this a culling.

I pointed out that in this situation, the person does not know which copy he is, so subjectively he has a 50% chance of dying overnight.  That is true, using causal chains to define identity, but the objection was raised that ‘since one copy survives, the person survives’ based on the ‘teleportation’ idea that the existence elsewhere of a person with the same memory and functioning is equivalent to the original person surviving.

So to be clear, we can combine a culling with teleportation as follows: both copies are destroyed overnight, but elsewhere a new copy is created that is identical to what the copies would have been like had they survived.

Is it still true that the person has a subjective 50% chance to die overnight?  If causal chains are the definition, then depending on the unreliability of the teleporter and how it was done, the chance of dying might be more like 100%.  But as we have seen, definitions of personal identity are not important.  What matters is whether the person should be unhappy about this state of affairs; in other words, whether his utility function is decreased by conducting the culling.

Using U = M Q, it obviously is decreased, since M is halved and Q is unchanged.  So as far as I can see, the only point of contention that might remain is whether this is a reasonable utility function.  That is what the next thought experiment will address.

3) The Riker brothers

Will T. Riker tried to teleport from his spaceship down to a planet, but due to a freak storm, there was a malfunction.  Luckily he was reconstructed back on the ship, fully intact, and the ship left the area.

Unknown to those on the ship, a copy of him also materialized on the planet..  He survived, and years later, the two were reunited when Will’s new ship passed by.  Now known as Tom, the copy that was on the planet did not join Star Fleet but went on to have many adventures of his own, often supporting rebel causes that Will would not.  Will and Tom over their lifetimes played important but often conflicting roles in galactic events.  They married different women and had children of their own.

The two brothers became very different people, it seems obvious to say – similar in appearance, but as different as two brothers typically are.  It should be obvious that killing one of them (say, Will) would not be morally OK, just because Tom is still alive somewhere.  Based on functionalism, Will is just as conscious as any other person; his measure (amount of consciousness) is not half that of a normal person.

When did they become two people, rather than one?  Did it happen as soon as the first bit of information they received was different?  If that were so, then it would be morally imperative to measure many bits of information to differentiate your own MWI copies.  But no one believes that.

No, what matters is that Rikers’ measure increased during the accident.  As soon as there were two copies of him – even before they opened their eyes and saw different surroundings – they were two different (though initially similar) people and each had his own value.  Their later experiences just differentiated them further.

4) The 'snatch'

Suppose an alien decided to ‘clone’ (copy) a human, much as Riker was copied.  The alien scans Earth and gets the information he needs, then goes back to his home planet, many light years away.  Then he produces the clone.

Does this in any way affect the original on Earth?  Is half of his consciousness suddenly snatched away?  It seems obvious to me that it is not.  If it is not, then measure is what matters, not ‘first person probabilities’.

5) The groggy mornings

Bob always wakes up to his alarm clock at 7:00 every morning.  However, he is always groggy at first, and some of his memory does not ‘come on line’ until 7:05.

During this time, Bob does not remember how old he is.  Since there is genuine subjective uncertainty on his part, he can use the Reflection Argument to guess his current age.  The effective probability of his being a certain age is proportional to his measure during that year.  Thus, we can talk about his expectation value for his age, the age which he is 90% ‘likely’ to be younger than, etc.

If he is mortal, then his measure decreases with time, so that his expected age and so on fall into typical human parameters.

If he were immortal, then his expected age would diverge, and the ‘chance’ that his age would be normal is 0%.  Clearly that is not the case.  Thus, having a long time tail in the measure distribution is not immortality.






     


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list@...
To unsubscribe from this group, send email to everything-list+unsubscribe@...
For more options, visit this group at http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---

[daddycaylor]

I noticed someone taking my name in vain. ;)  (though experiment where
I, Tom, am a clone of Will Riker)  The magic of thought experiments,
it's amazing.  I felt my measure decrease, but only after I read the
thought experiment.

I trust this will not derail anyone's personal identity here, but I
have some subjectively random thoughts.  The importance of, the
necessary dependence of any model of reality on, our subjective view
of things has struck me lately.  For instance, I brought up the
"should I worry about a clone dying?" question a while back.  This
seems totally subjective to me.  And that's not a put-down, it's an
observation.  What if we started thinking about making a clone as sort
of like having a baby?  Then it would suddenly change from being
something uncertain to something exhilating, plus there probably
wouldn't be any diapers involved.  I'm not suggesting we do that, just
performing a thought experiment.  When we die, the "personal identity"
pain is less if we have passed on something of ourselves in some way.
But then of course we care if that passed-on stuff is discontinued,
for instance if one of our children die, or a symphony we've written
is forgotten.  The Golden Rule still applies.

The latest Scientific American has an article about non-locality, and
it seems to me that this is related to this topic, through the causal-
chain aspect of it.  One thought-picture that was used to try to
convey non-locality is that it is like a fist punching in Chicago and
a face being hurt in Los Angeles.  So it occurred to me that it is
only in the presence of a consciously-aware assignment of cause that a
causal chain is present.  We are not omniscient, there is always
something in a universe which cannot be predicted by any model within
that universe.  There are always those faces feeling pain for which we
have no knowledge of the cause, so how can we claim that everything
has a cause?  So effectively there are things happening in any
universe which have no cause.

But what does that mean (subjectively of course) to us?  It shouldn't
stop us from trying to find causes and do predictions, since this
works for everything that we need to work, macroscopic things, local
things.  Just rambing here, it seems to me that the whole quantum
entanglement/non-locality thing is fairly intuitive.  It is based on
the fact that you need at least three things to have meaning.  (For
instance, the distance between two points by themselves is
meaningless.)  And one of those things seems to be consciousness.

Tom (Riker's brother)

P.S. On your groggy morning, does the fact that you can have an
infinitely long tail but have a finite area underneath it have any
bearing?

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list@...
To unsubscribe from this group, send email to everything-list+unsubscribe@...
For more options, visit this group at http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---

[Wei Dai]

Jack, welcome back. I no longer read every post here, but I read this post
and found your positions pretty close to my own. This one, especially, I
totally agree with:

> The important thing to realize is that _definitions don't matter_!
> Predictions, decisions, appropriate emotions to a situation - these are
> completely independent of definitions of personal identity.

This one is more problematic:

> So if our utility function is U = M Q, where M is the guy's measure (which
> is constant here) and Q is his quality of life factor (which we can assume
> to be constant), [...]

The ASSA/RSSA and QTI debates can be rephrased as whether U should equal
M*Q, or just Q, but that is an "ought" question. If we accept the standard
view in decision theory that the utility function is completely subjective,
then that means the ASSA/RSSA debate can't be resolved by objective
arguments.

We can get around this a bit by asking what most people's utility functions
actually are, instead of what they ought to be. Are they closer to M*Q, or
Q? I'm afraid that for most people, it's closer to Q than M*Q. One might
have expected that evolution would have programmed us to have U=M*Q, but
that doesn't seem to have been the case. I have a couple of speculations as
to why:

1. M cannot be perceived directly. It can be inferred, but that takes a lot
of work.

2. In our EEA, M couldn't increase, only decrease. (Because there were no
mind-copying machines.) So evolution could essentially simulate the effect
of U=M*Q with U=Q plus fear of pain and fear of dying, and that's what it
did because it's a lot easier than getting the brain to compute M. (For a
similar reason, we value sex instead of number of offspring.)

Initially I was also an advocate of ASSA until I realized that it's
ultimately a subjective question of values. I think once mind-copying
machines are invented, there will be a much greater selection pressure
towards U=M*Q. But given that U=Q is closer to the reality today, I'm not
sure what good it would do to "taking a stand against QS/QI".
 


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list@...
To unsubscribe from this group, send email to everything-list+unsubscribe@...
For more options, visit this group at http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---

[Günther Greindl]

Jack, Wei Dai,

> machines are invented, there will be a much greater selection pressure
> towards U=M*Q. But given that U=Q is closer to the reality today, I'm not
> sure what good it would do to "taking a stand against QS/QI".

To "translate":
U=M*Q is 3rd person POV (hypothetical; viewed from outside
platonia/spacetime worm/insert fav. metaphysics)

U=Q is 1st POV (a cognitive agent reasoning if he/she/it will have next
experience)

Again, it seems that QI is conceded, and an emotional argument made to
care for all successors. I see no contradiction, and no refutation or
whatever of QI.

 >The real key point at which the QS fallacy appears seems to be that
 >some people find it inconcievable that they will not have a future.

Having only a finite number of successor moments is standard materialist
assumption. It is computationalism which seems to suggest otherwise.
Inconceivability does not enter the picture.

 >Thus, they assume that they will survive and only need to take into
 >account effective probabilities that are conditional on survival.

That is correct, as not surviving is not an experience. As long as a
successor with "your" memories of previous moments exists, you survive.

 > 5) The groggy mornings
 >If he were immortal, then his expected age would diverge, and the
 >‘chance’ that his age would be normal is 0%.

Ok, this is ASSA reasoning. But it does not follow with RSSA. So, it is
just the old argument that with ASSA you don't have immortality but with
RSSA you do?

A "refutation" of QI would also require refuting it under RSSA,
otherwise you simply have claim _if_ ASSA _then_ no QI, which does not
seem disputed.

As written in a previous post, I think an RSSA-reasoner should also care
for all his successors, and thus not engage in a QS-experiment or stuff
like that.

Cheers,
Günther

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list@...
To unsubscribe from this group, send email to everything-list+unsubscribe@...
For more options, visit this group at http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---

[Stathis Papaioannou-2]

2009/2/25 Jack Mallah <jackmallah@...>:

> 1) The fair trade
>
> This is the teleportation or Star Trek transporter thought experiment.  A person is disintegrated, while a physically identical copy is created elsewhere.
>
> Even on Star Trek, not everyone was comfortable with doing this.  The first question is: Is the original person killed, or is he merely moved from one place to another?  The second question is, should he be worried?
>
> The answer to the first question depends on the definition of personal identity.  If it is a causal chain, then if the transporter is reliable, the causal chain will continue.  However, if the copy was only created due to extreme luck and its memory (though coincidentally identical to that of the original) is not determined by that of the original, then the chain was ended and a new one started.
>
> The second question is more important.
>
> Since we are considering the situation before the experiment, we have to use Caring Measure here.  The temptation is to skip such complications because there is no splitting and no change in measure, but skipping it here can lead to confusion in more complicated situations.
>
> The utility function I'll use is oversimplified for most people in terms of being so utilitarian (as opposed to conservatively history-respecting, which might oppose 'teleportation') but will serve.
>
> So if our utility function is U = M Q, where M is the guy's measure (which is constant here) and Q is his quality of life factor (which we can assume to be constant), we see that it does not depend on whether or not the teleportation is done.  (In practice, Q should be better afterwards, or there is no reason to do it.)  Therefore it is OK to do it.  It is a fair trade.
>
> 2) The unfair trade
>
> Now we come to the situation where there are 2 ‘copies’ of a person in the evening, but one will be removed overnight, leaving just one from then on.  I’ll call this a culling.
>
> I pointed out that in this situation, the person does not know which copy he is, so subjectively he has a 50% chance of dying overnight.  That is true, using causal chains to define identity, but the objection was raised that ‘since one copy survives, the person survives’ based on the ‘teleportation’ idea that the existence elsewhere of a person with the same memory and functioning is equivalent to the original person surviving.
>
> So to be clear, we can combine a culling with teleportation as follows: both copies are destroyed overnight, but elsewhere a new copy is created that is identical to what the copies would have been like had they survived.
>
> Is it still true that the person has a subjective 50% chance to die overnight?  If causal chains are the definition, then depending on the unreliability of the teleporter and how it was done, the chance of dying might be more like 100%.  But as we have seen, definitions of personal identity are not important.  What matters is whether the person should be unhappy about this state of affairs; in other words, whether his utility function is decreased by conducting the culling.
>
> Using U = M Q, it obviously is decreased, since M is halved and Q is unchanged.  So as far as I can see, the only point of contention that might remain is whether this is a reasonable utility function.  That is what the next thought experiment will address.

If you're not worried about the fair trade, then to be consistent you
shouldn't be worried about the unfair trade either. In the fair trade,
one version of you A disappears overnight, and a new version of you B
is created elsewhere in the morning. The unfair trade is the same,
except that there is an extra version of you A' which disappears
overnight. Now why should the *addition* of another version make you
nervous when you wouldn't have been nervous otherwise? Sure, you don't
know whether you are A or A', but the situation is symmetrical: if you
are A the presence of A' should make no difference to you, and if you
are A' the presence of A should make no difference to you. And if
something makes no difference to you, it shouldn't impact on your
utility function.

> 3) The Riker brothers
>
> Will T. Riker tried to teleport from his spaceship down to a planet, but due to a freak storm, there was a malfunction.  Luckily he was reconstructed back on the ship, fully intact, and the ship left the area.
>
> Unknown to those on the ship, a copy of him also materialized on the planet..  He survived, and years later, the two were reunited when Will’s new ship passed by.  Now known as Tom, the copy that was on the planet did not join Star Fleet but went on to have many adventures of his own, often supporting rebel causes that Will would not.  Will and Tom over their lifetimes played important but often conflicting roles in galactic events.  They married different women and had children of their own.
>
> The two brothers became very different people, it seems obvious to say – similar in appearance, but as different as two brothers typically are.  It should be obvious that killing one of them (say, Will) would not be morally OK, just because Tom is still alive somewhere.  Based on functionalism, Will is just as conscious as any other person; his measure (amount of consciousness) is not half that of a normal person.
>
> When did they become two people, rather than one?  Did it happen as soon as the first bit of information they received was different?  If that were so, then it would be morally imperative to measure many bits of information to differentiate your own MWI copies.  But no one believes that.
>
> No, what matters is that Rikers’ measure increased during the accident.  As soon as there were two copies of him – even before they opened their eyes and saw different surroundings – they were two different (though initially similar) people and each had his own value.  Their later experiences just differentiated them further.

That Riker's measure increased is not the important thing here: it is
that the two Rikers differentiated. Killing one of them after they had
differentiated would be wrong, but killing one of them before they had
differentiated would be OK. To give another example, if whenever the
teleporter operates two copies are created, one on the surface of the
planet and one at the centre of the planet's sun, then it wouldn't
worry me to use it: I would expect to certainly find myself on the
planet. On the other hand, if the teleporter is like the one in "The
Prestige", I would be very anxious about using it, since I would have
a 1/2 chance of an unpleasant death.

You might not agree with the above appraisal, just as someone may
refuse to use the teleporter, sticking with the belief that it would
mean certain death no matter what arguments are used to try to
persuade him otherwise. If so, this would seem the reason you don't
accept QS, and we have reached an impasse.

> 4) The 'snatch'
>
> Suppose an alien decided to ‘clone’ (copy) a human, much as Riker was copied.  The alien scans Earth and gets the information he needs, then goes back to his home planet, many light years away.  Then he produces the clone.
>
> Does this in any way affect the original on Earth?  Is half of his consciousness suddenly snatched away?  It seems obvious to me that it is not.  If it is not, then measure is what matters, not ‘first person probabilities’.

The cloned person should expect to suddenly find himself at the
distant planet with 50% probability. This is equivalent to
non-destructive teleportation, or destructive teleportation to two
separate locations.

> 5) The groggy mornings
>
> Bob always wakes up to his alarm clock at 7:00 every morning.  However, he is always groggy at first, and some of his memory does not ‘come on line’ until 7:05.
>
> During this time, Bob does not remember how old he is.  Since there is genuine subjective uncertainty on his part, he can use the Reflection Argument to guess his current age.  The effective probability of his being a certain age is proportional to his measure during that year.  Thus, we can talk about his expectation value for his age, the age which he is 90% ‘likely’ to be younger than, etc.
>
> If he is mortal, then his measure decreases with time, so that his expected age and so on fall into typical human parameters.
>
> If he were immortal, then his expected age would diverge, and the ‘chance’ that his age would be normal is 0%.  Clearly that is not the case.  Thus, having a long time tail in the measure distribution is not immortality.

If the probability function has an infinite tail but most of the area
under the curve is to the left of a finite age, then if he forgets how
old he is he should bet that he is younger than this age. But if the
tail is infinite, that still means he can expect to live forever.


--
Stathis Papaioannou

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list@...
To unsubscribe from this group, send email to everything-list+unsubscribe@...
For more options, visit this group at http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---