Le 07-oct.-06, à 16:35, Stathis Papaioannou a écrit :
> Bruno marchal writes:
>> Le 05-oct.-06, à 20:49, markpeaty a écrit :
>>> I started to read [the English version of] your discourse on Origin
>>> Physical Laws and Sensations. I will read more later. It is certainly
>>> very interesting and thought provoking. It makes me think of 'Reasons
>>> and Persons' by Derek Parfitt. His book is very dry in places but
>>> mostly very well worth the effort of ploughing through it.
>> Parfit is good. I stop to follow him when he insists that we are
>> I paraphrase myself sometimes by the slogan MANY TYPES NO TOKEN.
> Can you explain the disagreement with Parfit? My reading of chapter 99
> R & P is that a "token" is a particular instantiation of a person
> while a "type"
> is the ensemble of related instantiations. "Mary Smith" is a type,
> "Mary Smith
> coming out of replicator no. 978 at 11:05 AM" is a token.
When I say "MANY TYPES NO TOKEN", I assume comp *and* the conclusion I
derive from it, that is the reversal between physics and "number
theoretical machine theory" (say). In particular I take from granted
that my next "observer moment" is somehow determined by two things: a
"proportion" of computational histories going through my actual
computational state, and the proportion of consistent extensions, which
are related to a "proportion" of similar computational histories.
So with comp "Mary Smith coming out of a replicator no. 978 at 11:05
AM" is a type. It is the type of all (2^aleph_0) histories going
through that event (supposedly well 3-described).
From a third person point of view, if you are willing to say that the
natural numbers are token (I am neutral on that), then it would make
sense to see the nth step of an immaterial execution of a DU, (or an
enumeration of the true sigma_1 arithmetical sentences) as (immaterial)
tokens. But even in that case, there would be no sense to attribute
tokenness to "Mary Smith coming out of a replicator no. 978 at 11:05
AM", because there is no way to privilege one instantiation from
another. We must take them all, and they constitute highly undecidable
> It appears that in this terminology (actually due to Bernard Williams,
> not Parfit)
> once generated a token remains the same token until there is another
> but my preference is to generalise the term and say that a token has
> only transient
> existence, which then makes "token" equivalent to "observer moment".
OK. I prefer.
With comp it has to correspond to the third person Observer Moment (OM
They are the true Sigma_1 sentences, or the accessible states by the UD.
> This is
> literally true, given that from moment to moment, even in the absence
> of teleportation
> etc., the atoms in your body turn over such that after a certain time
> none of the
> matter in your body is the "same", and before this time the fact that
> some of the
> matter in your body is the "same" is accidental and makes no
> difference to your
> conscious experience.
Assuming bodies. I see the point.
> As to whether I am token or type: obviously, literally,
> I-who-write-this-now am a
This looks like the first person OM. It is different from the preceding
The 3-OM are enumerable, even recursively enumerable.
The 1-OM are enumerable but not recursively enumerable (for those who
have the Cutland, it is a simple consequence of Rice theorem).
And the similarity classes of the 1-OM (= states plus its relative
proportion) has the power of the continuum.
> My present token is included in the set of related tokens in the
> past, future,
> other branches of the multiverse, surreptitious emulations of my mind
> made by aliens,
> and so on: the type. Note that the definition of a particular token
> (especially in my
> generalised sense, fixed to a specific and unique position in the
> multiverse) can be
> made completely unambiguous,
How? With comp (with the multiverse = UD*) you have to bet on a level
of description of you, and then, even in the lucky case of a correct
bet, I still don't see how you will discover you present token, if only
because of the many undistinguishable computational histories going
through that state (which I recall you can only bet on).
This was about your 3-OM token. The situation is even more difficult
for the 1-OM token, which is determined in the neighborhood of the
And frankly if you believe a recording can be conscious, don't forget
to look for the infinitely any emulation any recording before telling
me where is your 1-OM in UD*.
> while the definition of a type is necessarily vague and
> fuzzy arround the edges. For example, if a being exists somewhere with
> 70% of
> my memories and 30% of your memories, should he be included in my
> type, your type,
> a new type, or some combination of these? It is only because we
> experience a linear
> existence from birth to death, so that only a single token is extant
> at a time and there
> is clear physical continuity from one token to the next, that we can
> ignore the distinction
> between token and type and consider the definition of a particular
> type to be
I agree. But thought experiments involving self-duplication are here to
remind such distinctions. Like the superposition is contagious in the
quantum multiverse, with comp typeness is contagious: Mary Smith, being
digitally duplicable, is a type and when she leaves the replicator 978,
she does not instantiate in "Mary Smith coming out of a replicator no.
978 at 11:05 AM", she just particularizes herself into the more
particular type of the "Mary Smith coming out of a replicator no. 978
at 11:05 AM". But there are many of such types, even not counting those
emulated by the "recordings" emulated in the UD* (or the sigma_1 true
> As for what matters in survival, that is a contingent fact of the way
> our brains have
We can bet that survival of (many pov) values is necessary in the long
> It isn't as simple as saying that survival of type is what matters,
> except in the
> abovementioned linear existence with which we are familiar. One
> concise way to put it
> is that what matters in survival is that there should exist at least
> one token (or observer
> moment, or instantiation) which has my present token's memories in its
> past. Moreover,
> it is important that where several such tokens exist, as I may feel
> myself "becoming" any
> one of them with equal probability, as many as possible should have
> good experiences.
> This works for simple branchings, but it becomes complicated when we
> consider mergings
> and partial memory loss. Another problem is tokens in parallel worlds:
> I could say I don't
> survive if no future tokens exist who have my present memories,
OK, but I really think that it is less misleading to reason on a block
reality (like UD*), instead of taking the risk to confuse first person
time (which cannot not exist) and third person time (which could not
> even though other tokens
> exist in the multiverse who branched off some time ago, but then that
> would be like saying
> I don't survive if I experience any memory loss at all, for example
> due to medications like
> midazolam. It could be that every second, tokens are finding
> themselves in hellish multiverse
> branches where they suffer horribly then die, which is equivalent to
> suffering horribly then
> having the memory of the experience erased.
Hoping that is possible.
> We can describe exactly what happens in each
> of these cases using the token or observer moment terminology,
For the 3-OM, we can only bet on descriptions. We don't know our
The 1-OM can only hope their neighborhoods are enough coherent so that
their goals make sense.
> but reconciling this with
> psychological survival is problematic because our brains did not
> evolve to cope with these
> sorts of situations.
OK. But it is perhaps the same as jumping with a parachute. We can
(The lobian interview shows that the tension between the first person
views and the third person views (and other hypostases) will be with us
forever so that we have to learn forever ... and even after that we can
still expect to be surprised infinitely often and force to cope with
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To post to this group, send email to email@example.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at