Re: Memory Merging Possible For Close Duplicates

From: Lee Corbin (
Date: Mon Mar 17 2008 - 18:17:07 MDT

Mike writes

> The original context was a fictional
> scenario where you listened to a 20 minute tale of your wife's day.
> There is a level of detail that is "just the facts," detail that
> includes humorous situations perhaps only your wife would find
> relevant, details that includes particular facial expressions...
> I was using 'nuances' to include those details which provide only
> subtle refinements on a reasonably useful answer to the question,
> "How was your day?"

Thanks for the explanation.

> In other words... you missed that point.

Unfortunately, this is not the only time that this has ever happened to me.

> It was perhaps of tertiary importance to what I was saying. It has
> become a distraction. I would have aggressively trimmed this from
> my reply, but I think it served as a self-referential example.

I see :-)

>> It is similarity of structure, ultimately. Of course, we have various
> [snip]
>> Suppose that by some *obvious* [1] isomorphism, today
>> Lee is one bit string where two 0's happen to be changed to 1's.
>> By all measures of similarity, those two strings are still very similar.
> Your definition of similar is not sufficient for me to make sense of
> your bit string analogy. I "get" that there might be some reference
> to TM or Life Board in there, but I'd have to guess too much.

Okay, I started a new thread called "Similarity of Structure".

> I'm learning that when I assume too much agreement I will find
> later that whatever building goes up will inevitably be torn down
> to its unstable foundation.

Well, perhaps not "inevitably". But you are dead right that this
happens far too often, and we all have to be on the lookout for it.

>> If the teleporter or copying machinery is working well, there
>> will be very few bit errors.
> I'll grant that you have a perfect copy maker including the dimension
> of time, such that all bits recording one's existence up to the point
> of the copy are absolutely identical. Each instance has their new
> experience encoded and appended to the identical stream of bits that
> were absolutely perfectly copied. Using the Sa->Sb->etc notation from
> another current thread, I would imagine normal/expected
> causation-induced changes between states over a small interval of time
> to be similar (though I am not 100% certain how to accurately measure
> similar to five nines precision, right?)

I'm not certain either, but I believe that it can be done. We can take
it up further in the "Similarity of Structure" thread.

>> > [...]memories [..] are used to determine identity up to the point of copy [...]
>> Yes. Or new experiences which create or destroy small parts of the structure.
> You are allowing new experience to create or destroy parts of the
> structure? That is inconsistent with your definition of preserving
> information as the True history of an entity's existence.

I don't think that the "history" as such---that is, a detailed record
of how all the little changes get made day in and day out in an
individual---is important. What is important is that my memory,
say, of my elementary school not undergo damage. But it's okay
if I forget the name of a certain leader of a foreign nation. I call
that "creating" and "destroying" small parts of the structure. Bad
word choice? Clearly something was destroyed---information
of some kind was lost---when I forgot whats-his-name.

Now it's also true, incidentally, that the more we can, for example,
remember about details of our daily life, the better that our identities
are preserved over time. What unfortunately will make me almost
not the same person 40 years from now---new medical tech aside
---is that I will have forgotten so much about the way I am now.

> Once the log is open for editing, there will be no integrity to any
> measure.

You mean, for example, when once we're uploaded and we have
the power to edit any of our memories easily and at-will? Well,
yes, then someone would have the ability to utterly destroy his
personal "integrity" if you will, his wholeness and what it is about
him that makes him who he is. But I suspect that I'm not reading
you right. You need to elaborate?

> By this possibility I could simply edit-in or edit-out any memory
> necessary to force the checksum in your similarity measure.

Oh, it's not a checksum, which is a many-to-one mapping. It's
a perfect hash, say, or a scan of the kind that can be counted
upon for re-creating someone. E.g., we assume that the
information, probably sent bit-by-bit by Scotty's teleporter,
is enough to re-create a very, very convincing replica of
Captain Kirk. Damn, it occurs to me that I could have saved
a lot of words in the new Similarity of Structure post.

> This is all made up TE anyway, so even if I can't hack your
> memory gracefully I should be able to do so by brute force.
> Oops, didn't get it right on the first try... maybe I'll get it on
> the second attempt, or fifty-second attempt.

If you mean the dystopian nightmare of the authorities being
able to redact or alter someone's memories, why, yes, no
reason in principle that they couldn't hack my memories.
If they went far enough, I'd turn from a Winston Smith
into an O'Brien in no time.

> I have consistently used "belief" to indicate a subjective understanding
> of the facts available to a clone. (eg: the clone believes itself to
> be... or I believe X from a subjective measure.) I certainly
> understand the confusion this may cause since the word belief is
> overloaded with meaning such that it should perhaps not be used at
> all.

Yes. I use "belief" like that for the relationship between what
people think they know and what is true. Eliezer has described
this in detail in Overcoming Bias sometime in the last few days.
Anyway, I guess that the word is being used the same way right
now here in this thread, and probably need concern us no longer.

>> > close duplicates should merge at least as easily as they are created?
>> I don't see how that last statement follows from anything above.
> I do not know how to proceed. The statement does not follow from
> anything above, it was the starting point for my participation in this
> thread.


> It has taken this dialog to reach the point of agreement above.
> I imagine this is where in an in-person conversation that one
> of us would offer to refresh drinks, then start again on a different
> topic upon returning. :)

Oh, okay. But you don't defend "close duplicates should merge
at least as easily as they are created"? Or 'twas someone else
said that?

> I imagine this is where in an in-person conversation that one
> of us would offer to refresh drinks, then start again on a different
> topic upon returning. :)

Entirely possible. Sometimes these kinds of discussions really
do peter out :-) But surely something I've said above in this
not-very-short post seems either wrong or too confusing.


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:02 MDT