Re: Natural selection

From: Jacques Mallah <>
Date: Wed, 05 Jul 2000 19:11:29 EDT

>From: Russell Standish <>
>Jacques Mallah wrote:
> > I didn't ask for a full theory. I require a definition of observer
>because 1) it seems obvious that, in principle, there is no such definition
>that would have all the properties you want. The only way to dissuade me
>of that would be to provide said definition, and 2) without the definition,
>I don't know what you are trying to say.
>So that we're quite clear what I'm talking about, the observer
>referred to is the Anthropic Principle observer. It would be nice to
>have a clear definition of what that is - eg is it required to
>self-aware, as Max Tegmark implies. All I say about this observer is
>a) It must be capable of computation (though not necessarily universal
>b) It must experience a time dimension in which to perform these
>c) It must be able to project from the ensemble of possibilities to
>the actual outcome it sees.
>That's it. b) is required for a). a) is required for the Schmidhuber
>"bitstring" ensemble to be experienced. c) is required for the
>Anthropic Principle.
>I don't see any inconsistencies between these properties. There is
>possibly an inconsistency between c) and computationalism, which I
>have pointed out. If true, there would be an inconsistency between AUH
>and computationalism. However, this is a completely spearate argument.

    "The AP observer"? I asked for the defintion of an observer. Most
importantly, how could one distinguish one observer from another,
observer-moments belonging to one observer from those belonging to another.
This you did not address at all.
    I don't know what your c) is supposed to mean.

> > >The way the statements M(x) and M(x') that you gave are phrased, they
>refer to the same "generic" observer, wondering why it sees an ordered
> >
> > If there is such a "generic" observer, then is he the only observer
>ever to be in existance? If so, then maybe I could understand what you
>Not the only observer, but the only class of observers. The "generic"
>observer can only have properties that all observers must share.

    So the "generic" observer is really not an observer at all, but instead
...? Not a set of observers, from what it sounds like. Perhaps you mean
that any observer would have the same measure distribution as any other?
Absolute measures?

> > Otherwise, surely the Bayesian probabilities that he is the various
> > observers come into play, which you refuse to consider.
> >
> > > > Suppose I see a guy named Harry, who verifiably came into being
>due to 'random fluctuations'. Surely, whatever is meant by 'Harry', that
>is one of his characteristics and it's what makes him so popular at
> > >
> > >This would be a truly remarkable observation, and would probably put
>into grave doubt the whole idea of the SSA and the Anthropic Principle.
>Instead, it would make the idea of an intelligent designer much more
> >
> > Now that's a real red herring. Forget that and answer the question:
>is it one of Harry's characteristics? And if so, shouldn't Harry still be
>surprised to have such an origin? And if that too, then he is clearly
>considering the measure of other observers, which he knows is larger than
>the measure of freaks like him.
>Yes, but Harry is still only considering the generic class of
>observer. He is surprised to find himself a freak, but can recognise
>that his case is not impossible.

    Supposing it is a characteristic of him as an observer. Then he is
considering the Bayesian probability that an observer, prior to knowing who
he is, would be a freak. By considering prior Bayesian probabilities of
being different observers, he must be using absolute measure. Period.

> > In the AUH, all possibilities exist. So no information is needed to
>specify which ones exist. But of course there is stucture.
> > Consider the UTM analogy of the universal distribution. The
>universal distribution is the distribution of output strings produced by a
>UTM, given a uniform distribution of input strings.
> >
> > >This is my point - all measure distributions must be relative to some
>information processing entity (be they observer or UTM or whatever). These
>"observers" needn't be concrete, but they must have some properties to
>enable information in the form of measure distributions to be extracted.
> >
> > First, as I have said many times on this list, for the UTM analogy
>to work there must exist some special UTM or distribution of UTMs which is
>to be used. This does *not* mean information is required, because it
>should be possible to know this distribution based only on special
>properties that it ought to have. (Just as it does not take information to
>know that pi > 3.)
> > If on the other hand that is not sufficient, then nature would need to
>'arbritrarily' make a choice, in which case the zero information hypothesis
>should be considered FALSIFIED.
>Wrong. There may well be a special UTM, or distribution of UTMs, in
>which case what you say applies. However, it is simply not necessary
>for the Universal Prior argument to work.

    Sure it is. No special distribution, no universal prior.

> > Second, if we consider a measure distibution of conscious
>computations, then obviously the definition of 'conscious' acts as a
>filter. To go back to a bitstring analogy, suppose there is a uniform
>distribution 'D' of strings, and we want the measure distribution of such
>strings given an additional property, c. Obviously M(c) will not (in
>general) be uniform. Again - no information is required to specify 'D'
>because it is uniform, yet it leads to a nontrivial M(c).
> > M(c) is quite objective; anyone who calculates it would get the same
>result. In our case, we are interested in consciousness, so we choose that
>as a filter. Me being conscious has nothing to do with the result, just as
>an unconscious computer could calculate M(c).
>Maybe this is true, but you've made several assumptions. One of which
>is that the ensemble of all conscious entities must implement a well
>defined distribution of UTMs. I don't need to assume this, and in the
>absence of a good reason, I think it is a mistake to make such an
> >assumption.

    I have assumed no such thing. Of course as I said, a special
distribution of UTMs must implement all conscious computations, but that's
quite different. You may think you don't need it, but then you are wrong as

                         - - - - - - -
               Jacques Mallah (
         Physicist / Many Worlder / Devil's Advocate
"I know what no one else knows" - 'Runaway Train', Soul Asylum
         My URL:

Get Your Private, Free E-mail from MSN Hotmail at
Received on Wed Jul 05 2000 - 16:16:20 PDT

This archive was generated by hypermail 2.3.0 : Fri Feb 16 2018 - 13:20:07 PST