Ben Goertzel writes:
>I read your argument for the UDA, and there's nothing there that
>particularly worries me.
Good. I don't like to worry people. (Only those attached
dogmatically to BOTH comp AND the existence of a stuffy
substancial universe should perhaps be worried).
You seem to be making points about the limitations
>of the folk-psychology notion of identity, rather than about the actual
>nature of the universe...
Then you should disagree at some point of the reasoning, for the
reasoning is intended, at least, to show that it follows from
the computationalist hypothesis, that physics is a subbranch of
(machine) psychology, and that the actual nature of the universe
can and must be recovered by machine psychology.
(I do use some minimal Folk Psychology in UDA, and that can be
considered as a weakness, and that is one of the motivation---
for eliminating the need---to substitute it (folk psychology)
by machine self-referential discourses in the Arithmetical-UDA).
>> >When you say "sum over all computational histories", what if we
>> just fix a
>> >bound N, and then say "sum over all computational histories of
>> >info. content <= N." Finite-information-content-universe, no Godel
>> >problems. So what's the issue?
>> The main reason is that, once we postulate that we are turing emulable,
>> (i.e. the computationalist hypothesis comp), then there is a form
>> of indeterminacy which occurs and which force us to take into account the
>> incompleteness phenomenon.
>I'm sorry, but I don't get it. Could you please elaborate?
Physics is taken as what is invariant in all possible (consistent)
anticipation by (enough rich) machine, and this from the point of
view of the machines. If arithmetic was complete, we would get
just propositional calculus. But arithmetic is incomplete.
This introduces nuances between proof, truth, consistency, etc.
The technical part of the thesis shows that the invariant propositions
about their probable neighborhoods (for
possible anticipating machines) structure themtselves into a sort
of quantum logic accompagned by some renormalization problem (which
could be fatal for comp (making comp popperian-falsifiable)).
This follows from the nuances which are made necessary by the
Godel's incompleteness theorems, but also Lob and Solovay
fundamental generalization of it. But it's better grasping first
the UDA before tackling the AUDA, which is "just" the translation
of the UDA in the language of a "Lobian" machine.