Bruno Marchal writes (quoting SP):

> >>> ...a controlled
> >>> experiment in which measure can be turned up and down leaving
> >>> everything else
> >>> the same, such as having an AI running on several computers in 
> >>> perfect
> >>> lockstep.
> >>
> >>
> >> I think that the idea that a lower measure OM will appear more complex
> >> is a consequence of Komogorov like ASSA theories (a-la Hal Finney,
> >> Mallah, etc.). OK?
> >
> > I understand the basic principle, but I have trouble getting my mind 
> > around
> > the idea of defining a measure when every possible computation exists.
> I am not sure I understand. All real number exist, for example, and it 
> is the reason why we can put a measure on it. All computations exist 
> (this is equivalent with arithmetical realism) yet some are or at least 
> could be relatively more frequent than others.

Sure, but it's the details that are mind-boggling. Why do dog-computations 
bark and cat-computations meow? If there is a definite mathematical answer 
how do we even begin to fathom it? Or would you go the reductionist route 
of starting with basic physical laws, on which chemistry, biology, psychology 
etc. are built, the more basic sciences supporting the less basic?

> >>
> >> I agree from some 1 pov. But 1 plural pov here would lead to some 
> >> "Bell
> >> inequalities violation". That is: sharable experiments which shows
> >> indirectly the presence of some alternate computations.
> >
> > I don't understand this statement. I am suggesting that the computers 
> > are
> > running exactly the same program - same circuitry, same software, same
> > initial conditions, all on a classical scale. I don't see that there 
> > is any way
> > for the AI to know which computer he was running on (if that question 
> > is
> > even meaningful) or how many computers were running.
> I know it looks counterintuitive, but an AI can know which computer is 
> running and how many they are. It is a consequence of comp, and the UDA 
> shows why. The answer is:
> the computer which is running are the relative universal number which 
> exist in arithmetical platonia (arithmetical truth is already a 
> universal video game, if you want, and it is the simplest). How many 
> are they? 2^aleph_zero.
> I have already explain it here:
> It is a key point and we can come back on it if you have some 
> difficulties.

Well now I'm confused! I thought the whole point of the earlier part of the UDA 
as discussed in the cited post (and many others of yours) is that you *can't* 
know the details of your implementation, such as what type of computer you are 
being run on, how fast it is running, if there are arbitrary delays in the 
and so on. Are you now saying that if I am being run on the 3rd of 100 PC's in 
the basement of the local university computer science department, but everyone 
is keeping this a secret from me, there is a way I can figure out what's going 
all by myself?!

> >>> If I
> >>> were the AI the only advantage I can think of in having multiple
> >>> computers running
> >>> is for backup in case some of them broke down; beyond that, I 
> >>> wouldn't
> >>> care if there
> >>> were one copy or a million copies of me running in parallel.
> >>
> >> Except, as I said above, for the relative probabilities. But this is
> >> equivalent with accepting a well done back-up will not change your
> >> "normal" measure.
> >
> > Yes, I think what you mean by "relative probabilities" is that if 
> > there were
> > several possible versions of "me next moment", then I would be more 
> > likely
> > to experience the one with higher measure. It is only relative to the 
> > other
> > possibilities that measure makes a subjective difference.
> Ah but you get the point now!

So, as long as this *relative* measure does not come into play, the absolute 
measure makes no difference?

Stathis Papaioannou
Be one of the first to try Windows Live Mail.
You received this message because you are subscribed to the Google Groups 
"Everything List" group.
To post to this group, send email to
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at

Reply via email to