Re: Time Machines, FTL, and P=NP

Michael Ash wrote:
> James Burns <burns.87@[EMAIL PROTECTED]
> wrote:
> 
>>Actually, I do think that my free will would go away in
>>certain causality-violating circumstances. But we may not be
>>in disagreement (about this point) because we may not be agreeing
>>about what free will is, etc.
> 
> I think you're right. We agree on what happens, just not on
> what it means. I as simply a means for discovering the best
> way to pursuade someone. You see it as, basically, a quantitative
> difference so large that it becomes qualitative. That is, a
> machine which gives you the perfect means of pursuading someone
> means that you circumvent their free will.
> 
> My argument against your position would be that there are
> *probably* situations where a person cannot be pursuaded by
> any means at your disposal. In that case your machine will either
> fail, meaning they have free will, or it will bend their mind,
> meaning that it doesn't work the way that we've discussed.

Okay. If the machine fails, then we're back to square one; no
causality violation, no loss of free will. I have a hard
time imagining the person could not be persuaded /in principle/,
because it seems to me I could also /in principle/ select for
futures in which she turns into a potted plant or a whale.

However, I think I need to use a more nuanced model of this
whole future-selection-by-causal-loop scheme.

If the different futures were horses, then the causal-loop
scheme would be to shoot the favorite, then the next-favorite,
and then on down the line, until we get to an old beat-up
nag (sure to lose in any fair race) who satisfies the universe's
consistency condition. The successful nag could do this by
satisfying /my/ goal, maybe by having a cosmic ray kill
the correct neuron in my lady fair's brain at just the
correct moment. It could do this by altering my own brain
chemistry so that I think (and only think) that I have had
a fantastic night. It could also have the correct bits toggle
in the message I send back to myself to read "Whoo hoo!
Line 137 is a winner." And many many other possible
ways to make the universe consistent.

All of these nags were increadibly unlikely to win the race
without my time machine. However, some of the incredibly
unlikely outcomes will be much more unlikely than other
incredibly unlikely outcomes. These even-more-unlikely
outcomes will (probably!) not be the way the universe
turns out consistent. The horse that will win will be one
of the first that I come to that satisfy consistency.

The problem I see in the whole scheme is the difficulty
in judging the relative likelihood of fantastically unlikely
outcomes. Which is more likely: that she will have
a convenient change of mind or that I will have a convenient
change of memory? I dunno.

There are some rules we can deduce though. The mechanism
should be as robust as possible. The condition that you're
selecting futures for should NOT be judged by you, and
should be judged in as robust as possible, too, in the same
sense as the causal loop mechanism.

One thing I worry about is having my future being made
consistent by a massive solar flare or something. I don't see
how this can be avoided entirely, since eliminating
the future-ward end of the loop will always stop the
paging through futures right at the one with the disaster.
I suppose one should put the future-ward end of the mechanism
in a bomb shelter or something, where it could survive
long enough say "Not this one!"

If you limit what you ask for to something unlikely but
still extremely more likely than, for example, all the
radioactive atoms near the mechanism decaying at once,
then I think you'll be okay. (Though how will you know
until you've tried?)

I think we're in agreement on a lot here, although I
don't think of using the causal loop as "bending their mind"
I think of it as selecting one of the the vanishingly
rare futures in which she would have changed her mind
on her own. Probably just a difference in terminology,
but let me point out that your way of putting things
implies some sort of influence flowing from me to her
that will just not be found, if looked for.

[...]
> Pretend for a moment that we're not talking about time
> machines, but instead about faeries. The discussion is
> based around the question of what we could do if faeries
> exist. Could we build faerie-powered starships?
> Faerie-powered lockpicks? What design would you use to
> construct a 200 mile-per-gallon faerie-catalyzed Cadillac?
> 
> It may be fun speculation but from a scientific standpoint
> it's all garbage. The term isn't defined well enough to
> discuss it in any context other than fun.
> 
> Thus it is with causality, I think. Fundamentally it's
> outside our understanding, so discussion of the consequences
> is necessarily non-scientific.

Sorry, I don't see causality as fundamentally outside our
understanding. I understand that talk using the word
"causality" and its cousins has been frowned upon for much of
the twentieth century, but it looks to me as though the
actual use of the ideas never went away.

There has been some very good work in dealing with causality.
Judea Pearl's home page, highly respect researcher in this area
http://bayes.cs.ucla.edu/jp_home.html
A couple of "gentle intruductions"
http://singapore.cs.ucla.edu/LECTURE/lecture_sec1.htm
http://singapore.cs.ucla.edu/IJCAI99/index.html

I understand that there is some controversy to the idea that
everything we need about causality can be modeled by Bayesian
nets, but the argument that there is something uncapturable
about causality reminds me too much of all the arguments
that /consciousness/ has something uncapturable.
http://www.uni-konstanz.de/FuF/Philo/Philosophie/Spohn/preprints/pdf/PS62-Spohn.pdf
Bayesian Nets Are All There Is To Causal Dependence

--------------------
Sorry, I think I may have been perfectly unclear in this last
bit about causality and Bayesian nets. Let me just say
I think you'll find the references very interesting.

Jim Burns