Re: Perpetual motion via entropy disposal (was Re: effectiveperfection)

From: Eliezer S. Yudkowsky (
Date: Sun Nov 26 2000 - 23:49:02 MST

"Daniel E. Azzopardi" wrote:
> On Sun, 26 Nov 2000, Eliezer S. Yudkowsky wrote:
> > surfaces. Sigh... I either need to actually read my set of the Feynman
> > lectures, find a good physics encyclopedia, or give up physical science.
> My recommendation would be to stick with Feynman, but don't use physical
> science arguments unless they are quantitative and/or predictive.

I'm afraid that I do not possess the skill to use physical arguments
quantitatively, except to make complexity-theory type statements
("exponential", "linear", "logarithmic", and so on). As for using them
predictively, I'm not sure what you mean by this; if you mean what I would
mean by it - that every statement must have observable consequences and
must be a consequence of rules deduced from observations - then I very
strongly agree.

> > Paul Davies sez: "It is easy to dream up scenarios that produce
> > unphysical or paradoxical consequences--building perpetual motion
> > machines, or even travelling backwards in time. To physicists, these are
> > alarming notions."
> > Who knows, Davies may turn out to be right; but even so, this strikes me
> > as the sort of thing that future historians call a 'blind spot', like the
> > unbreakability of the sound barrier or the impossibility of
> > heavier-than-air flight. "Despite numerous demonstrations that their own
> I think Physicists of our time are generally are much less willing to
> dismiss things out of hand (alarming notion!=physical impossibility).

I see physicists of our time using their "alarming notion" indicator
predictively; e.g., they are saying that because certain things alarm
physicists, they must alarm the Universe as well. If a naked singularity
is possible, then there must be some kind of effect that rules out the
formation, and so on. They then go looking for the effect. Sometimes
this works, in which case I have no objection. Certainly, faced with a
classical-physics proposal for perpetual motion, I would start with my
preconceived notion that classical perpetual motion is impossible, and
work backwards from there.

Ultimately, however, it is quite possible that - however alarming a naked
singularity is to physicists - there will simply *be* naked singularities
anyway. Some physicists seem to be crossing the line between a valuable
heuristic, which is one thing, and a state of mind in which there is no
longer any value in going and looking for naked singularities or trying to
build them.

> However, I must take issue with the analogies you make here, especially as
> you are either glossing over, or overlooking things which make these
> different kinds of "barriers", rather than different magnitudes. The sound
> barrier/heavier than air arguement is oft quoted w.r.t. travel at speeds
> greater than c (tantamount to time travel, which I'll come to).

Please note that *nowhere* do I make the comparision between
faster-than-sound and FTL; believe me, I find the comparision as odious as
you do. FTL is not a speed limit. It is built into the structure of
causality. You cannot go faster than light, globally, except by warping
space so that you are still travelling slower than light, locally. And
you cannot warp space without permitting time travel. Now, it so happens
that I think time travel is possible. But to actually go faster than
light would require outright magic.

The "local FTL" limit is imposed by the Universe itself. The Universe
speaks to us, and it defines causality in a way that is counterintuitive;
it says that there is no space of simultaneity, and that the structure of
causality rules out FTL. This was an alarming notion to the physicists of
the time - though it doesn't alarm me at all, since I grew up decades
later. What we have now is the Universe speaking through General
Relativity, saying that a single correct direction of time is prohibited;
we have numerous descriptions of systems which may be impossible to
*construct*, such as wormholes, but which still contain internally
consistent CTCs, and we have a bunch of physicists calling closed timelike
curves an "alarming notion". I agree that the notion of killing your own
grandfather alarms physicists, but we have absolutely no experimental
reason to believe that the notion alarms the Universe. It conflicts with
our notions of causality, but our notions of causality are demonstrably
not the same notions of causality possessed by the Universe. So when I
hear someone calling a closed timelike curve an "alarming notion", it
sounds to me like a very shaky assumption.

> No
> physical theory said faster than sound flight was *physically* impossible.
> There is a theory which says faster than light travel is physically
> impossible though (Relativity, both general and special). Though there are
> some (theoretical, not observed) metrics which appear to violate this
> locally, globally flat space time is a pretty good approximation.

Isn't that assuming what you want to prove? Let's say that you construct
a wormhole between Sol and Proxima Centauri; space is then no longer
globally flat. It might be better to say that astronomical observations
show no evidence that the metric of spacetime in other galaxies, to the
limits of observation, differs from the observed local metric - but that
says nothing about whether one solar system contains a time machine.

> Flat
> space time is special relativity(SR), and whatever way you write it, and
> however you do it, globally SR has never been violated and there is no
> theoretical way to violate it either - QM obeys it, even including EPR
> paradoxes and mesoscopic systems.

What is the speed of propagation of state-vector reduction? Never mind;
the refusal of most physicists to think about state-vector reduction as a
real physical phenomenon, rather than a philosophical or observational
one, is a totally different blind spot.

> So - in contrast to the vague reasons given for the limits to
> human technology as regards supersonic flight, reliable(*) physical theory
> has given us concrete physical limits to our universe, as we understand it
> now. There is no loophole without an effect that has yet to be discovered.
> Let me say that again another way: though we may construct theories that
> don't disallow certain things (eg: time travel) these theories do not
> belong to the minimal consistent set of physical theories.
> < (*) aside: how reliable is SR? How many tests of SR do we do each day
> that CERN / SLAC / FermiLab run? Millions? How many times has SR been
> wrong? Never. Would we see it if it was wrong just once? Oh yes!>

Again, I do not dispute Special Relativity. I do not ever expect to see
Special Relativity violated. The reason I do not believe in Greg Bear's
speculations about "descriptor theory" is that it would require a global
space of simultaneity, which is Just Not Allowed. I "get" Special

> Time travel: We need accept only one of: universal flat space time,

Isn't that sort of like a single space of simultaneity? Spacetime is
*locally* flat (no naked singularities, please!). Globally, the
lightcones point in all kinds of different directions, and you can pick
any individual light cone and declare it to be This Side Up; the rest will
rotate into place while preserving all observed physical laws, just as
Special Relativity says you can pick any space of simultaneity you like.
The spirit of General Relativity - as *opposed* to our untrustworthy
intuitions about causality - mediates as strongly against declaring a
particular, "objectively" correct direction of causality, as Special
Relativity mediates against an objective space of simultaneity.

> upper
> speed limit for information transfer=c,

Locally? Sure. Globally? Another question entirely.

> or causality,

By "causality", I presume you mean the set of intuitions about cause and
effect that have evolved to be present within the mind of the theorizing
physicists? One can hardly advocate that such intuitions be sovereign
over experiment, or even over the particular predictions of current

> to run into immense
> difficulties constructing a theory which allows time travel. In fact, with
> out accepting causality we run into immense problems constructing any
> logical theory.

*This* is the blind spot. I don't believe in what you call "causality".
I think that when you say "causality", you are referring to a set of
preconceptions which will eventually be disproved by the experimental
construction of a closed timelike curve.

> > theories permitted the possibility of time travel / negative energy /
> > naked singularities etc., twentieth-century physicists still refused to
> > believe, calling them 'alarming notions'." This rant is more about time
> > travel then about thermodynamics, but real physical laws don't look like
> > they're about to be violated every second Tuesday. Sometimes physics
> Thermodynamics is something you should be concerned with. Locally, at
> scales <10^-10 m (say) physics is CPT invariant, yet the emergent
> properties of matter are anything but (as you have stated). If you hold
> that intelligence is an emergent property of correctly /near optimally
> arranged matter, than one would have thought we're going to need to be
> able to understand the transistion phase from quantum to statistical
> mechanics as the best modelled physical analogue we have.

I disagree. If the functional elements of an intelligence do not depend
on details of the quantum-mechanical level, an understanding of the
transition phase is not practically necessary to constructing
intelligence. To the extent that the functional elements of intelligence
do not depend on details which are *only* achievable at the
quantum-mechanical level, understanding of physics is not theoretically
necessary to constructing intelligence.

> > "permissible fluctuations in entropy", but did you ever hear of a
> > permissible fluctuation in mass-energy?
> Vacuum pair creation, on short time scales.

I stand corrected, then; you can get away with it for sufficiently small
values of delta t.

> > If current physical theory offends a physicist, so what? God does play
> > dice with the Universe, Einstein, get over it. Even the more imaginative
> > physicists, like David Deutsch, suffer from the same disease, confining
> Physicists must work in science fact, not fiction. For example, we have
> several candidates for a GUT. Which one do we choose? We can choose our
> favorite, we can choose the one that says every particle is really an n-th
> dimensional donut because it lets us do something we'd really like to do
> (eat donuts), but all that means nothing if we don't have physical
> evidence for the theory.

GUTs are epicycles. GUTs are another pet peeve of mine. Who moved these
people's cheese? Why are they wasting time - I won't say "wasting money",
since "a physicist's salary is never wasted", but why are they wasting the
far more precious resource of physicists' time - on searching for
particles? We will see no new interesting theories and no new interesting
technology as a result of applying the reductionistic paradigm that took
us down to the quark level, because that is no longer where the real
questions are. GUTs throw hugely increasing amounts of complexity in
exchange for vanishingly small returns of prediction, and this should be
enough to tell us that GUTs are on the wrong track.

I, Eliezer Yudkowsky, being of sound mind, do make this testable
prediction: The next major discovery in physics - if there are any major
advances in physics pre-Singularity - will be on the order of the
Michelson-Morley experiment, and it will occur when somebody constructs a
simple, obvious in retrospect, and relatively cheap experiment that
distinguishes when state-vector reduction occurs in a system. The next
major discovery will not be in the tenth decimal place; it will cover one
of the huge, gaping holes in our current description of the Universe.

> No, it isn't. GR must obey SR over large enough distances

That is not something I've ever heard. Why?

If true, it would seem to completely rule out constructs such as wormholes
and Tipler cylinders, which may be impossible to *construct* due to Cosmic
Censorship, but which do have descriptions that appear *internally*
consistent under current physical law.

> Sorry, but that is faulty reasoning. If you
> did understand GR completely (which you don't), you still would have no
> right to say "Ah, but we can do this in our model (if we push all our
> paremeters to one corner of the phase space) therefore we can do this in
> reality". In addition, you are attempting to use a technology
> (mathematical constructs around GR) to fault the apparent narrow
> mindedness of physicists who use and develop that technology as their own
> tool. To do that, you better be on a very firm foundation - and it's clear
> you're not.
> > Oh, never mind. I'm not a physicist. You can get personally annoyed with
> > physical theories, but you're supposed to buy the right to do so with a
> > physics doctorate. This system works so well generally that I would have
> > no real objection to being slapped down by it personally. Not my job.
> > </RANT>
> I like this though ;-)

Yes, I thought you would.

One of the ways that I can prevent conflicts with accepted contemporary
theory is by making sure that at least one physicist agrees with me about
any particular point. That modern physicists have a "blind spot" about
time travel is my own, causal interpretation of a reason why they make
certain statements, but it is also a fact that a significant minority of
physicists do think that time travel is possible; I thus have a reasonable
expectation that I am not making an absurd statement when I say that time
travel is not prohibited by current physical law. Similarly, when I say
that state-vector reduction is a real physical event and not an observer
effect or a philosophical way of describing a system, I am agreeing with,
e.g., Sir Roger Penrose, which doesn't mean that I am necessarily right,
but it does hopefully mean that I am not making a statement in material
conflict with existing physics.

> You can't just say that, since there are no observational consequences of
> the many worlds model. You are ruling something out in almost the same way
> as I'd rule out the class of GR metrics that allow time travel, except one
> senario has testable experimental consequences, the other doesn't ;-)
> Interpretation of QM is philosophy.

No, it is not. A system in which state-vector reduction, e.g. a
measurement, occurs, evolves differently than a system in which no
measurement occurs. That's the basic "paradox" (thing our intuitions
don't like) about quantum theory; if you measure which slit a photon
passes through, it will wind up in a different place, not because of
Heisenberg or anything, but because superposed states can cancel out as
well as adding up. A many-worlds theory states that no superposed state
ever dies, and we simply wind up in a randomly selected one of them. If
so, why can we correctly make the prediction that certain superposed
states will cancel out?

> > quantum collapse were an observational-effect illusion, it would make no
> > difference what the probability amplitudes were; we would have the same
> > observed chance of winding up in a .2 probability Universe as a .8
> No, the probabilities would still be different.

If all the superposed states exist and no quantum collapse occurs, so that
a .2 amplitude and a .8 amplitude exist equally, with the world branching
to go both ways, then why do we, the observer, make the prediction that we
have a .64 probability of appearing to go down the first branch and only a
.04 probability of going down the second branch?

> Hmmm.. I've gone on rather longer than I'd hoped, but the points I've
> attempted to make are these:
> If one is going to use physical theories for some thing concrete
> (ie: predictive) then use the theory correctly - don't use a theory for an
> analogy, use it to calculate, that is what it is there for. Entropy is a
> precisely defined quantity for example (but be aware of the tacit
> assumptions made in defining a quantity).

Entropy is observer relative; that is, depending on how much information
an external system has about an internal system, the apparent "entropy" of
that system may change. There are some precisely defined properties, like
"space" or "time", that are not observer-independent - but "entropy"
depends on the observer's state of mind.

> If one is going to speculate about science fiction (a possible
> effect which hasn't been observed yet, that would have "scientific"
> reprecussions) then speculate, but don't do it within the framework of the
> general consensus, since those theories were formulated to explain only
> observed effects. Certainly one shouldn't use the fact that an
> interpretation of an individual theory allows something to conclude that
> that something exists... minimise your assumptions.

In essence, what I am doing is not - from my own twisted perspective -
speculating; rather, I am calling attention to the totally unjustified
speculation of many current physicists that "closed timelike curves" are
hated by the Universe. In fact, it was *in* science fiction that I first
found the speculation that the Universe hates a paradox - but we have *no*
justification for this! It is something that would be convenient for
physicists if true, but nobody has any idea whether it's true! The
Universe may indeed *turn out* to hate closed timelike curves, but I see
nothing that allows us to conclude this from current evidence. I do not,
Ifni forbid, insist that - for example - Special Relativity be modified to
permit local FTL. I do, however, get disturbed when I hear a physicist
objecting to an innocent theory on the grounds that it would permit
"closed timelike curves"; this person is making an assumption which has
not been justified. Calling it a "consequence of the logical definition
of causality" horrifies me. It sounds like a reversion to Plato. A
logical definition proves nothing.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:35 MDT