[Haskell-cafe] Quanta. Was: Wikipedia on first-class object

jerzy.karczmarczuk at info.unicaen.fr jerzy.karczmarczuk at info.unicaen.fr
Sun Dec 30 06:27:32 EST 2007


Ketil Malde writes: 

> I guess I should disclaim the rest of my post right away: I don't know
> much about quantum anything, beyond what I read in the newspapers.

I answer here, but there were other contributions, of Ryan Ingram, Miguel
Mitrofanov, and Andrew Bromage, which I acknowledge. People, we are
reopening (recursively) a Pandora box, and this discussion will lead
nowhere, as always, when people speculate whether the Universe is a Turing
machine... 

I would like to point out that it started on a subject NOT related to
computing, but to simulation. This was, btw. one of the points Feynman
stressed upon in his talks on the relation between quanta and information
processing. 

But Feynman was a physicist, and we have different sensibilities from a
typical computer scientist, for whom the Universe is a specific "model".
For us, things change, you speak about the "state transformers". A particle
moves, and some Turingard will say that this is just some data processing.
Some say that "thinking is computing", which for me is stupid as hell,
since nobody really knows what thinking is. Frankly, reducing the world to
computerese is as sensible, as works of La Mettrie (1701-1751) about the
human soul as *mechanism*. 

Ketil, you say about my objection that we can simulate quanta: 

> You could raise the same argument for (digital) computers compared to
> brains - although my brain might be able to, it's not practical for it
> to do the computations performed even by simple computer programs. 
> 
> But the difference is quantitative and practical, not qualitative and
> theoretical.  (Arguably, I know. I invoke Occam.)

No, it is qualitative and theoretical. You can simulate *some models of
quantum systems*, not the quantum reality, since we have simply no idea
what is the quantum information, and how to cope with the non-separability
(EPR). This implies a non-modularity of the simulating programs. If you
manage to *really* split an entangled system, and send one half of it to
another galaxy, you will *really* face the EPR paradox, an experiment on
Earth conditions the issue of the measurement faraway. But physically, the
experiments there are independent. So, in a simulation, the measurement here
should do something horrible with the random number generator used to
generate the measurement instance faraway. You have a kind of particularly
nasty side-effect. (BTW. Amr Sabry really reasons in terms of these side-
effects, although he tries to be purely functional...) On the other hand,
a physicist will tell you, that in Nature there cannot be any "side-effects"
in quanta, the unitarity forbids them. In two words, a simulator of a
quantum system becomes fast as complex as the simulated system itself... 

> If I understand correctly, a quantum computer might solve problems in
> NP [conditions discussed by others...]

> As far as I can tell, it doesn't imply the
> ability to compute anthing that wasn't computable before.

Bother...
Look, ALL, ABSOLUTELY ALL what all those "computationalists" (I call them
Turingards, which is a very impolite term, look into the French dict.
what is the meaning of "ringard"), *reduce* the behaviour to information
processing, but Nature does not process information, whatever you may say
about it. This is *our* interpretation of physical phenomena. The
"information" is a distilled concept. Of course, there is entropy, whose
relation to information is extremely profound. But Nature does not solve
equations, nor implements algorithms. Nature does not compare things for
equality; on the other hand it "has" some equalities built-in, such as the
*true* indistinguishability of quantum particles, which invalidates all
classical combinatorics of the state counting, and changes the entropy.
I still don't know whether anybody knows what is the relation between the
entropy of quantum systems and their "informational contents"... 

Again, think a bit about the simulation, not about the computability. You
cannot even simulate a classical system, frankly. There is chaos, which
requires an infinite real-number precision, to make the simulation "good". 

In relativistics you are in a bad shape. Try to simulate a black hole, with
a complete decoupling between an outer world frame, in which an object
takes an infinite time to fall through the event horizon, and the victim,
for whom it is finite. 

> Now if we can wrap up the topic of whether machines can think, next
> session we'll discuss whether ships can swim.

I don't understand your point. We know what swimming is: floating and
moving autonomously. Thinking is different, since our thinking is (at least
for some of us) conscious, and we have no idea what is the conscience.
For goodness sake, I have *REALLY* the impression that those guys who
speak about computability of the Universe, have the mentality of 18 century
thinkers for whom the world was simple and mechanistic. Or even the
mentality of people contemporary of Democritus, for whom everything
"reduced" to some dance of atoms. 

I am as sure as I can modestly be, that in less than 50 years people will
laugh loud about the current ideas of "brain as computer", as we laugh
now about the 18 century attempts to reduce living organisms to mechanical
automata. Well, OK. /cum grano salis/: we *may* say that organisms are
automata. The point is that is just a phrasing with absolutely no meaning
nor practical consequences. 

Jerzy Karczmarczuk 




More information about the Haskell-Cafe mailing list