The Big Picture
Page 17
92
Big Picture - UK final proofs.indd 92
20/07/2016 10:02:41
01
02
12
03
04
Reality Emerges
05
06
07
08
09
10
11
W
12
ith our Bayesian knowledge- building tool kit in hand, we can
13
return to fleshing out some of the ideas behind poetic natural-
14
ism. In particular, the innocuous- seeming but secretly pro-
15
found idea that there are many ways of talking about the world, each of
16
which captures a different aspect of the underlying whole.
17
The progress of human knowledge has bequeathed to us a couple of in-
18
sights that, taken together, suggest a world that is profoundly different from
19
the picture we construct from our everyday experience. There is conserva-
20
tion of momentum: the universe doesn’t need a mover; constant motion is
21
natural and expected. It is tempting to hypothesize— cautiously, always with
22
the prospect of changing our minds if it doesn’t work— that the universe
23
doesn’t need to be created, caused, or even sustained. It can simply be. Then
24
there is conservation of information. The universe evolves by marching from
25
one moment to the next in a way that depends only on its present state. It
26
neither aims toward future goals nor relies on its previous history.
27
These discoveries indicate that the world operates by itself, free of any
28
external guidance. Together they have dramatically increased our credence
29
in naturalism: there is only one world, the natural world, operating accord-
30
ing to the laws of physics. But they also highlight a looming question: Why
31
does the world of our everyday experience seem so different from the world
32
of fundamental physics? Why aren’t the basic workings of reality perfectly
33
obvious at first glance? Why is the vocabulary we use to describe the every-
34
day world— causes, purposes, reasons why— so different from that of the
S35
microscopic world— constant motion, Laplacian patterns?
N36
93
Big Picture - UK final proofs.indd 93
20/07/2016 10:02:41
T H E B IG PIC T U R E
01
This brings us to the “poetic” part of poetic naturalism. While there is
02
one world, there are many ways of talking about it. We refer to these ways
03
as “models” or “theories” or “vocabularies” or “stories”; it doesn’t matter.
04
Aristotle and his contemporaries weren’t just making things up; they told
05
a reasonable story about the world they actually observed. Science has dis-
06
covered another set of stories, harder to perceive but of greater precision and
07
wider applicability. It’s not good enough that the stories succeed individu-
08
ally; they have to fit together.
09
•
10
11
One pivotal word enables that reconciliation between all the different stories:
12
emergence. Like many magical words, it’s extremely powerful but also tricky
13
and liable to be misused in the wrong hands. A property of a system is “emer-
14
gent” if it is not part of a detailed “fundamental” description of the system,
15
but it becomes useful or even inevitable when we look at the system more
16
broadly. A naturalist believes that human behavior emerges from the complex
17
interplay of the atoms and forces that make up individual human beings.
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35S
36N
The Starry Night. (Painting by Vincent van Gogh)
94
Big Picture - UK final proofs.indd 94
20/07/2016 10:02:41
R E A l I t y E M E R g E S
Emergence is ubiquitous. Consider a painting, such as van Gogh’s The
01
Starry Night. The canvas and paint constitute a physical artifact; on one
02
level, it is just a collection of certain atoms in certain locations. There is
03
nothing to the painting other than those atoms. Van Gogh didn’t infuse it
04
with any form of spiritual energy; he put the paint onto the canvas. If the
05
atoms making up the paint had been put in different locations, it would
06
have been a different painting.
07
But it’s obvious that specifying an arrangement of atoms isn’t the only way
08
of talking about this physical artifact, and it’s not even the best way for most
09
purposes. When we talk about The Starry Night, we refer to the color palette,
10
the mood it evokes, the swirling of the moon and stars in the sky, and perhaps
11
to van Gogh’s period in the asylum at Saint- Paul de Mausole. All of these
12
higher- level concepts are something in addition to a dry (but accurate) list of
13
all the atoms that make up the paint. They are emergent properties.
14
The classic example of emergence, one you should constantly return to
15
whenever these things get confusing, involves the air in the room around
16
you. That air is a gas, and we can speak of it as having various properties: a
17
temperature, a density, a humidity, a velocity, and so on. We think of the
18
air as a continuous fluid, and all of those properties take on numerical val-
19
ues at every point in the room. (Remember that gases, like liquids, are flu-
20
ids.) But we know that the air isn’t “really” a fluid. It we look at it very
21
closely, down at a microscopic level, we see that it’s composed of individual
22
atoms and molecules— mostly nitrogen and oxygen, with trace bits of other
23
elements and compounds. One way of talking about the air would simply
24
be to list every one of those molecules— perhaps 1028 of them— and specify
25
their positions, velocities, orientations in space, and so on. This is some-
26
times called kinetic theory, and it’s a perfectly legitimate way of talking.
27
28
29
&nbs
p; 30
31
32
33
34
S35
Two ways of thinking about air: as a collection of discrete molecules, or as a smooth fluid.
N36
95
Big Picture - UK final proofs.indd 95
20/07/2016 10:02:41
T H E B IG PIC T U R E
01
Specifying the state of each molecule at every moment in time is a consistent
02
and self- contained description of the system; if you were as smart as Laplace’s
03
Demon, that would be enough to determine the state at any other time. In
04
practice it’s incredibly cumbersome, and nobody ever talks that way.
05
Describing the air in terms of its macroscopic fluid properties such as
06
temperature and density is also a perfectly legitimate way of talking. Just as
07
there are equations that can tell us how the individual molecules bump into
08
one another and move over time, there are separate equations that tell us
09
how the fluid parameters evolve over time. And the good news is, you don’t
10
need to be nearly as smart as Laplace’s Demon to actually find the solution;
11
real computers are completely up to the task. Atmospheric scientists and
12
aeronautical engineers solve such equations every day.
13
So the fluid description and the molecular description are two different
14
ways of talking about the air, both of which— at least in certain
15
circumstances— tell very precise and useful stories about how air behaves.
16
This example illustrates a number of features that commonly appear in dis-
17
cussions of emergence:
18
19
• The different stories or theories use utterly different vocabu-
20
laries; they are different ontologies, despite describing the
21
same underlying reality. In one we talk about the density,
22
pressure, and viscosity of the fluid; in the other we talk about
23
the position and velocity of all the individual molecules.
24
Each story comes with an elaborate set of ingredients—
25
objects, properties, processes, relations— and those ingredi-
26
ents can be wildly different from one story to another, even
27
if they are all “true.”
28
29
• Each theory has a particular domain of applicability. The fluid
30
description wouldn’t be legitimate if the number of mole-
31
cules in a region were so small that the effects of particular
32
molecules were important individually, rather than only in
33
aggregate. The molecular description is effective under wider
34
circumstances, but still not always; we could imagine pack-
35S
ing enough molecules into a small enough region of space
36N
96
Big Picture - UK final proofs.indd 96
20/07/2016 10:02:41
R E A l I t y E M E R g E S
that they collapsed to make a black hole, and the molecular
01
vocabulary would no longer be appropriate.
02
03
• Within their respective domains of applicability, each theory
04
is autonomous— complete and self- contained, neither relying
05
on the other. If we’re speaking the fluid language, we describe
06
the air using density and pressure and so on. Specifying those
07
quantities is enough to answer whatever questions we have
08
about the air, according to that theory. In particular, we don’t
09
need to ever refer to any ideas about molecules and their
10
properties. Historically, we talked about air pressure and ve-
11
locity long before we knew it was made of molecules. Like-
12
wise, when we are talking about molecules, we don’t ever
13
have to use words like “pressure” or “viscosity”— those con-
14
cepts simply don’t apply.
15
16
The important takeaway here is that stories can invoke utterly different
17
ideas, and yet accurately describe the same underlying stuff. This will be
18
crucially important down the line. Organisms can be alive even if their
19
constituent atoms are not. Animals can be conscious even if their cells are
20
not. People can make choices even if the very concept of “choice” doesn’t
21
apply to the pieces of which they are made.
22
23
•
24
If we have two different theories that both accurately describe the same
25
underlying reality, they must be related to each other and mutually consis-
26
tent. Sometimes that relationship is simple and transparent; other times we
27
just have to trust that it’s there.
28
The case of fluid dynamics emerging from molecules is as simple as it
29
gets. One theory can directly be obtained from the other by a process
30
known as coarse- graining. There is an explicit map from one theory (mole-
31
cules) to the other (fluid). A particular state in the first theory— a list of all
32
the molecules, their positions, and velocities— corresponds to some par-
33
ticular state in the second one— a density and pressure and velocity of the
34
fluid at every point.
S35
N36
97
Big Picture - UK final proofs.indd 97
20/07/2016 10:02:41
T H E B IG PIC T U R E
01
Moreover, many different states in the molecular theory get mapped
02
to the same state in the fluid one. When this is the case, we often call
03
the first theory the “microscopic” or “ fine- grained” or “fundamental” one,
04
and the second the “macroscopic” or “ coarse- grained” or “emergent” or “ef-
05
fective” one. These labels aren’t absolute. To a biologist working with an
06
emergent theory of cells and tissue, the theory of atoms and their interac-
07
tions might be a microscopic description; to a string theorist working on
08
the quantum theory of gravity, superstrings might be the microscopic enti-
09
ties, and atoms are emergent. One person’s microscopic is another person’s
10
macroscopic.
11
We want our theories to give physical predictions that are consistent
12
with each other. I
magine that a state x in the microscopic theory evolves
13
into some state y. And imagine that the “emergence” map sends x and y to 14
states X and Y in the emergent fluid theory. Then it had better be the case 15
that X evolves to Y under the rules of the emergent theory, at least with very 16
high probability. Starting with a microscopic state, the process “evolve for-
17
ward in time, and see what that corresponds to in the emergent theory”
18
should give the same answer as “see what it corresponds to in the emergent
19
theory, then evolve forward in time.”
20
states in the
states in the
21
microscopic theory
emergent/effective theory
22
23
evolution
through
Y
24
y
time
25
26
27
x
28
X
29
30
emergence
31
Emergence of one theory from another. Boxes in each image represent different possible 32
states the entire system could be in, as described by each theory. Time evolution and emer-33
gence should be compatible: microstates that map to the same emergent state should
34
evolve into microstates that also map to the same emergent state. Several microstates map 35S
to each emergent state.
36N
98
Big Picture - UK final proofs.indd 98
20/07/2016 10:02:42
R E A l I t y E M E R g E S
Coarse- graining goes one way— from microscopic to macroscopic— but
01
not the other way. You can’t discover the properties of the microscopic
02
theory just from knowing the macroscopic theory. Indeed, emergent theo-
03
ries can be multiply realizable: there can, in principle, be many distinct
04
microscopic theories that are incompatible with one another but compati-
05
ble with the same emergent description. You can understand the air as a
06
fluid without knowing anything about its molecular composition, or even
07
if there is a description in terms of particles at all.
08
The reason why emergence is so helpful is that different theories are not
09
created equal. Within its domain of applicability, the emergent fluid theory
10
is enormously more computationally efficient than the microscopic mo-
11
lecular theory. It’s easier to write down a few fluid variables than the states
12
of all those molecules. Typically— though not necessarily— the theory that
13
has a wider domain of applicability will also be the one that is more com-