The Big Picture
Page 58
The Chinese Room thought experiment forces those of us who think con-
21
sciousness is purely physical to confront what a dramatic claim we are mak-
22
ing. Even if we don’t purport to have a fully fleshed- out understanding of
23
consciousness, we should try to be clear about what kinds of things could
24
possibly qualify as “conscious.” In the Chinese Room, that question is
25
raised about a pile of papers and an instruction book, but really those are
26
just colorful ways of talking about the information and processing inside a
27
computer. If we believe “consciousness” is just a way of talking about under-
28
lying physical events, what kind of uncomfortable situations does that
29
commit us to?
30
The one system we generally agree is conscious is a human being—
31
mostly the brain, but we can include the rest of the body if you like. A hu-
32
man can be thought of as a configuration of several trillion cells. If the
33
physical world is all there is, we have to think that consciousness results
34
from the particular motions and interactions of all those cells, with one
S35
another, and with the outside world. It is not supposed to be the fact that
N36
34 1
Big Picture - UK final proofs.indd 341
20/07/2016 10:02:54
T H E B IG PIC T U R E
01
cells are “cells” that matters, only how they interact with one another, the
02
dynamic patterns they carve out in space as they move through time. That’s
03
the consciousness version of multiple realizability, sometimes called sub-
04
strate independence— many different substances could embody the patterns
05
of conscious thought.
06
And if that’s true, then all kinds of things could be conscious.
07
Imagine that we take one neuron in your brain, and study what it does
08
until we have it absolutely figured out. We know precisely what signals it
09
will send out in response to any conceivable signals that might be coming
10
in. Then, without making any other changes to you, we remove that neuron
11
and replace it with an artificial machine that behaves in precisely the same
12
way, as far as inputs and outputs are concerned. A “neuristor,” as in Hein-
13
lein’s self- aware computer, Mike. But unlike Mike, you are almost entirely
14
made of your ordinary biological cells, except for this one replacement neu-
15
ristor. Are you still conscious?
16
Most people would answer yes, a person with one neuron replaced by an
17
equivalently behaving neuristor is still conscious. So what if we replace two
18
neurons? Or a few hundred million? By hypothesis, all of your external ac-
19
tions will be unaltered— at least, if the world is wholly physical and your
20
brain isn’t affected by interactions with any immaterial soul substance that
21
communicates with organic neurons but not with neuristors. A person
22
with every single one of their neurons replaced by artificial machines that
23
interact in the same way would indisputably pass the Turing test. Would it
24
qualify as being conscious?
25
We can’t prove that such an automated thinking machine would be con-
26
scious. It’s logically possible that a phase transition occurs somewhere along
27
the way as we gradually replace neurons one by one, even if we can’t predict
28
exactly when it would happen. But we have neither evidence nor reason to
29
believe that there is any such phase transition. Following Turing, if a cyborg
30
hybrid of neurons and neuristors behaves in exactly the same way as an or-
31
dinary human brain would, we should attribute to it consciousness and all
32
that goes along with it.
33
Even before John Searle presented the Chinese Room experiment, phi-
34
losopher Ned Block discussed the possibility of simulating a brain using the
35S
entire population of China. (Why everyone picks China for these thought
36N
experiments is left as an exercise.) There are many more neurons in the brain
342
Big Picture - UK final proofs.indd 342
20/07/2016 10:02:54
W h At t h I n K S ?
than there are people in China or even the whole world, but by thought-
01
experiment standards that’s not much of an obstacle. Would a collection of
02
people running around sending messages to one another, in perfect mim-
03
icry of the electrochemical signals in a human connectome, qualify as “con-
04
scious”? Is there any sense in which that population of people— collectively,
05
not as individuals— would possess inner experiences and understanding?
06
Imagine mapping a person’s connectome, not only at one moment in
07
time but as it develops through life. Then— since we’re already committed
08
to hopelessly impractical thought experiments— imagine that we record
09
absolutely every time a signal crosses a synapse in that person’s lifetime.
10
Store all of that information on a hard drive, or write it down on (a ridicu-
11
lously large number of) pieces of paper. Would that record of a person’s
12
mental processes itself be “conscious”? Do we actually need development
13
through time, or would a static representation of the evolution of the phys-
14
ical state of a person’s brain manage to capture the essence of consciousness?
15
16
•
17
These examples are fanciful but illustrative. Yes, reproducing the processes
18
of the brain with some completely different kind of substance (whether
19
neuristors or people) should certainly count as consciousness. But no, print-
20
ing things out onto a static representation of those processes should not.
21
From a poetic- naturalism perspective, when we talk about conscious-
22
ness we’re not discovering some fundamental kind of stuff out there in the
23
universe. It’s not like searching for the virus that causes a known disease,
24
where we know perfectly well what kind of thing we are looking for and
25
merely want to detect it with our instruments so that we can describe what
26
it is. Like “entropy” and “
heat,” the concepts of “consciousness” and “under-
27
standing” are ones that we invent in order to give ourselves more useful and
28
efficient descriptions of the world. We should judge a conception of what
29
consciousness really is on the basis of whether it provides a useful way of
30
talking about the world— one that accurately fits the data and offers insight
31
into what is going on.
32
A form of multiple realizability must be true at some level. Like the Ship
33
of Theseus, most of the individual atoms and many of the cells in any hu-
34
man body are replaced by equivalent copies each year. Not every one— the
S35
atoms in your tooth enamel are thought to be essentially permanent, for
N36
343
Big Picture - UK final proofs.indd 343
20/07/2016 10:02:54
T H E B IG PIC T U R E
01
example. But who “you” are is defined by the pattern that your atoms form
02
and the actions that they collectively take, not their specific identities as
03
individual particles. It seems reasonable that consciousness would have the
04
same property.
05
And if we are creating a definition of consciousness, surely “how the
06
system behaves over time” has to play a crucial role. If any element of con-
07
sciousness is absolutely necessary, it should be the ability to have thoughts.
08
That unmistakably involves evolution through time. The presence of con-
09
sciousness also implies something about apprehending the outside world
10
and interacting with it appropriately. A system that simply sits still, main-
11
taining the same configuration at every moment of time, cannot be thought
12
of as conscious, no matter how complex it may be or whatever it may repre-
13
sent. A printout of what our brain does wouldn’t qualify.
14
Imagine you were trying to develop an effective theory of how human
15
beings behave, but without any recourse to their inner mental states. That
16
is, you are playing the role of an old- time behaviorist: person receives input,
17
person behaves accordingly, without any unobservable nonsense about an
18
inner life.
19
If you wanted to make a good theory, you would end up reinventing the
20
idea of inner mental states. Part of the reason is straightforward: the sen-
21
sory input might be hearing someone ask, “How are you feeling?” and the
22
induced reaction might be “I’m a little gloomy at the moment, to be hon-
23
est.” The easiest way to account for such behavior is to imagine that there is
24
a mental state labeled “gloomy,” and that our subject is in that state at the
25
moment.
26
But there’s also another reason. Even when an individual behaves in
27
ways that do not overtly refer to their inner mental state, real human behav-
28
ior is extremely complex. It’s not like two billiard balls coming together on
29
a pool table, where you can reliably predict what will happen with relatively
30
little information (angle of impact, spin, velocities, and so on). Two differ-
31
ent people, or even the same person in slightly different circumstances, can
32
react very differently to the same input. The best way to explain that is by
33
invoking internal variables— there is something going on inside the per-
34
son’s head, and we had better take it into account if we want to correctly
35S
predict how they will behave. (When someone you know well is behaving
36N
strangely, remember: it might not be about you.)
34 4
Big Picture - UK final proofs.indd 344
20/07/2016 10:02:54
W h At t h I n K S ?
If we weren’t familiar with consciousness already, in other words, we’d
01
have to invent it. The fact that people experience inner states as well as outer
02
stimuli is absolutely central to who they are and how they behave. Inner
03
lives aren’t divorced from outer actions.
04
Daniel Dennett has made essentially this point with what he calls the
05
intentional stance. There are many circumstances in which it is useful to speak 06
as if certain things have attitudes or intentions. We therefore, quite sensibly,
07
speak that way— we attribute intentionality to all sorts of things, because
08
that’s part of a theory that provides a good account of the thing’s behavior.
09
Talking “as if” is the only thing we ever do, as there is no metaphysically dis-
10
tinct “aboutness” connecting different parts of the physical world, just rela-
11
tionships between different pieces of matter. Just as when we discussed the
12
emergence of “purpose” in chapter 35, we can think of intentions and atti-
13
tudes and conscious states as concepts that play essential roles in a higher- level
14
emergent theory describing the same underlying physical reality.
15
What Turing was trying to capture in his imitation game was the idea
16
that what matters about thinking is how a system would respond to stimuli,
17
for example, to questions presented to it by typing on a terminal. A com-
18
plete video and audio recording of the life of a human being wouldn’t be
19
“conscious,” even if it precisely captured everything that person had done
20
to date, because the recording wouldn’t be able to extrapolate that behavior
21
into the future. We couldn’t ask it questions or interact with it.
22
Many of the computer programs that have attempted to pass cut- rate
23
versions of the Turing test have been souped-up chat bots— simple systems
24
that can spit out preprogrammed sentences to a variety of possible ques-
25
tions. It is easy to fool them, not only because they don’t have the kind of
26
detailed contextual knowledge of the outside world that any normal person
27
would have, but because they don’t have memories even of the conversation
28
they have been having, much less ways to integrate such memories into the
29
rest of the discussion. In order to do so, they would have to have inner men-
30
tal states that depended on their entire histories in an integrated way, as
31
well as the ability to
conjure up hypothetical future situations, all along
32
distinguishing the past from the future, themselves from their environ-
33
ment, and reality from imagination. As Turing suggested, a program that
34
was really good enough to convincingly sustain human- level interactions
S35
would have to be actually thinking.
N36
345
Big Picture - UK final proofs.indd 345
20/07/2016 10:02:54
T H E B IG PIC T U R E
01
•
02
03
Cynthia Breazeal, a roboticist at MIT, leads a group that has constructed a
04
number of experiments in “social robotics.” One of their most charming
05
efforts was a robot puppet named Leonardo, who had a body created by
06
Stan Winston Studio, a special- effects team that had worked on such Hol-
07
lywood blockbusters as The Terminator and Jurassic Park. Equipped with 08
more than sixty miniature motors that enabled a rich palette of movement
09
and facial expressions, Leonardo bore more than a passing resemblance to
10
Gizmo from the Steven Spielberg film Gremlins.
11
The ability to have facial expressions, it turns out, is enormously useful
12
in talking to human beings. Brains work better when they’re inside bodies.
13
Leonardo interacted with the researchers in Breazeal’s lab, both reading
14
their expressions and exhibiting his own. He was also programmed with a
15
theory of mind— he kept track of not only his own knowledge (from what
16
his video- camera eyes picked up happening in front of him) but also the
17
knowledge of other people (from what he saw them doing). Leonardo’s ac-
18
tions were not all preprogrammed; he learned new behaviors through inter-
19
acting with humans, mimicking gestures and responses he witnessed in
20
others. Without knowing anything about his programming, anyone watch-
21
ing Leonardo in action could easily tell whether he was happy, sad, afraid,
22
or confused, just by observing his expressions.
23
One illustrative experiment with Leonardo was a type of false- belief
24
task: checking that a subject understands that a different person might hold
25
a certain belief even if that belief is not true. (Humans seem to develop this
26
capacity around the age of four years old; younger children labor under the
27
misconception that everyone has the same beliefs.) Leonardo watches one
28