<<

. 3
( 4 .)



>>

in more than one way. In (a), the image can be seen as either a
black vase or two white faces. In (b), either a young woman or
an old woman can be seen. However, you cannot “see” both
interpretations at the same time; your mind is always locked
onto one or the other. At any particular instant the figures are
not ambiguous; they are a consistent representation of what you
believe you are seeing. You see the vase or two faces; you see
a young woman or an old woman. Even though the data
entering your brain is ambiguous, your instantaneous conscious
experience of the image is not ambiguous. Your brain has
scoured the incoming data for a match. When found, you are
conscious only of the consistent features of the match, not the
inconsistent features of the raw data.
The Inner Light Theory of Consciousness
132

Let™s look at an example to show just how powerful the
approach of “matching” is. The images in Fig. 8-8 were created
by degrading pictures of three common scenes, all of which you
would immediately recognize. The resulting image quality is so
poor that they hardly look like pictures at all; they seem more
like random ink blots. Suppose we conduct an experiment
where we show these three degraded figures to a group of 100
people and asked them to identify the pictures. How many
correct responses would we expect? Of course, the answer is
zero; these images are so poor that it would be impossible for
anyone to do much better than guessing.
But now suppose that we redo the experiment with one
significant change; we make it a multiple choice test. We start
by telling our subjects that the three original images were (1)
Abraham Lincoln, (2) a sunset, and (3) the Eiffel tower, in no
particular order. We again ask them to identify each picture,
using this additional information. After looking for a few
moments, all 100 of our subjects come up with the correct
answers. In other words, by narrowing the choices we have
enormously improved the ability to identify patterns in
ambiguous, incomplete, and noisy data. As in this example, we
have changed a task that was virtually impossible, into one that
can be carried out with perfect reliability.

The Subreality Machine in Operation
How does this relate to an inner reality? When we move
around in the world, our brains are flooded with raw
information from the senses. This data stream is so large, and
such poor quality, that it would be impossible for the brain to
analyze it for every possible pattern. The brain is simply not
powerful enough to do this. For instance, suppose you walk into
an office building for the first time. Your brain is suddenly
inundated with information from your eyes and ears about the
new environment. It responds by searching these data for what
it expects to find, desks, chairs, people, computers, telephones,
carpeting, and so on. When a match is found, the brain labels
Chapter 8: The Function of the Subreality Machine 133




FIGURE 8-8
Degraded images. These images
cannot be recognized as they are.
However, it is a simple task to
match them with the original
photographs from which they
were derived: The Eiffel tower,
Abraham Lincoln, and a sunset
(clockwise from top-left).




it, and then moves onto portions of the raw data that have not
been recognized. This continues until the brain believes it
understands the surroundings well enough to carry out its
planed activities. And none of this is surprising; it is not much
more than the common sense view of how our minds work.
But now let™s reexamine this process using an additional
assumption. We have already discussed how the analysis of
sensory information is enormously difficult. Of course, this is
a relative statement; it is “enormously difficult” compared to
The Inner Light Theory of Consciousness
134

what? The assumption we will make is that sensory analysis is
difficult according to two criteria, the brain™s computational
power and its memory capabilities.
To understand the first of these, imagine you see a chair
when you walk into the new office. How long does it take you
to recognize it as a chair? Of course, this happens very quickly,
perhaps a tenth of a second. But how long would it take you to
recognize it as one very specific chair, say, one that was part of
your family™s furniture when you were growing up? Since this
is a more difficult task, it will take much longer, perhaps a few
seconds. This is important because we live in a world where
critical movements need to be made in a fraction of a second.
If it took you a few seconds to identify a nearby alligator, you
would be his lunch! The point is, the time it takes to complete
a mental task depends on the difficulty of the task and the
computational power of the brain. When we say that “sensory
analysis is enormously difficult compared to the brain™s
computational power,” we are commenting on the types of
mental tasks that can be carried out within a fraction of a
second. Specifically, within this key time constraint, we can
sort objects into general categories, but not recognize specific
entities, or search for particular characteristics.
After you enter the office and identify the chair, the next
task for your brain is to take an appropriate action concerning
this object. This is where the criteria concerning memory
capabilities comes in. How do you know what this object is
for, what its characteristics are, how it is used, its potential
dangers, and so on? There are two obvious ways that you can
obtain this information. First, your brain could search the
sensory data it is receiving to answer these questions. Second,
you could rely on your past experiences with this type of object.
That is, you could retrieve your accumulated knowledge
concerning “chairs” and assume that this particular chair has the
same characteristics. Our assumption that “sensory analysis is
enormously difficult compared to the brain™s memory
capabilities” means that the second option is faster that the first.
Chapter 8: The Function of the Subreality Machine 135

That is, it is faster for the brain to retrieve known information
about objects in general, than it is for the brain to deduce this
information each time it encounters the object.
Since the brain is a product of natural selection, it should be
highly adapted to its function and environment. If sensory
analysis is extremely difficult compared to the brain™s
computational power and its memory capabilities, this should
shape the way that our mental processes are carried out. Given
these assumptions, we now ask, how would we expect the brain
to operate?
Again we will use the example of walking into a strange
office. In this new situation the brain must quickly identify
those things in the environment that are critical to its survival.
It must do the most that it can in the first fraction of a second,
the timescale that critical events happen in our world. And the
best it can do is to categorize the key elements of the scene, the
main features that will dictate the appropriate movements that
must be made. From the sensory data, it recognizes the area as
a typical office, containing a desk, chair, table, and a man.
However, it determines little or nothing about the particular
characteristics of these things; it only knows that they are
typical members of their categories. This is all the brain can
know in the first fraction of a second; its computational powers
are not sufficient to extract anything else from the sensory data.
But the brain needs to have detailed information about these
objects in order to move our bodies among them in a productive
way. The quickest way for it to attain this information is from
its own memory, what it has previously learned about objects in
these particular categories. While these stored generalizations
may not be accurate, they are the best that the brain can do,
given the time constraint it is working under.
Keep in mind that the function of the brain can be divided
into three parts, (1) analyze the sensory data to understand the
environment, (2) decide where to move, and (3) coordinate the
movement. Accordingly, step one must produce a “description”
of the local environment that can be used by steps two and
The Inner Light Theory of Consciousness
136

three. Given the assumptions that we have made, we would
expect that this “description” would be composed of two parts,
coarse information about a few key elements in the nearby
environment, with the remaining details filled in from stored
memories.
In short, the brain creates an inner reality that is (1) based
loosely on the surroundings, (2) consistent with previous
memories, and (3) free from noise, interference and ambiguity.
This important concept is the sixth major teaching of the Inner
Light theory:



Major Teaching #6:
The Function of the Subreality Machine
The subreality machine in the brain provides efficient
sensory analysis. It achieves this by inspecting the poor
quality data from the senses, and constructing an inner
reality that is an estimate of the actual environment.
This inner reality provides the consistent and noise-free
information needed to plan and execute movements.




The Capacity of our Brains
In order for this scenario to work, the brain must have
stored information about a vast number of categories of objects.
This leads us to ask, is it really possible that the brain could
categorize all of the familiar things that it knows? After all, we
are familiar with everything from the whiskers on a cat, to the
sound of a locomotive, to the taste of peanut butter. Aren™t
there just to many things that we are familiar with to make this
possible?
To answer this question, we can make a rough estimate of
just how many “things” a human knows. Of course, we can do
no better than a general approximation, since we haven™t
Chapter 8: The Function of the Subreality Machine 137

defined exactly what a “thing” is. For instance, a “thing” might
be the cat™s whiskers, or the whole cat, or all mammals in
general. Nevertheless, it is still useful to go through the
calculations to get a general idea of the size of the library stored
in each of our heads.
The key to making this estimate is a very simple principle:
we cannot know something unless we have learned it sometime
in our past. This is important, because we know very accurately
how long each of us has been learning things. For instance, a
typical adult has been alive for 30 years, which is the same as
10,950 days. This means they have been awake for about
175,000 hours, 10 million minutes, or 600 million seconds. The
question is, on the average, how often do we learn a new thing?
Is it every second? Every minute? Every hour?
To answer this, think about a motion picture that you saw
five to ten years ago. Now suppose that you are shown a one
second segment from this movie, along with a one second
segment that was shot for the movie but not included in the final
release. Could you reliably pick the one you had seen before?
Of course not, indicating that we do not learn new things on a
second-to-second time scale. But if the segments are made
longer, say ten minutes, your recognition would become much
more accurate. Making the segments an hour long would make
your recognition nearly perfect. Using this line of reasoning,
we can estimate that we learn one new “thing” about every ten
minutes or so. This corresponds to about six new things per
hour, 100 new things per day, 36,500 new things per year, and
about one million new things in an entire lifetime. Keep in mind
that this only pertains to long-term memory, those things that
can affect our mental capabilities years after they are learned.
At this instant you can probably recall hundreds of things from
the last one-hour of your life. However, nearly all these will
fade away, and not become a permanent part of who you are.
In short, our brains have a mental capacity of about one
million “things.” For comparison, this is about the same
number of sentences in an encyclopedia, giving us additional
The Inner Light Theory of Consciousness
138

reason to believe this estimate is reasonable. Of course, this
number may be off by a factor of ten or more either way,
especially since we have not really defined what a “thing” is.
The point is, our mental world consists of a finite number of
concepts that can be manipulated. Further, this finite number is
not a trillion, or even a billion, but only in the neighborhood of
about one million.
This is important because it allows us to compare our
mental capacity with the physical structure of the brain. We
know that the brain is composed of about 100 billion neurons,
making about 100 trillion synaptic connections. In other words,
the brain contains about 100,000 neurons and 100 million
synaptic connections for each concept that the mind can ever
process, seemingly more than sufficient to carry out the task.
Going back to our original question, is it possible that the
brain has the capacity to categorize all of the things that humans
know? While much of the brain™s operation remains a mystery,
the answer to this question seems to be a clear yes.
On a more philosophical note, this estimate of our mental
capacity is a bit unsettling, especially for scientists that are
accustom to dealing with very large numbers. For instance,
there are about a trillion stars in our Milky Way Galaxy, and a
billion trillion atoms in a single drop of water. Compared to
these enormous numbers, a brain capacity of one million
concepts seems quite small and almost insignificant.

Why Do We Dream?
The Inner Light Theory provides a very specific answer to
the question, What are dreams? Each of our minds contains a
subreality machine to facilitate the analysis of sensory data.
Dreams result when this machinery is operated without input
from the senses, resulting in an inner reality that does not
correspond to the external world. Dreams are the subreality

<<

. 3
( 4 .)



>>