Ray
Kurzweil has advanced a very popular theory that the Singularity is
near – a time where there will be an explosive growth in machine
intelligence, one that sees computers and robots becoming far smarter
than humans. To back up this theory, Kurzweil and his supporters
typically show graphs showing exponential growth in the power and
speed of computer hardware and computer memory.
But
in order for you to have anything like machine superintelligence, we
would need to see more than just computer hardware increasing by
orders of magnitude (factors such as 100, 1000 or 10,000). We would
also need to see computer software making similar strides. If today
someone were to create a computer with a million times the speed and
memory of the human mind, that computer would not be even as
intelligent as a mouse. For such a computer to become even as
intelligent as a mouse, you need strides in computer software on the
same scale as the strides in computer hardware.
One
problem is that computer software is not advancing at anything like
the rate of progress of computer hardware. Computer software is
progressing at a relatively slow rate, an issue I discussed in this blog
post. Computer software is not progressing at anything
like an exponential rate. The basic process of software development
today is not fundamentally different from the process of software
development in the 1990's – programmers slowly and laboriously
grinding out code line by line. Code generators can create lots of
code, but in a typical project most of the code still has to be
created manually.
It
would seem that we would need many centuries to complete the project
of creating the software needed for a computer to be as smart as a
human, assuming that most of it would have to be manually coded. But
many singularity enthusiasts assume that we will have a gigantic
shortcut – the ability to acquire much of this software by studying
the human brain.
We
can find an example of this type of thinking by looking at the
Wikipedia.org page which summarizes Ray Kurzweil's predictions. In
the predictions for the 2020's, we see this prediction:
The
above description is imagining the mother of all shortcuts – the
shortcut to end all shortcuts. The description imagines that we will
be able to get the needed software for an intelligent machine by
studying the operations of the human brain.
The
problem is that there is little reason to assume that we will be able
to do any such thing anytime in the next hundred years. Why? Because
the mystery of how human consciousness and human memory work is one
of the greatest mysteries of the universe, and nature does not give
up its secrets easily.
Philosophers
have long been troubled by what is called the hard problem of
consciousness – the problem of how mind can be produced by matter,
something that is fundamentally different. So far very little light
has been cast on this problem. We do not understand how the brain
stores memories, or how the brain produces thoughts. There is
actually little hope that we will be able to solve this problem any
time in our century.
You
might get the wrong idea from studies that look at which parts of the
brain have higher electrical activity when a particular brain
operation occurs. Such studies merely show correlations, without
throwing any light on how the mental activity is produced. Imagine a
3-year-old child studying a computer. Such a child may make some
observations that lead him to conclude that a particular green light
(at the front of a computer) blinks whenever some type of operation
occurs, such as a file save. But that doesn't really move the child
any closer to understanding the deep mysteries of how the computer is
computing, and how its data is being saved. Similarly, MRI studies on
electrical brain activity really don't take us very far in
understanding the mysteries of how the brain works.
Although
you hear about gray matter when people talk of brain cells, the human
brain is actually a pinkish color. We may use the term the Pink Wall
to refer to mysteries of the human brain which are blocking us from
understanding its inner secrets. We want to pierce this Pink Wall,
and make our way to the inner secrets of the brain. But the Pink Wall
seems all but impenetrable. We have little hope of being able to get
through it in the next several decades.
The Pink Wall
The
idea expressed in the italicized quotation above (that we can unravel
the secrets of the human brain through brain-scanning) seems like
wishful thinking. Imagine an extraterrestrial culture that had not
learned the details of electromagnetism, had not learned about atoms
or subatomic particles, and had not learned about quantum mechanics.
Such a culture might think it could learn the fundamental mysteries
of how matter is organized just by making scans of rocks with higher
and higher resolutions. But such a thinking would be fallacious. It
would seem to be just as fallacious to think that we can unravel the
secrets of the human mind by scanning brain tissue in higher and
higher resolutions.
Can
we, for example, imagine a scientist of the future saying something
like this?
I
couldn't understand before how a neuron produces a thought, but now
that I can view the neuron more closely with my new brain scanner,
now I can see how the thought is produced.
Or
can we imagine a scientist of the future saying something like this?
I
couldn't understand before how a piece of brain tissue produces a
thought, but now that I can view that tissue more closely with my new
brain scanner, now I can see how the thought is produced.
We
can't really imagine either of these things happening, because we
can't imagine how anything that a scientist could see in a scanner
could lead him to say, “Aha, there is a human thought being
produced.”
It
seems, therefore, that the Pink Wall will be blocking us for a long,
long time. We will not have any time soon any brain-scanning shortcut
that allows us to get the software for a silicon brain by borrowing
it from the human brain. It seems that for a long time, the only option for getting
the software for a silicon brain will be to build it through software
development processes similar to those now in use. Such processes
might take centuries before they could produce something like human
consciousness.
So
if you do not see the Singularity anytime in this century, blame it
on the slow process of software development and the Pink Wall which
blocks us from uncovering the sublime secrets of exactly how the
brain works.
No comments:
Post a Comment