Living Computers?
Back in the black-and-white days of 1950, the year before I was born, a British
mathematical genius called Alan Turing devised a game which he called the
"imitation game" (now known universally as the Turing Test) to determine whether
a machine can think. The test was to put a person at a terminal in a room apart
from a machine and another person. The person at the terminal should ask
questions of both the machine and the other person. If, after a certain time,
the questioner could not tell, from the responses, which was the human and which
was the machine, then the test had been passed and the machine could be said to
be intelligent. The test has become the touchstone of AI (Artificial
Intelligence) research and the model for an artificial thinking device has
become known as a Turing Machine.
But would the machine be truly conscious? Or would it be nothing more than a
highly complex fake? Roger Penrose believes the latter. In his books, The
Emperor's New Mind and Shadows of the Mind, Penrose claims that there are
aspects of human thinking that are not computational.
Let us take a slight detour here and remind ourselves how computers "think". For
a start, computers are not capable of original thought - all output from a
computer is a direct result of input supplied by the human programmer, or
"sense" data from the outside world converted into electrical impulses. Computer
software is written in the form of
algorithms:
solutions based upon the results of strictly logical steps, including
questioning: if this happens, do this or, if not, do something else. All of the
complex computing that goes on in the most sophisticated of today's computers
can be broken down into these basic steps. In fact, at the most basic level, the
computer is merely an array of on/off (yes/no; 1/0) switches. The power of the
computer comes from the fact that billions upon billions of these simple steps
can be carried out in every second of time. Having said all that, however, we
can but admire the sophistication of some of the latest computer programs, with
games and simulations being at the cutting edge of development. But are the
programs intelligent? Once again, even the experts disagree.
When Turing conceived his imitation game he predicted that, by the year 2000,
nobody would question the thinking ability of machines: "
I
believe that at the end of the century the use of words and general educated
opinion will have altered so much that one will be able to speak of machines
thinking without expecting to be contradicted" (
Alan
Turing, October 1950). He obviously thought that the processing power of
computers would increase exponentially over the ensuing 50 years. He was right
about the latter but wrong about the former. Today, even the most powerful
computers running the most sophisticated programmes have failed dismally to pass
the test. The
Loebner
contest, with a prize of $100,000 and a gold medal, is run yearly for Turing
Test challengers. The gold medal would be awarded to anyone who's program can
fool 3 out of 10 human judges, for a mere 5 minutes, into believing they were
conversing with another human. Nobody has ever come even close to the gold
medal.
So, despite the exaggerated expectations of the AI community, it seems we are
nowhere near producing a machine that can even mimic the human thought process,
never mind actually think for itself. Roger Penrose, himself another committed
materialist, believes we never will see that thinking machine:
"
It is within mathematics that we find the clearest
evidence that there must actually be something in our conscious thought
processes that eludes computation.'' Roger Penrose in Shadows of the
Mind, 1994.
Penrose's non-computability argument is essentially an application of the the
famous incompleteness theorem of Czech logician, Kurt Gödel. I thought I might
summarise this myself but I'll cop-out this time and quote the
BBC web site, h2g2:
Gödel's Proof
Speaking of mathematics, there is a rather famous result which is typically
brought up in discussions about paradox, although it is not strictly about
paradoxes. Kurt Gödel (1906-1978) proved in 1931 his famous First Incompleteness
Theorem. This theorem states that in any formal system complicated enough to be
able to talk about basic arithmetic, there will exist propositions which are
true, yet cannot be proved nor disproved within that formal system. This is sort
of the converse of a paradox, in which a statement can be proved, and yet is not
true.
Since Gödel's Theorem is not a paradox, it needs no resolution. It can be neatly
illustrated by looking at the statement, made in system X, 'This statement
cannot be proved true in system X'. This statement is clearly true, for if it
weren't, then it could be proved true, which would be absurd. There is, however,
no contradiction in supposing that it is true. It cannot be proved within system
X. We know that it is true, but that knowledge comes from thinking outside of
system X.
In other words, there are things that we, as humans, can see are true but these
truths cannot be proved by computation. Many in the field of AI research, as
well as scientists and philosophers involved in the study of consciousness, do
not agree with Penrose. As I have mentioned, Penrose is a materialist but he
thinks that physical science does not yet explain consciousness satisfactorily
and a new approach - a new physical theory - is required. Penrose is supported
in this view by American anaesthesiologist, Stuart Hameroff and together they
have proposed a model based on quantum collapse at the neural level in the
brain. Philosophically, Hameroff seems less inclined toward the physicalist view
than Penrose, suggesting in
this interview that there might be a "universal
proto-consciousness" to which we all have access.
Another scientist thinking along similar lines but even less impressed by the
purely physicalist view, was the physicist Evan Harris Walker. Towards the end of
his book, The Physics of Consciousness, Walker says: "
A
universe that has only matter cannot have consciousness and cannot have will".
Some physicalists would agree but for the opposite reason: they would deny the
existence of consciousness and free will. They would say that these things are
illusions arising out of the complex computation happening in the brain.
In the end, this is one subject where your opinion, and mine, and Martha's down
at the local newsagent, and Roger Penrose's are all equally valid. What do you
think? Could a machine make decisions in the same way that humans do? We make
judgements based upon evidence and, theoretically, a machine could do likewise.
But, hopefully, we temper our judgements with mercy and mercy is born of
emotion. When we don't apply mercy in our judgements, when we try to design a
set of rules - algorithms, if you like - and stick rigidly to those rules with
no room for compassion, the result is invariably cold and inhuman.
I listened to an example of this on the radio only yesterday: a First World War
soldier was refused a memorial because he had been shot as a deserter and a
coward. That was the result of applying the military rules of the time, without
emotion. But what really happened was that the soldier had been wounded,
recovered and sent back to the front
three times, each time fighting
heroically alongside his comrades in some of the most harrowing campaigns of
that terrible war. In the end, his nerves shattered and his mind in a state of
total confusion, he walked away from the line. Still, he did not hide but gave
himself up to the authorities. He was court marshalled and shot as a coward. No
matter how those who sat in judgement felt about this man, they were not allowed
to question the rules. Taken to extremes, this dehumanising application of
judgement without mercy can, and has, resulted in the greatest atrocities in
history. The German SS were trained to mechanically apply the dictates of the
Nazi high command and became the chilling, emotionless killers we remember with
horror today.
It might, conceivably, be possible to write a program that would apply rigorous
checks and balances sufficiently well to enable decisions to be made avoiding
the kind of gross injustices mentioned above. Would that be, however, compassion
as we humans understand it? Compassion can wrench at your gut and bring tears to
your eyes. Injustice can generate anger and frustration to the point where one
might smash his own fist into a wall and break his own bones? Could a computer
ever feel in that way? Could an algorithm shed a tear?
You decide: your opinion is as valid as anyone else's.
