In "Free Will: The Illusion and the Reality, and How Our
Minds Rule the Day", I discussed a view that the multi-level capacity of
the human mind, i.e. its ability to loop back on its own processes, enabled our
exercise of free will.
Consistent with that, I argued we needed to "exercise"
our free will, since absent a disciplined approach, absent our self-watching of
our own decision making, that decision making could become controlled by
subconscious and predetermined factors. Free will could go away.
A question now is can computers exercise free will? That is an
interesting question in itself (at least to some; others within their free will
can decide that the question is not interesting). The question also shines a
spotlight on the still considerable skills of the human mind.
Computers: What They Can Do
Computers today have an array of amazing capabilities, but also
severe limitations.
Computers are fast, of that there is no doubt. And they are
getting faster. Computers are flexible, to the extreme; they can be programmed
to perform multiple tasks, just about any task. Computers can self-correct;
they can review their output and adjust factors and even coding to improve
their accuracy and performance.
For all that, computers have limitations. Computers are not yet
very good at sensory input. The human brain, thanks to several hundred million
years of evolution of life (or if you prefer due to the design of a god or
higher power) can integrate sight, sound, smell, touch and taste, and do so
essentially instantly. We can then store such integrated experiences, millions
of them, and match our current experiences, even if distorted, shifted or
disoriented, to the stored past experiences.
In contrast, computers can not (yet) do sensory integration. We do
have computers that can process visual input to navigate obstacles. We marvel
at that. But consider human's ability to experience and recount the rich
sensual tableau of a mother's kitchen during Thanksgiving preparation. That no
computer can do.
Computers, in the same vein, are not good at forward
visualization. Certainly computers can project forward the weather, but they
can't project forward a sensory image of what twelve inches of snow looks like,
and how to handle the children when school is cancelled.
Computers are not yet very good at meaning. Humans are. Humans can
take logical structures, symbolic shapes, remembered experiences, forward
visualizations, categorized information, and create meaning. Computers can link
information on these items. But that is akin to drawing lines on a paper.
Computers can not in any sophisticated way build integrated three-dimensional,
symbolic/visual/temporal constructs to create what we call meaning.
Computers have only limited ability to be self-reflective.
Computers certainly can execute feedback. They can have algorithms that compare
their calculated output or action against the goal, and correct the algorithms.
But humans have algorithms that are inherently self-referential. We are
conscious, and we are conscious of our consciousness. We are observers, and we
are observers of how we observer. We are thinkers, and we can think about how
we think.
Computers, so far, do not have algorithms that are so inherently
self-reflective. If a computer has an algorithm for observing the terrain, that
algorithm can not turn inward and observe itself observing. If a computer has
an algorithm for correlating text passages across millions of input documents,
that algorithm can't correlate the bit streams internal to the itself that are
generated by the process of correlating text passages.
Can Computers Exercise Free Will?
Let's start with what we mean by free will, or at least a common
sense, but decidedly non-rigorous, definition of free will. Let's say free will
would be the ability to select among alternatives to best advance goals, and do
so in creative ways that may or may not extend from prior conditions or
experiences.
I would then say computers can make free choices. Computers can
look at situations with multiple options, and select one in way that extends
beyond the deterministic limits of their programming.
For all that, I would say, however, that computers can not make
free choices anywhere comparable to the range that humans can.
Let's examine where computers make free choices. Let's start with
a computer controlled vehicle, by itself, in a complex but static terrain,
faced with a decision on which of three roads to take. Such a computer/vehicle
combination could survey the roads, identify hazards, calculate physical
parameters, assess probabilities, then run Monte Carlo simulations to pick the
best choice.
This would reasonably resemble free will. Why? Because the link
between the initial conditions plus computer code, and the final outcome, that
link, though in some sense determined, is so intricate, that the concept of
cause and effect starts to be devoid of meaning. If two of the three roads were
acceptable, and of almost equal weighting, the extensive series of calculations
the computer executed, and the potential for the outcome of those calculations
to be sensitive to minor variations, means essentially no ability exists to
predict the outcome from the input.
So I will count that as the ability to choose between options
without the choice being decidedly determined by initial conditions.
But Human Free Will
That is nice, computers can guide a motorized vehicle through
rough, unknown terrain in experimental situations.
But put that vehicle into a war. Several things happen.
Computers can not ascribe meaning to the attributes of war. Death,
destruction, mercy, justice, sacrifice, justification, horror, subjugation,
honor, bravery, and on and on, the computer can not in any way integrate those
critical and important consequences and characteristics of war into any
realistic sense of meaning and ethics. Certainly we can develop algorithms to
convert those items to numbers, but even with that, the computer can not
inherently attribute to those numbers the meaning the underlying attributes of
war have for humans.