A machine is always designed to fulfil a function. Machines
are designed to open a tin of tuna, to crimp hair, to guide
space shuttles to Mars or to assemble other machines. The
questions we ask of a machine are of the nature ‘what
does it do?’ or ‘what’s it for?’ In
other words, we are ‘behaviourist’ or ‘functionalist’
about machines. We concern ourselves only with their ostensible
consequences. It is rare therefore for us to ask if a machine
knows what it is doing. Whilst the more inquisitive mind may
reasonably ask ‘how does it work?’ it is a different
thing altogether to ponder whether a machine knows that it’s
opening a tin or to ask if the machine thinks about the hair
it is crimping. Indeed it may seem an absurd thing to even
contemplate, that is until the machine becomes more complex.
For it seems that as we add functions to our machine and ask
it to do more we come closer to admitting the possibility
that our machine is capable of intelligent thought. The question
at hand asks us whether a machine could ever reach a level
of sophistication whereby it became aware of the functions
it was performing and understood the behaviour it was exhibiting.
Whatever the answer to this, Alan Turing professes to hold
the criteria by which any claim at true Artificial Intelligence
is to be judged. ‘The Turing Test’, whereby a
human agent is unable to distinguish between a computer and
human, hidden from sight, by the answers each give to a series
of spontaneous questions, is often upheld as the relevant
yardstick. It aims to establish intelligence by the behaviour
a machine exhibits. This may seem reasonable as often we will
judge another person to be intelligent based upon the behaviour
they exhibit. However, The Turing Test is far from controversial.
According to some it is one thing to exhibit intelligent
behaviour but something quite different to understand that
behaviour. Jon Searle’s ‘Chinese Room’ thought
experiment is intended to demonstrate this. The Chinese Room
translates English into Chinese, one inserts an English sentence
into a hatch at one side of the room and moments later the
corresponding Chinese sentence is produced from a hatch at
the other end. It seems as though the Room is exhibiting intelligent
behaviour in translating English to Chinese and one may be
tempted to assert that the room understands (or is conscious
of) Chinese. However Searle then takes us inside the room
and we find a group of people sitting at desks passing symbols
between each other. In front of each person is a set of instructions
of the ilk ‘if you receive symbol ‘x’ then
pass it to the right, if you receive symbol ‘y’
pass it to the left, if you receive symbol ‘z’
then keep it on your desk’ The collective actions of
this group of people are designed to translate an English
sentence into Chinese but nowhere within the room is Chinese
understood. The people only understand the detached instructions
before them and there is no collective understanding of what
is being done. This demonstrates that whilst a machine may
demonstrate intelligent behaviour this does not mean that
it understands what it is doing. For the sake of argument,
it will be useful for us to imagine that it were possible
to build a machine which exhibited the same behaviour as ourselves.
A machine which responded to the same stimuli and had human-like
ostensible responses.
Searle has warned us that it may be erroneous and misleading
to conflate intelligent behaviour with conscious understanding
but let us now suppose that the machine was identical to us
in the way it functioned as well as the way it behaved. Therefore
as well as having identical inputs and outputs it now shared
identical processes. The only difference being that it was
made of different material (not flesh and blood). For example
then, as a replacement the firing ‘c-fibres’ (thought
to be the process by which a human feels pain) we have some
electric impulses being sent down a fibre optic cable causing
the machine to yelp realistically as if in agony. Is our machine
now conscious of pain? It seems that our intuition urges us
to resist this conclusion. Why is this? Ned Block thinks that
it may be because we don’t believe that metal and glass
could ever have sensations surely this would be miraculous.
Did it not take the magic of a Blue Fairy or some such to
bestow feelings and sensations upon the wooden Pinocchio?
Block however thinks this is a peculiar reservation. Why do
we find the idea of metal being conscious any less plausible
than the idea of grey spongy stuff (i.e. brain matter) being
conscious? Considering that our mental life does seem miraculous,
how can we hold any firm intuitions about the possibility
of mental life being enabled by any other material?
There are still many troublesome issues though. Not least
that to suppose that all human sensations/thoughts/beliefs
and emotions can be reduced to mechanical processes is surely
to fail to take account of all the necessary phenomena. Humans
are more complex than the input-process-output model. Some
human’s have know senses (they are blind and deaf etc)
and may be categorised as having no inputs. Yet they can still
think and display meaningful behaviour. Other’s (perhaps
total paralytics) display no outputs yet one suspects they
have thoughts and sensations. Furthermore a functionalist
account of consciousness cannot give an account of every individual
emotion. A human is aware that there are many different qualitative
types of anger and that these will feel different each time
and (one suspects) from human to human. How could a machine
with only a mechanism for each function expect to represent
TO ITSELF the spectrum and qualitative differences of these
various types?
It seems that were we to deem a machine ‘conscious’
we would have to do so one evidence more than just displaying
intelligent behaviour. And whilst we shouldn’t resist
on the grounds of our intuitions, it seems that one would
be hard-pressed to give an adequately functional account of
the human mind let alone how a machine could duplicate it.
A machine may well emerge that is capable of displaying intelligent
behaviour, perhaps indistinguishable from that of humans but
this would not necessarily make it a conscious entity. It
seems most prudent to remain carefully agnostic as to whether
a machine could ever be conscious.
Therefore when we consider this new machine’s candidacy
for consciousness what criteria should we consider?
|
