By Paul Almond, 9 August 2003
Introduction
John Searle, an American philosopher, is well known for
arguing against a position in the areas of cognitive science, artificial
intelligence (AI) and philosophy that he calls strong AI, and which he describes as follows:
|
‘...according to
strong AI, the computer is not merely a tool in the study of the mind; rather,
the appropriately programmed computer really is a mind, in the sense that
computers given the right programs can be literally said to understand and have
other cognitive states. In strong AI, because the programmed computer has
cognitive states, the programs are not mere tools that enable us to test
psychological explanations; rather, the programs are themselves the
explanations.’ [1]
|
Strong AI is the
idea that machines that could behave as if they were conscious would
necessarily be conscious.
This essay will
attempt to weaken Searle’s case, using an argument based on Charles Darwin’s
theory of evolution by natural selection.
Searle’s View of Consciousness
Searle regards consciousness as real. Unlike some detractors
of the strong AI case he does not propose a metaphysical cause of
consciousness, but instead regards it as emergent behaviour produced by the
‘causality’ in some physical systems. In his book The Mystery of
Consciousness [2] he states (Chapter 1):
|
‘Consciousness is caused by lower-level neuronal processes
in the brain and is itself a feature of the brain. Because it is a feature that
emerges from certain neuronal activities, we can think of it as an ‘emergent
property’ of the brain. An emergent property of a system is one that is
causally explained by the behaviour of the elements of the system; but it is
not a property of any individual elements and it cannot be explained simply as
a summation of the properties of those elements. The liquidity of water is a
good example: the behaviour of the H2O molecules explains liquidity
but the individual molecules are not liquid.’
|
In other words,
Searle regards the specific physical processes going inside a system as being
critical with regard to whether or not it is conscious.
Searle accepts a
position that he calls weak AI as being valid. In The Mystery of
Consciousness (Chapter 1) he defines weak AI as follows:
|
‘...the view that the
computer is a useful tool in doing simulations of the mind, as it is useful in
doing simulations of just about anything we can describe precisely, such as
weather patterns or the flow of money in the economy.’
|
Searle’s View of AI
Searle uses the Chinese
Room Argument, which will not be analysed here, in an attempt to show that
computer software that can produce the external appearance of conscious
behaviour is not necessarily sufficient for consciousness. He does not have the
view, however, that only human brains can be conscious; rather, he appears to
think that there is a set of possible machines that could be made which could
be conscious, the human brain being a member of this set, and that whether or
not a machine is a member of ‘the set of conscious machines’ depends on its
internal workings and not its externally observed behaviour.
So what does Searle
really think of computers and consciousness? Although he states that he does
not attempt to prove that computers cannot be conscious, it is clear that he
does not take the idea too seriously. In The Mystery of Consciousness
(Conclusion) he states:
|
‘I can’t prove this chair
is not conscious. If by some miracle all chairs suddenly became conscious there
is no argument that could disprove it. Similarly, I do not offer a proof that
computers are not conscious.’
|
This is important to
my case against Searle. He does not merely argue against the idea that an
appropriately programmed computer would guarantee consciousness: he also makes
the clear suggestion that there are no grounds at all for presuming
consciousness to exist in an appropriately programmed computer. In The
Mystery of Consciousness (Conclusion) he states:
|
‘To try to create
consciousness by creating a machine which behaves as if it were conscious is
similarly irrelevant, because the behaviour by itself is irrelevant.’
|
This claim can be
shown to have severe problems when considered in relation to Darwin’s theory of
evolution.
The Argument Against Searle
In The Mystery of
Consciousness (Chapter 2), Searle states:
|
‘I, for one, am always
amazed by the specificity of biological systems, and, in the case of the brain,
the specificity takes a form you could not have predicted just from knowing
what it does. If you were designing an organic machine to pump blood you might
come up with something like a heart, but if you were designing a machine to
produce consciousness, who would think of a hundred billion neurons?’
|
That may be worthy
of amazement, but if Searle was right I think something else should be worthy
of amazement: that Darwinian evolution should solve the problem of making a
genuinely conscious machine, as opposed to a machine that merely mimics
consciousness, and that it should be able to solve Searle’s design problem by
finding such a ‘specific’ arrangement of matter when, as far as evolution is
concerned, genuine consciousness would have absolutely no more survival
value than ‘faked’ consciousness.
This is the problem
for Searle’s reasoning: Darwin’s theory of evolution is not familiar with
Searle’s ideas about consciousness. Searle states that attempting to create
a conscious machine by making one that behaves as if it were conscious is
irrelevant, yet his own claims, combined with the view that consciousness
exists in humans and that human biology was produced by Darwinian evolution
suggest that this is exactly what evolution did
I shall explain why
this should be the case. Darwinian evolution works according to two mechanisms:
random variation and natural selection. Random variation involves some random
change occurring in the ‘recipe’ that makes an organism and natural selection
involves the environment determining the chances of reproduction that organisms
have and therefore selecting which, of all the ‘recipes’ that are in the world,
will become more common and will be the basis for future recipes to which later
random variations are made.
The important point
here is that the natural selection process has no regard at all for how elegant
the internal workings of an organism are and it lacks any motivation to
actually produce consciousness. Features of an organism are selected, in
evolution, only for their benefits in increasing the average number of
offspring that an organism will have.
It is easy to see
why features of organisms that cause them to behave as if they were conscious
could be selected and why any mutations that lead in this direction could be
favoured. Human behaviour is extremely sophisticated and our ability to model
our environment, plan our actions and determine the best way to act to favour
our own survival, as well as cooperating with other members of our species, is
certainly of some use to us in propagating our genes. That is the only
criterion that is relevant. Evolution has made a series of ‘design’ changes and
those that have increased the tendency to conscious behaviour have been
selected because they have bestowed survival advantages on the organisms with
them.
Why should evolution
have any tendency to produce real consciousness, though? It would be irrelevant
to say that consciousness gives certain advantages because, according to
Searle, a machine that was not conscious could have exactly the same
behaviour and it would, therefore, have exactly the same advantages.
Just as Searle would regard a human designer attempting to produce a machine
that behaves in a certain way as being unlikely to produce consciousness, then
he should view any process that produces machines only to satisfy
requirements about their externally observable behaviour as similarly unlikely
to produce consciousness. Evolution is such a process.
In other words, if
Searle is right, why are we actually conscious? Why are we not a civilisation
full of ‘zombies’ who do all the things that conscious beings do, but have none
of the mental states associated with being conscious?
I think that this puts Searle’s position in serious trouble and, while
it is not a straightforward refutation of his claim, I suggest that it weakens
his case severely. If Searle is right about everything else then he is left
with the problem that his own view of consciousness actually negates the power
of Darwin’s theory of evolution to explain its existence. This is not a trivial
matter. Darwin’s theory is a major success story of science. Indeed, Searle
appears to admire the theory, stating in The Mystery of Consciousness
(Chapter 5):
|
‘Darwin’s greatest achievement was to show that the appearance of
purpose, planning, teleology, and intentionality in the origin and development
of human and animal species was entirely an illusion.’
|
There is nothing in evolution to produce the type of consciousness
proposed by Searle. Whereas biological specificity can usually be explained by
evolution, Searle is proposing the existence of very high specificity in
biology with no cause for its origins.
Arguments that could be used to defend Searle’s case
I am aware that a
number of arguments could be made against all this. In anticipation of these,
here are some of the more obvious arguments that I think could be made against
my reasoning and my replies to such arguments.
Argument:
You have
misrepresented Searle. Searle never said that only human brains can be
conscious. He accepts that other conscious machines can exist.
My reply:
I have not misrepresented
Searle. Although he accepts that other conscious machines can exist, he clearly
rejects the idea that attempting to create consciousness by duplicating its
externally observable behaviour would be likely to succeed. To repeat a quote
from him that I previously used in this essay, ‘To try to create consciousness
by creating a machine which behaves as if it were conscious is similarly
irrelevant, because the behaviour by itself is irrelevant.’ If you say that
using a particular method to try to create something is ‘irrelevant’ you are
not only implying that it would not be a logically necessary requirement that
using that method would result in success, but that success would be unlikely
to follow from use of that method in itself. For example, if trying to create
consciousness by duplicating the right behaviour had a ninety percent chance of
success, or even a ten percent chance, it would not be ‘irrelevant’: it would
be a process with a reasonable statistical expectation of creating
consciousness.
Argument:
It is absurd for you
to say that ‘fake’ consciousness would be more likely to evolve than real
consciousness. Why should evolution contrive to make something that has all the
contrivance of faking consciousness rather than something that just has it? Why would evolution create animals that are
made in such an apparently contrived way?
My reply:
Anyone asking this
really has not understood what I am saying. Firstly, I am not saying
that fake consciousness would be more likely to evolve than real consciousness;
however, I am saying that this is a conclusion that is implied by
Searle’s reasoning, if we hold it to be correct (which I do not). If conscious
machines have less specificity than machines that fake consciousness and you
are presented with a machine that has been made to produce conscious behaviour
you should conclude, unless you have a good reason for thinking otherwise, that
this machine is more likely to have real consciousness than the ability to fake
it. You cannot have it both ways.
Argument:
In the reply that you just gave you said that if we view conscious
machines as having less specificity than machines that fake consciousness then
a machine made to generate the behaviour of consciousness should be viewed as
more likely to be conscious than to be faking consciousness unless there is a
‘good reason for thinking otherwise’. Searle gives good reasons for thinking
otherwise. For example, his Chinese Room argument shows that there are good
reasons for thinking that computation, in itself, is not adequate for consciousness.
He also argues that the case for computation being adequate is actually
incoherent. Given this, could we not say that conscious machines do have
less specificity than machines that fake it, but that computers are a special
case and that computers are shown by Searle’s logic to be a bad choice of
system for trying to create consciousness?
My reply:
This would actually be a misrepresentation of Searle’s logic. Searle
does not try to argue that computers have something in their construction that makes
them unlikely to be conscious. His arguments merely attempt to show that
computation that makes a machine behave in the same way as a conscious machine
does not necessarily mean it is conscious. I do not think Searle’s arguments
are valid, but even if we accepted them, and if we regarded conscious machines
as having less specificity than machines that fake it, we would still be left
with the strong statistically based suggestion that any machine that behaves in
a conscious way is actually conscious. In addition, the case you have just made
does not merely fail to follow from Searle’s reasoning: Searle’s position
actually contradicts it. In The Mystery of Consciousness Searle attempts
to show that the strong AI position is incoherent because, he suggests, the
interpretation of a machine as a being a computer, is subjective. He states
(Chapter 1):
|
‘…computation is not a machine process like neuron firing or internal
combustion; rather, computation is an abstract mathematical process that exists
only relative to conscious observers and interpreters.’
|
Because Searle has the opinion that it is only observers that assign
computational interpretations to machines, then he could not validly use an
argument that suggested that a computer with conscious behaviour was
particularly more unlikely than any another arbitrarily selected machine with
conscious behaviour to actually be conscious: a device would only be a computer
because a human has decided to call it one and, leaving this subjective
interpretation aside, we would simply be left with a set of machines, which
work in various ways, that happen to have the same behaviour as conscious
machines.
Argument:
How do you know that
there is not something intrinsic to biology or the process of Darwinian
evolution that results in consciousness? If this were the case then
consciousness would tend to result often when those sorts of processes were
making brains, but it would still mean that there was no reason to presume that
machines made by other means were conscious.
My reply:
In The Mystery of Consciousness Searle states:
|
‘…some brain processes are sufficient
to cause consciousness. This is just a fact about how nature works.’
|
He frequently says
words to this effect. This idea of associating consciousness with ‘something’
in biology, or the processes of evolution, which causes consciousness to emerge
for some unexplained reason, has a fallacy that may not be quite obvious at
first glance. It makes what is being proposed a system of high specificity with
no explanation for its origins. We should not deceive ourselves by thinking
that a Darwinian explanation of the origins of apparently conscious behaviour
provides that explanation in the case of inner mental processes, the existence
of which is not implied by such behaviour.
To make this point more clear I will provide an analogy:
Let us imagine that
we lived on a dark world where no light shone and where we had never
experienced it, but instead sensed our environment by other means. Let us
suppose that we found animals with extremely complex eyes. Someone may theorise
that light used to shine on this world and that the process of evolution
produced eyes as a result, but how would someone sound who said, ‘There was
never any light. These eyes did not appear because there was an evolutionary
advantage. The emergent properties that eyes have are just something that
biological systems just happen to generate?’ It would be clear that such a
statement was suspect. Such a person would be arguing for a huge amount of specificity
in a system and suggesting that it happens because biology and evolution just
happen to do things this way, irrespective of any survival value of this
system.
This is the problem
with Searle’s reasoning: ultimately, it puts consciousness outside the realm of
evolution and merely makes it an unexplained ‘emergent property’ of the process
of evolution. If we were to find this plausible we could explain any
aspect of human biology in these terms. For example, we could suggest that the
heart exists because evolution tends to cause the emergence of hearts,
regardless of whether or not they have survival value. We could do this for
every system in the human body, until we were left with no systems at all. We
would then be faced with the bizarre idea of an evolution process in which nothing
actually evolved, but in which every system originated because it was simply a
characteristic of the process to generate it, regardless of its survival value.
Because the evolution mechanism was no longer actually causing anything to
evolve it would not matter what the evolution mechanism was and we could do
away with the entire theory. This would leave us with:
- Evolution is a process which does not
specifying, because nothing actually evolves in it.
- Biological systems have high specificity
and they are produced by evolution because they are emergent properties of
evolution. They do not evolve. They are just produced as by-products of
evolution, because of the nature of what evolution entails.
If we found
ourselves forced to use reasoning like this we would consider it highly
unsatisfactory. Declaring a system of high specificity to be a natural result
of a process when we have no reason for presuming that process to be capable of
such specificity does not deal with the problem of probability: it merely hides
low probabilities from a casual observer. Searle is asking us to adopt a
position on consciousness that we would find unsatisfactory, given our
knowledge of evolution, to adopt on any other biological system of high
specificity.
Argument:
Suppose that an
appreciable number of machines that act conscious are conscious? Humans
are clearly conscious, but not much luck would be needed for that. Similarly,
you do not know that if you showed Searle a machine made a by a process that
produces machines to behave with apparent consciousness he would denounce it as
‘fake consciousness.’ He may agree that it has a reasonable chance of being
conscious.
My reply:
Once more, I think
that Searle’s statement, ‘To try to create consciousness by creating a
machine which behaves as if it were conscious is similarly irrelevant, because
the behaviour by itself is irrelevant,’ is implying that Searle would not take
this position.
We could consider this thought experiment though:
Let us imagine that
we made one hundred different machines, each of which has the appearance of
conscious behaviour and each of which has been made using a different method of
producing machines that have conscious behaviour, the one hundred possible
methods being randomly chosen from all the possible methods of doing this.
We put all these
machines in front of Searle and ask him if he thinks that there is at least a
fifty percent chance that he is looking at least one conscious machine. If he
gives a negative answer we increase the number of machines by one and keep
going until he does say that there is a fifty percent chance that there is a
conscious machine in front of him.
If Searle agrees
that there is a fifty percent chance that one of these machines is conscious
when the number of such machines is relatively low then he is weakening his own
assertion that externally observable behaviour is irrelevant. On the other
hand, if he waits until there is a vast number of machines in front of him
before giving such an answer he is making consciousness more of a problem, in
that he is increasing its specificity and making it less plausible that it
would exist without a mechanism such as Darwinian evolution to bring it about.
It should also be
noted that this may not be a thought experiment forever. If adequate advances
in AI are made in Searle’s lifetime he could be put in exactly this situation.
Argument:
You are unfairly
accusing Searle of arguing that the externally observable behaviour of a system
is irrelevant. He accepts that we may reasonably judge other people to be
conscious, based on our observations of their behaviour.
My reply:
Searle does not
actually say this. While he does say that we can judge other people to be
conscious, because of their behaviour, he qualifies it by saying that we also
take into account the fact that their brains have a structure similar to our
own. In The Mystery of Consciousness (Conclusion) he states:
|
‘…we can typically
tell when other people are conscious by their behavior, for example-but the
epistemic relevance depends on certain background assumptions. It rests on the
assumptions that other people are causally similar to me, and that similar
causes are likely to produce similar effects.’
|
The qualification
requiring similar causality for the behaviour is yet another assertion by
Searle that he rejects behaviour, in itself, as a strong indication of
consciousness.
Argument:
How do we know we have not just been lucky?
My reply:
We cannot prove
that we have not been lucky and if we accepted Searle’s arguments about
consciousness we would have to accept that we might be conscious merely by
virtue of a high degree of luck, whether such luck is a sequence of mutations
occurring to produce real consciousness with no reason or whether such luck is
that the process of Darwinian evolution just happens to produce
consciousness when we might imagine many other things that it could produce,
but does not. Any reliance on such good fortune for the existence of our
consciousness does however weaken Searle’s argument: it would certainly be
better for Searle’s case if there was no reliance on such extreme luck for our
existence.
Argument:
Maybe we have not
been lucky. Maybe life has started on a massive number of planets. On most of these
planets civilisations full of zombies may exist, but on a very small number of
them real consciousness may emerge. It would seem like extreme luck to those
organisms that were conscious, but this would not be the case. Someone is going
to be conscious if enough species are created with conscious behaviour.
Furthermore, some quantum physicists think that a theory called the many
worlds interpretation (MWI) correctly describes the universe. In MWI many
worlds exist apart from our own and life and evolution could have occurred in a
vast number of such worlds. With such a large number of civilisations full of
zombies existing, someone actually has to be conscious and whoever that
is will feel extremely lucky.
My reply:
I would be surprised
if Searle himself used an anthropic argument like this. It would still leave
his case weakened, because alien life and, in the case of MWI, entire
cosmologies, are now being introduced to support it. Specific models of reality
are now needed to make Searle’s case tenable and, if they are introduced, it is
now dependent on a lot more than just his reasoning.
Conclusion
Searle’s view of
consciousness implies that either:
1. physical systems
that are conscious have a very high degree of specificity and machines made
solely to produce conscious behaviour are unlikely to have such specificity. In
this case, there is no reason for Darwinian evolution to have produced such
specificity in humans and Searle is in the position of arguing for the
existence of high specificity in biology that is not accounted for by
evolution. This weakens his entire case, but creationists might like it.
or
2. we should say
that a machine that behaves like a
conscious machine, while not necessarily conscious by virtue of this (according
to Searle’s reasoning), is likely to be conscious, even if we do not know
anything about the causality in that system that produces that behaviour,
because the specificity of machines that behave as if they are conscious and
are conscious is not substantially higher than that of machines that merely
behave as if they are conscious. In this case Searle’s assertion that the
externally observable behaviour of a machine is irrelevant, with regard to
whether or not it is conscious, is weakened.
References
[1] Searle, J. R. (1980). Minds, brains and computers. The
Behavioral and Brain Sciences 3:417-457
[2] Searle, J. R. (1997). The
Mystery of Consciousness. New York: The New York Review of Books.
