Massively
Parallel |
|
You frequently see brain
processes described as 'massively parallel'. The analogy being
suggested is with parallel processing in computers, where a
process is divided up and different parts of it are carried out
simultaneously by different processors, thereby finishing the job
more quickly.
Now the first and best reason why the brain isn't a
parallel processor, of course, is because it isn't a computer at all. But even if it were, there
would be no good reason to think the brain is a parallel, rather
than a serial processor. Massively parallel processing
specifically is a particular, sophisticated
form of parallel processing which allows hundreds of processors to
work efficiently on the same overall job, such as searching a large
database. Each processor typically has its own memory and operating
system; where the activities of the different processors overlap,
the relationship is carefully managed so that they do not have to
remain in step, and this requires careful prior programming. This
does not seem to me even remotely like the way the brain is
organised, so far as we understand it. There do not seem to be any
good candidates within the brain for the role of processor (neurons
are surely too simple)and the discrete, managed approach of
parallel processing contrasts starkly with the prolific
interconnectedness of the brain.
Parallel processing isn't
such a big deal, in any case. Turing himself established that all
computers are essentially the same: fundamentally they differ
only in speed and storage capacity. Parallel processing
does not affect what a computer can or can't do: it's just
a tactic which may allow some additional speed. It doesn't
always have even that advantage - some tasks cannot be done in
parallel anyway.
The visual system is often quoted as the best
example of parallel processing, but what this seems to refer to
is the existence of separate channels in which different kinds of feature
recognition take place simultaneously: one dealing with colour,
another with location in space, for example. This kind
of activity certainly justifies the description 'parallel processing' in one
sense, but not at all in the sense which implies a
resemblance to parallel-processing computers. Moreover, while this kind of parallelism occurs in the
visual and certain other systems, it is not typical of brain processes overall,
and there are, so far as I know, no parallel channels
which could plausibly be associated with conscious
processes.
Another
reason for talking about parallel processing seems to be the
way a set of widely-separated sections of the cortex
may be activated by a particular conscious experience. You could
fairly say that these different sections were responding in parallel, but it looks
more like the whole of the cortex being available for use in
any single task. Certainly there is no sign here of the segregation of processes characteristic
of parallel processing.
It seems likely that part of the
appeal of the 'massively parallel' idea stems from a confusion with
the 'pandemonium' or 'multiple drafts' kind of theory, in which
different potentially conscious threads of thought develop
separately and compete for dominance. But the existence of these
multiple threads at the conscious, or potentially conscious level,
would imply nothing about parallelism at the operating level. Even
Roger Penrose
seems prey to this
misconception when in 'The Emperor's New Mind' he seems to assume
that parallel processing would necessarily involve thinking
consciously about two different things at once. But suppose we had a
program, running which gave rise to consciousness. Running it on a
serial processor and running it on a parallel processor would
achieve exactly the same result (possibly a little quicker on the
parallel processor). Running it on a parallel processor would no
more give rise to multiple consciousness than solving a problem with
a parallel processor gives rise to multiple answers.
Daniel Dennett
presents an intriguing
argument in 'the Intentional Stance' about why the brain's
supposed parallelism might give it a vital turn of speed that computers
can never match (which might mean computers can never handle typical
human tasks at real-time speeds - an uncharacteristically downbeat
thought). There's something weird going on
here: it's being suggested that the brain works in the same
way that some powerful computers do, and that this is a way of
working that computers may not be able to manage, and that
therefore they might not be able to keep up with
the brain - I think I must have misunderstood
somewhere. Even less fathomably, in
'Consciousness Explained', Dennett suggests consciousness might
involve a parallel processor implementing a virtual serial
processor. What would
be the point of that?
|
|
A computer |
|
People are always asserting that the brain is
a computer. The comparison goes back to the days of Babbage, and in a
sense, mechanical brains are exactly what computers were always
supposed to be. It's only natural that people look for analogies
(when I was young - well, younger - they used to compare the brain
to a telephone switchboard) and computers are the obvious choice. In
fact, however, there are no good analogies in the case of the brain.
It is sui generis - utterly unlike anything else
in the world (except another brain), and it differs fundamentally
from a digital computer.
For one thing,
the brain has no software. Universality is one
of the essential characteristics of a computer - allowing for limitations of
speed and storage, any computer can implement any function.
By contrast, there is no way to load arbitrarily chosen new
programs into the operating level of the brain: what it does
is largely hard-wired. There is a degree of plasticity which allows it
to recover from injury, and an ability to modify
behaviour through learning: but neither resemble the loading of a
new program. Only at the conscious level, at a painfully
slow rate, can the brain 'run' programs. In addition, the brain
is not a discrete-state machine. Computers, ideally, switch from
one definite state to another with no transition. No real object
behaves exactly like that in practice, but brains don't even
come close, relying on a maze of interconnected physical
and chemical processes none of which is instant.
Now it can be argued, I realise, that the function of
neurons is exactly to reduce this chaos to binary signals
- either a neuron fires or it doesn't, after all - but
the rate of firing and the complex series of events
which determine it are not digital. I would add as
a third point that computers are artefacts, while
brains generate themselves - an important
distinction.
Gerald Edelman stresses that
the brain is not a computer: no two brains are wired the same way,
he points out, and reality is not a program. Although there are
counter-arguments, I think these are essentially sound points which
draw attention to the falseness of the computer paradigm.
Curiously, John
Searle , the leading sceptic about
the possibility of generating consciousness by running a program,
accepts that the brain is a computer. Latterly, however, he has
argued that virtually everything in
the world can be counted as a computer, including, for example a
window. Now it is perfectly true that you can in theory interpret
anything as being anything else if you're perverse enough. But for
rational beings the choice of interpretation is not arbitrary. One
minimum condition is surely that the thing
you're interpreting as an X must do most of the work of
being an X - not the rules you apply in your interpretation.
In order to use, say, a window as a computer, you need
interpretive rules which effectively do all the computing
themselves. |
|
Illusory self |
|
More than one author has come to the conclusion
that the self is an illusion, but so far none of them has
refused royalties on that account. Scepticism about the
reality of the self comes naturally to some materialists
because of the historical association of selfhood with
belief in the soul and hence with
dualism (denial of the
self is also, on the other hand, an important doctrine of
Buddhism, which aims to dispel the illusion), but it's one of those
'Sunday best' theories that no-one actually adopts, or could adopt,
in real life: indeed, when applied to one's own self, it comes close
to self-contradiction. I believe there is no self; it follows that
there is no 'I'; and it follows that nobody believes that there is
no self.
Why
does this view, which surely runs strongly contrary to all our
intuitions, appeal to some people? I think one reason is
the strongly analytical tendency of the modern mind. We tend to
look for final explanations at an atomic level, or at least in the
disassembled parts of any mechanism. But by the time we've got down
to working out how individual neurons work, we're already gone below
the level where the self could reasonably be expected to be
recognisable. It ought not to be too difficult to see through this
confusion (we don't say 'That's just a collection of cogs
and springs - this clock you're talking about is an illusion',
after all), but our intuitions tell us the self is unified and
indivisible: so we feel that if it's anywhere it ought to be
lying on the bench among the other dismantled pieces. Actually,
of course, the fact that something is constituted or composed of
parts doesn't stop it being a real, unified entity at the same
time.
By labelling the
self as an illusion, of course, people do not necessarily mean
to say that it has no existence at all; only that it is something
other, and lesser, than what it appears to be. In particular,
what is meant is that we are not the real source of our own
decisions or behaviour - we just think we are (a kind of
epiphenomenalism). Such
a view might rest on straightforward determinism, reinforced by
findings such as those of
Benjamin Libet , but it is often reinforced by the observation that our
conscious thoughts seem to come into our minds from nowhere:
you can't catch yourself deciding to think something, however
quickly you turn round.
There is really nothing mysterious about this.
The processes which give rise to conscious thoughts cannot
themselves be conscious, or they in turn would need to be
underpinned by other conscious processes, and so on. The act of
making a decision must always precede, if only marginally, conscious
awareness of having made a decision. Discussions of Libet's research
tend to assume that if a decision is conscious there must be
immediate consciousness of the decision, but the two are actually
distinct. Some would disagree, of course; some theories make a
second-order awareness
of conscious thoughts the very thing that makes them conscious:
Libet's results perhaps are a real problem for those who take
this view.
Ultimately, the fact
of selfhood, and the moral agency and responsibility it supports,
are features of the world which are just too salient to be
denied: the task is not to debate their existence, but explain their
nature.
|
|
There
is something it is like |
|
This innocent-looking little
phrase, which I believe comes from Nagel's famous paper 'What is it like
to be a bat?'
(summary of the paper -
'how would I know?') has played a quite astonishing role
in raising the morale of the lovers of qualia. Whenever they're
on the brink of throwing in the towel and admitting that
it's been confused nonsense all along, they repeat the mantra
and everyone brightens up again. The idea is that when you see
something red, it isn't just a matter of acquiring some information
about the light hitting your eye: there is something it is
like to
see the colour red.
To me, this is about as
sensible as trying to include carnal knowledge in epistemology, or
debating the ontology of the 'it' that does the raining. (Come to think
of it, some idle philosopher has probably done that last one). When
we talk about a thing being like something, that's what we
mean - it's like something else. If I eat ostrich, and
someone asks me what it's like, I don't screw up my face and say
'Uh, well I can't tell you, but there was an ineffable experience
which it was like'. I say 'A bit like beef, with a slightly less
granular texture.'
So, if there is something it is like to
see red, what is it? Seeing puce? |
| |