By Michael
Hanlon
Last updated at 8:45 AM on 11th August 2009
There are only a
handful of scientific revolutions that would really change the world. An
immortality pill would be one. A time machine would be another.
Faster-than-light
travel, allowing the stars to be explored in a human lifetime, would be on the
shortlist, too.
To my mind, however,
the creation of an artificial mind would probably trump all of these - a
development that would throw up an array of bewildering and complex moral and
philosophical quandaries. Amazingly, it might also be within reach.
For while time
machines, eternal life potions and Star Trek-style warp drives are as far away
as ever, a team of scientists in Switzerland is claiming that a fully-functioning replica of
a human brain could be built by 2020.
This isn't just
pie-in-the-sky. The Blue Brain project, led by computer genius Henry Markram -
who is also the director of the Centre for Neuroscience & Technology and
the Brain Mind Institute - has for the past five years been engineering the
mammalian brain, the most complex object known in the Universe, using some of
the most powerful supercomputers in the world.
And last month,
Professor Markram claimed, at a conference in Oxford,
that he plans to build an electronic human brain 'within ten years'.
If he is right,
nothing will be the same again. But can such an extraordinary claim be
credible? When we think of artificial minds, we inevitably think of the sort of
machines that have starred in dozens of sci-fi movies.
Indeed, most
scientists - and science fiction writers - have tended to concentrate on the
nuts and bolts of robotics: how you make artificial muscles; how you make a
machine see and hear; how you give it realistic skin and enough tendons and
ligaments underneath that skin to allow it to smile convincingly.
But what tends to be
glossed over is by far the most complex problem of all: how you make a machine
think.
This problem is one
of the central questions of modern philosophy and goes to the very heart of
what we know, or rather do not know, about the human mind.
Most of us imagine
that the brain is rather like a computer. And in many ways, it is. It processes
data and can store quite prodigious amounts of information.
'They
are copying a brain without understanding it'
//////////////////////////////////////////////////////////////////////////////
But in other ways, a
brain is quite unlike a computer. For while our computers are brilliant
at calculating the weather forecast and modelling the effects of nuclear
explosions - tasks most often assigned to the most powerful machines - they
still cannot 'think'.
We cannot be sure
this is the case. But no one thinks that the laptop on your desk or even the
powerful mainframes used by the Met
Office can, in any meaningful sense, have a mind.
So what is it, in
that three pounds of grey jelly, that gives rise to the feeling of conscious
self-awareness, the thoughts and emotions, the agonies and ecstasies that
comprise being a human being?
This is a question
that has troubled scientists and philosophers for centuries. The traditional
answer was to assume that some sort of 'soul' pervades the brain, a mysterious
'ghost in the machine' which gives rise to the feeling of self and
consciousness.
If this is the case,
then computers, being machines not flesh and blood, will never think. We will
never be able to build a robot that will feel pain or get angry, and the Blue
Brain project will fail.
But very few
scientists still subscribe to this traditional 'dualist' view - 'dualist' because
it assumes 'mind' and 'matter' are two separate
things.
Instead, most
neuroscientists believe that our feelings of self-awareness, pain, love and so
on are simply the result of the countless billions of electrical and chemical
impulses that flit between its equally countless billions of neurons.
So
if you build something that works exactly like a brain, consciousness, at least
in theory, will follow.
In fact, several
teams are working to prove this is the case by attempting to build an
electronic brain. They are not attempting to build flesh and blood brains like
modern-day Dr Frankensteins.
They are using
powerful mainframe computers to 'model' a brain. But, they say, the result will
be just the same.
Two years ago, a team
at IBM's Almaden research lab at
Half a mouse brain
consists of about eight million neurons, each of which can form around 8,000
links with neighbouring cells.
Creating a virtual
version of this pushes a computer to the limit, even machines which, like the
BlueGene, can perform 20trillion calculations a second.
The 'mouse'
simulation was run for about ten seconds at a speed a tenth as fast as an
actual rodent brain operates. Nevertheless, the scientists said they detected
tell-tale patterns believed to correspond with the 'thoughts' seen by scanners
in real-life mouse brains.
It is just possible a
fleeting, mousey, 'consciousness' emerged in the mind of this machine. But
building a thinking, remembering human mind is more
difficult. Many neuroscientists claim the human brain is too complicated to
copy.
'Turning
it off might be seen as murder'
//////////////////////////////////////////////////////////////////////////////
Markram's team is
undaunted. They are using one of the most powerful computers in the world to
replicate the actions of the 100billion neurons in the human brain. It is this
approach - essentially copying how a brain works without necessarily
understanding all of its actions - that will lead to success, the team hopes.
And if so, what then?
Well, a mind, however
fleeting and however shorn of the inevitable complexities and nuances that come
from being embedded in a body, is still a mind, a 'person'. We would
effectively have created a 'brain in a vat'. Conscious,
aware, capable of feeling, pain, desire. And probably
terrified.
And if it were
modelled on a human brain, we would then have real ethical dilemmas. If our
'brain' - effectively just a piece of extremely impressive computer software -
could be said to know it exists, then do we assign it rights?
Would turning it off
constitute murder? Would performing experiments upon it constitute torture?
And there are other
questions, too, questions at the centre of the nurture versus nature debate.
Would this human mind, for example, automatically feel guilt or would it need
to be 'taught' a sense of morality first? And how would it respond to religion?
Indeed, are these questions that a human mind asks of its own accord, or must
it be taught to ask them first?
Thankfully, we are
probably a long way from having to confront these issues. It is important to
stress that not one scientist has provided anything like a convincing
explanation for how the brain works, let alone shown for sure that it would be
possible to replicate this in a machine.
Not one computer or
robot has come near passing the famous 'Turing Test', devised by the brilliant Cambridge
scientist Alan Turing in 1950, to prove whether a machine could think.
It is a simple test
in which someone is asked to communicate, using a screen and keyboard, with a
computer trying to mimic a human, and another, real human. If the judge cannot
tell the machine from the other person, the computer has 'passed' the test. So
far, every computer we have built has failed.
Yet, if the Blue Brain project succeeds, in a few decades - perhaps sooner - we will be looking at the creation of a new intelligent lifeform on Earth. And the ethical dilemmas we face when it comes to experimenting on animals in the name of science will pale into insignificance when faced with the potential torments of our new machine mind.