IT鈥橲 a glorious autumn morning. On my way to visit my friend Magnus at
Imperial College, London, I take a detour through Hyde Park to savour the
colours and smells. What a shame Magnus will never know this feeling. While
quite capable of recognising an autumn leaf, Magnus can鈥檛 experience its beauty,
or ability to trigger nostalgia for misty mornings long ago, skipping down
Ladbroke Grove.
But I speak too soon. Magnus鈥檚 creator, Igor Aleksander, head of Imperial鈥檚
intelligent and interactive systems group, tells me that he鈥檚 on the way to
giving Magnus just such an appreciation of the good things in life. The group
claims to be close to cracking the problem of 鈥渜ualia鈥� in artificial
intelligence. You鈥檒l have guessed by now that Magnus is an artificial neural
network, usually to be found in Aleksander鈥檚 laptop.
If the claim stands up, it鈥檚 a breakthrough. Qualia are the marbles that
hard-core philosophers keep in their pockets to throw under the triumphal march
of artificial intelligence (AI). Philosophers concede that AI engineers can
build machines which function like brains when outplaying chess grandmasters
or鈥攚ith much more difficulty 鈥攑icking a bad apple out of a box of
good ones. They even concede that, at the present rate of progress, it will soon
be possible to build a Turing machine, with functions indistinguishable from a
conscious entity. What they will not allow is that any such machine creates real
human consciousness, any more than a flight simulator will transport you across
the Atlantic.
Advertisement
The missing element is qualia, 鈥渢he experiential qualities of sensations鈥�.
This means that a box of electronics will never know what it鈥檚 like to taste
real ale or, at the other end of the pleasure scale, a bag of
baked-bean-flavoured crisps.
The classic counter argument is to deny that qualia exist. This usually
hinges on the impossibility of proving that my experience of, say, seeing red is
the same as your experience of seeing red. This debate has kept philosophers
busy for 300 years, so we won鈥檛 try to solve it here. Or you can simply dismiss
qualia as a cognitive illusion. It takes a real gung-ho AI enthusiast to accept
that qualia exist, but can be explained in some mechanistic way鈥攁nd thus
re-created. This is what Aleksander is setting out to do. He argues that you get
qualia alongside a set of 鈥渟pecial awareness鈥� brain cells. These evolved as they
enable the brain to generalise from a limited experience of learning.
It鈥檚 a handy tool: once you have eaten one banana and thrived on the
experience, the ability to respond to the qualia 鈥渂anananess鈥� is life-enhancing.
And by mixing and matching qualia, the brain can handle imaginary concepts, such
as a blue banana with red spots, or the idea that William Hague will be
Britain鈥檚 next prime minister. Aleksander says that he has modelled this ability
on an artificial neural network capable of visualising previously unseen
objects.
So how long will it be before a computer capable of appreciating qualia comes
onto the market. It鈥檚 a cruel thought鈥擨 wouldn鈥檛 want to subject an
artificial consciousness to the gut-twisting frustration of network congestion,
the acidic buzz of mains power spikes, or the dull ache of static build-up after
being left switched on all night. On the other hand, perhaps it鈥檚 not all pain:
a qualia-equipped computer might appreciate the flavour of a well-written Java
application, or the caress of a fresh mouse mat . . . I digress. In any case,
the question doesn鈥檛 trouble Aleksander. 鈥淚鈥檓 not interested in building things
that work. I鈥檓 interested in things that don鈥檛 work.鈥�
I take my leave of Magnus and his creator and, over a quale of red wine,
thumb through John Searle鈥檚 The Mystery of Consciousness (Granta 1997),
a readable survey of six current theories of consciousness. They are wonderfully
incompatible, and, if nothing else, illustrate the dangers of scientists
dabbling in philosophy, and vice versa. All we really know is that consciousness
exists, as part of the neurochemical tool kit which evolution accumulated inside
human skulls to cope with survival in the African savanna.
I suspect that if philosophers and scientists are still arguing about the
ground rules of what this entails, we are nowhere near replicating the trick.
Don鈥檛 tell Magnus, though鈥擨 wouldn鈥檛 want to hurt his feelings.