The images derive from computer instructions that mutate randomly. The
least popular are ignored and the most popular allowed to continue developing.
Sims says the computer makes the difficult calculations needed to produce
unusual patterns and textures, but people make the aesthetic decisions.

He compares the images’ development to generations of living plants
and animals, and says the computer instructions are like genetic descriptions.
‘You can breed images that you couldn’t design,’ he says.

Swimming through the sculpture at the BAC gallery in south London is
a very pleasant and unusual experience. The Sea of Sun is a group of 26
small chambers separated by patterned chain curtains. Each chamber measures
about 1.5 by 2 metres, and you move about between them in any direction
by ‘swimming’ through the curtains. The 28 kilometres of anodised aluminium
chain, with their 1.4 million links, were assembled by hand by sculptor
Andrew Sabin. They show the faces of nine of his relatives, the word green
and psychedelic coloured squiggles and amoeba shapes.

Sabin sees links between the sculpture, virtual reality and artificial
intelligence. He says that, as in virtual reality, people move about at
their own will in an environment which is a product of his imagination,
realised with the help of a computer. Glimpses of successive pictures can
be seen through the chains and these change as each curtain is drawn aside.
The designs, the movement of the chains and the feeling of slight isolation
from visitors in other chambers combine to give it the exotic flavour of
a waking dream.

The pictures represent Sabin’s thoughts and memories, but he hopes they
are similar enough to everyone else’s for people to recognise their own
thoughts in it. He says the problem of transferring them to the computer
reflects the challenge of transferring intelligence to computers – artificial
intelligence, in which computers carry out tasks which would require intelligence
when done by people. But he sees even stronger links between art and science
in general: ‘I think scientists are attempting to describe the world. Artists
are trying to do the same thing. The same ideas can inform both fields.’

*The BAC gallery, Lavender Hill, Battersea, London SW11 (071 223 2223)
until 15 January. Admission costs £2, £1 for concessions.

Artists and critics are divided over whether computer art should be
more honest. Should you be able to tell if a computer has had a part in
creating a picture? Does the artist who uses computer rotation, enlargement
or retouching deceive the public?

An exhibition opening in Birmingham next week will fuel the debate.
‘Whose culture is it anyway?’ is organised by a network of artists who use
new technology, Zap Art International. Its members use computers in hugely
different ways, but the pictures they produce are not like the sterile doodlings
that anyone can generate with a computer graphics program. They look like
‘real’ art, an expression on paper of imagination and emotion.

Peter Gudynas of Zap Art says the artists involved are ‘promoting our
own creativity rather than the hardware’. Some of them use computers very
subtly, so that you can only be certain a computer has been involved by
peering closely at the picture and looking for the jagged lines. The same
pictures might just as well have been made with silk-screen printing or
airbrushing.

But others clearly could not have been done without the help of a machine.
Gudynas’s own art emphasises the input of the computer: ‘I’ve been trying
to exploit the fact that I use a low resolution (coarse) graphics package.
For example, I blow up or pixelate (divide into blocks) part of the image,
effects you couldn’t do without the computer.’

Some of the most accessible and beautiful works are those done by Charlie
Ward, who works with Photoshop, a photograph retouching program on an Apple
Mac. She draws her inspiration from her experience of other cultures and
alternative medicine, and uses her own pictures for meditation and hypnosis.
Only the most observant viewer would realise that she uses a computer.

Gudynas poses the question which has been asked of computer artists
for years: ‘Why use a computer to do traditional things? Computers have
an aesthetic of their own.’ Several of the artists are incorporating computer-generated
art in collage – an art form widely used among Dadaists and surrealists,
and then decades later by pop artists. Like computer artists, the surrealists
were experimenting with a new way of creating images, photography.

The pop artists of the 1960s, says Gudynas, were ‘producing a commentary
on their media landscape’, making art from the icons of the time. Now, however:
‘We use the technology available to us to comment on what’s happening around
us now, such as virtual reality and image overload.’ The computer artists
draw their inspiration from sources as diverse as vegetables, the Gulf War,
video self-portraits and alternative medicine. Gudynas’ brother Bernard,
also a computer artist, challenges his fellow artists to invent new ways
of making images. ‘It is not as easy as you think,’ he says, ‘not that many
people think about it at all.’

But is it art? ‘All art is theft,’ declares Malcolm Garrett. Even so,
several of his Zap Art colleagues are attempting to make computer art, which
is in theory infinitely reproduceable, a saleable commodity, by issuing
limited edition prints.

Entrance is free. Tel: 021 440 4221. A limited edition of 600 posters
is on sale from Zap Art International, 33-41 Dallington Street, London EC₁V
0BB. Tel: 071 250 3888.

Both chips are microprocessors which have memory and communications
links built in so that they can be easily linked to similar processors to
form parallel computers. This novel form of computer can process large amounts
of data very quickly by dividing up tasks between a number of identical
processors – anything from two to thousands. Conventional serial computers
carry out their tasks one after another using a single very fast processor.

‘Their announcement is very well timed,’ says Tariq Durrani, professor
of signal processing at the University of Strathclyde and director of the
Scottish Transputer Centre. Parallel processing researchers from around
the world are meeting in Glasgow this week for the Transputer Applications
’91 conference.

Durrani, who is chairing the conference, does not see it as a head-on
competition between the two chips – they have different architectures and
capabilities. The real battle, he says, will be over who is first to develop
the software tools necessary to put the chips to use. It is hard to write
software for parallel computers because it has to run on many processors
simultaneously and coordinate their activities.

When Texas Instruments’ C40 chip was first announced last year it was
described as a parallel digital signal processing chip. Its similarity to
the transputer did not go unnoticed, and Bob Anderson, business development
manager for microprocessors with Texas Instruments’ British arm, admits
that the design of the chip was ‘influenced by current market leaders’.

Texas’s announcement last week was of a new standard for modules to
carry the chip which will make it suitable for general parallel processing.
The modules are small circuit boards, about 10 centimetres long and 6.5
centimetres wide, which carry a C40 chip plus extra memory and connections
for communications and control of the chip.

The modules and other hardware devices are not being developed by Texas
itself but by smaller companies that specialise in parallel computing. Most
of these are in Britain, which Anderson describes as ‘the home of parallel
processing’. Similarly the software is mostly being developed in Britain.
Some of the hardware and software being developed will allow transputers
and C40 chips to work together in the same computer.

Texas claims to have the edge on the transputer in terms of speed but
Inmos, the transputer’s manufacturer which is now owned by SGS-Thomson of
France, dismisses its claims. ‘Texas Instruments is five years behind Inmos.
We don’t see the C40 as a threat to the transputer. It’s a very fast digital
signal processor but the transputer is a much more rounded part for general
²¹±è±è±ô¾±³¦²¹³Ù¾±´Ç²Ô²õ.’

Such computing problems are known as the Grand Challenges. They also
include modelling the flow of turbulent fluids, and the Human Genome Project,
an international programme to discover the entire genetic blueprint of human
beings.

Grand Challenges require such huge quantities of data to be processed
that even today’s fastest computers would need decades to finish the calculations.
But machines which can perform a million million calculations a second,
a rate known as a teraflop, should be able to cut this time to a few days.

Parystec, a small German company based at Aachen, last week revealed
its plans for its GC range of computers, which it intends to start building
at the end of this year. The most powerful model operates at just under
half a teraflop.

Falk Kubler, the founder and chairman says: ‘We could increase that
by a factor of four by minor modifications,’ making it a teraflop machine
‘by the end of 1993’. A fortnight earlier, Cray Research of Eagan, Minnesota,
said it would make a teraflop computer in 1994.

Today’s top mainframe computers calculate a thousand times more slowly,
at one or two gigaflops. They are serial computers, working on one part
of a problem at a time. If the same calculations need to be performed on
several sets of data, they have to be dealt with in a queue. The teraflop
machines, in contrast, will be so-called massively parallel computers, which
split calculations between thousands of com paratively slow processors and
do them all at once.

Parsytec expects to use more than 65 000 processors to achieve a teraflop
computer. Its half teraflop GC computer – called the T9000 Transputer GC-5/16K
– will contain just over 16 000 transputer chips (Technology, 27 April 1991).
The T9000 is the second generation of the transputer, a processor which
was designed specifically for parallel computing.

Parsytec has deliberately made the computer as simple as possible, so
that the company can sell versions of different sizes based on the same
design, down to a machine which has only 64 transputers and a thousandth
the processing power of a teraflop machine.

A major problem with any large computer is communication: data have
to be moved between memories and processors without delaying the actual
calculations. This problem is potentially worse in parallel computers becaue
they do so many things simultaneously. The T9000 includes the ‘virtual channel’
concept to speed up data transfer. It sends information in packets, labelled
with their destination, which can take any path and reroute themselves if
necessary. Earlier transputers had preset routes for sending information.

Despite all these technical advances, neither the transputer nor the
GC computer has been tested in practice. The T9000 is not due to go into
production until the end of this year. Parsytec is confident that it will
have enough T9000s, working properly, to make 30 GC computers next year.
It already has six orders, from Germany’s national centres for fluid mechanics
and engine bearings, the Naval Research Laboratory in Washington and two
computer laboratories in Germany and the US.

British scientists would like to acquire massively parallel computers
too. Howard Wheal of the University of Southampton may be lucky because
Parsytec’s British office is on the campus.

Wheal is a neurophysiologist, working on the hippocampus. This part
of the brain is normally involved in transferring information between the
short-term and long-term memory, but it also triggers some types of epilepsy.
Wheal is using electrical circuits to mimic the activity of the hippocampus,
to see whether there are instabilities which could be linked to epileptic
episodes.

‘I use 3000 electrical circuits to define the electrical activity of
a single cell,’ he says, ‘and I can model 10 000 neurons with 10 000 connections
now. It takes seven to nine hours for a single simulation.’ His computer
works at a a rate of 20 million calculations per second. He would like to
add more neurons to the model. The human hippocampus contains a million
neurons with 10 000 connections, which a half teraflop computer could easily
simulate in a few hours.

‘The whole brain has about 10 ¹¹ neurons,’ says Wheal. Even
a teraflop computer could take several days to do a simple simulation of
the entire brain. Modelling such a large number of neurons could provide
clues to the causes and treatment of pathological conditions such as epilepsy.
Wheal says that epileptic episodes ‘may involve many nuclei in the brain.
One might put together a large number of nuclei to model its pattern of
²õ±è°ù±ð²¹»å.’