¿ìè¶ÌÊÓƵ

On A sunny spring day in Yorktown Heights, New York, researchers at IBM’s
Thomas J. Watson Research Center were preparing for a trip to Las Vegas.
They were not going to test a new gambling system, but the trip was almost
as outrageous a venture for the normally cloistered IBM scientists. They
were going to Nevada to sell a new high-performance computer system to the
film and television moguls gathered at a National Association of
Broadcasting exhibition last month. Selling products to Hollywood is not
something for which IBM’s research division is renowned. It is more usually
associated with scientific endeavours that have won its researchers five
Nobel and many other top scientific prizes. But times have changed, and the
Las Vegas trip is symbolic of the company’s new attitude to research.

The computer that the scientists were trying to sell Hollywood is the Power
Visualisation System. Costing between $500 000 and $2
million, the PVS had originally been developed for scientists and designers
who need to visualise complex data. But the ability of the PVS to handle the
large amounts of data generated by digitised photographic images quickly
attracted the attention of companies that produce special effects for the
screen. The research division had already sold four systems to the film
industry, including one to Boss Film Studios of Los Angeles, which produced
the effects for Batman Returns and Alien3. In travelling to Las Vegas, the
research division hoped to exploit this interest. But while the research
scientists are delighted with their unexpected success in the film world,
the novelty of working in a market that, for IBM, is culturally foreign is
proving somewhat disconcerting, not least in sartorial styles. At IBM, where
the conservative blue suit is something of a uniform, one shocked member of
the Yorktown Heights group was heard to remark, ‘Those Hollywood guys wear
black shirts and orange ties.’

Yorktown Heights is the headquarters of IBM’s research division, although
the company has a number of laboratories scattered around the world. About
1700 employees are housed in the Thomas J. Watson building, a sweeping arc
of stone and glass that overlooks wooded hills. Since the building was
completed in 1961, 47 years after the company was founded and just as its
commercial power was reaching its peak, researchers have been able to watch
for deer in the woods from the curved windows lining the corridors before
returning to the windowless laboratories where they have worked away,
insulated from commercial pressures. But more recently there has been a
shift of attention to the world beyond the perimeters of Yorktown Heights
and the boundaries of the IBM organisation.

The event which precipitated this shift was the company’s first-ever loss in
1991. A year later the $2.18 billion loss had escalated to Dollars
4.9 billion – one of the largest corporate losses in history. The roots of
IBM’s problems lie in its sheer size which, for many years, had enabled it
to dominate the computer market but eventually hindered the company from
sensing or responding to changes in the industry. ‘We as a company didn’t
feel the heat of the marketplace on our backside,’ says James McGroddy,
director of the research division.

THE HUNT FOR ‘WHITE SPACE’

Change has been sweeping through IBM , and the research division
has received its share of the fallout. The research budget, which in 1992
was around $550 million, will be reduced by around 10 per cent this
year, and about 300 staff will be shed, bringing the total employed down to
3000. But the biggest change is that the pressure is now on for the division
to remove the layers of insulation between itself and the marketplace.

The ‘customers’ of IBM’s research have traditionally been the company’s own
technology and product divisions. New inventions have had to work a slow
passage from the laboratory through development to the marketplace. But the
exploitation of research by the internal divisions is no longer enough now
that IBM’s profit margins are lower, says Philip Carnelley from the London
office of New Science Associates, a group of analysts specialising in
advanced and emerging technologies. IBM’s research division has been forced
to take a more active role in commercialising its inventions, he says.

According to McGroddy, there is now a greater effort by the division to
discover the needs of IBM’s customers in the outside world, and to direct
research accordingly. In addition, the division is actively looking for what
McGroddy calls ‘the white space’ – those areas of the market where IBM is
not active or where new technologies might offer opportunities.

‘This is tremendously important for IBM because if you look at the growth of
the computer industry over any extended period of time it’s the new things
that come along that provide a very significant fraction of the growth,’
says McGroddy. In the past decade, for example, personal computers and the
more powerful workstations have spawned markets worth billions of dollars.

But to accommodate activity in new areas, some established areas of research
are being reduced or halted. ‘Stopping things is much harder than starting
things,’ says McGroddy. The most significant reduction is in semiconductor
research. In 1967, IBM invented the dynamic random access memory, or DRAM,
chip, the main memory chip used in computers. DRAMs store a single bit of
information as an electrical charge in an electronic circuit formed from a
single transistor and a capacitor. Previously the main memory in computers
was made from magnetic materials. Semiconductor memory is considerably
faster and more compact than magnetic memory, though it is more expensive.
But over the years, other semiconductor manufacturers have developed their
own designs for storing information in electronic circuits and DRAMs have
become commodity products available from a number of suppliers. Since
expertise in DRAM technology no longer provides a means of distinguishing
IBM products from those of its competitors, it is no longer a priority for
research. Other research on semiconductors is being transferred to joint
projects with companies such as Toshiba in Japan and Siemens in Germany.

These kinds of decision are guided by ‘The 10-year outlook’, an annual
strategic view of the future which the research division is responsible for
developing for the company. One of the objects of the exercise is to
identify those white spaces, and the Power Visualisation System presented at
Las Vegas is an example of one product developed to fill a such as space.
The way in which the system was developed also provides a model of how the
research division might increasingly operate in the future, now that it can
leapfrog the company’s bureaucratic structures to get its inventions to
market early.

In 1990, the 10-year outlook identified visualisation as a technology with a
strong potential for growth in the following decade. One example of
visualisation is the transformation of complex data – such as the readings
of air pressure, wind speed and direction, temperature, precipitation and
cloud cover from a number of stations scattered around the world that form
the basis of weather forecasts – into three-dimensional pictures that might
use colour and animation to convey the dynamics of the data. Another is the
detailed pictorial representation of product designs, such as cars.

To exploit this market, the research division took the unprecedented step
of bringing together a group of researchers into an autonomous unit which is
the closest IBM has ever come to setting up an independent start-up company
to exploit new technology. The group included experts in high-performance
parallel computers, advanced graphics systems and marketing. IBM committed
an estimated $25 million to the project and the group moved into
premises away from the main research site, operating in secrecy. Working at
the kind of breakneck speed more usually associated with the entrepreneurial
start-ups of Silicon Valley, the computer industry’s heartland in
California, the group produced the PVS in only 13 months, unveiling the
machine in January 1992.

IBM’s view of the market for high-performance computers for these kind of
demanding applications is supported by Paul Drath of Cambashi, market
analysts in Cambridge, Britain, which specialises in scientific and
engineering computing. According to Drath, the market in Europe alone in
1991 was $4600 million, with growth predicted at 11 per cent. But
this is a relatively new and open field, with no supplier dominating, says
Drath.

The PVS project presents a stark contrast to the way IBM handled the
invention of a reduced instruction set computer (RISC), an approach to
microprocessor design now used by most computer manufacturers. Although an
IBM researcher, John Cocke, first put forward the basic ideas for RISC in
the 1970s, IBM did not seriously exploit the technology until 1990 and four
years after rival manufacturer Hewlett-Packard had moved all its machines
over to RISC.

MISSED OPPORTUNITIES

RISC is based on the concept that, in most programs, microprocessors spend
more than 80 per cent of their time performing a limited number of simple
instructions, such as ‘store’, ‘load’, ‘add’, ‘subtract’ and ‘multiply’. By
the mid-1970s, however, microprocessors were being designed to perform many
complicated instructions, in some cases more than 350. These instructions,
such as a ‘loop’ command where a program requires the processor to perform
an activity over and over again until a threshold is reached, could occupy
the processor for some time. Cocke envisioned the computer translating the
complex instructions written by a programmer into a sequence of simple
instructions, each of which could be handled extremely quickly by the
processor. The speed gained in handling the simple instructions more
efficiently outweighs the inefficiency of breaking down complex instructions
into simpler parts.

Although the research division built a RISC-based computer in the 1970s
(called the IBM 801 after Building 801 at Yorktown Heights, where it was
invented), there was little attempt to exploit the concept commercially. In
frustration at the lack of progress on the project, one of the
researchers, Joel Birnbaum, left IBM in 1984 and joined Hewlett-Packard
where he was given the resources to develop a set of microprocessors and
associated hardware and software. Although based on Cocke’s original ideas,
Birnbaum developed the technology in a different way, avoiding any disputes
over patents or copyright. Hewlett-Packard’s decision to base all its
computers on RISC has enabled it to produce some of the industry’s fastest
workstations. IBM, however, did not get a competitive RISC-based RS/6000
workstation to market until February 1990.

‘We weren’t smart enough to exploit it (RISC) as fully as it could have been
exploited,’ says McGroddy. ‘To do great work is in some ways easy. You
provide a set of resources and hire a great set of people and great work
will be done. The hard part is to take that work and make it vital to a very
complex enterprise such as IBM.’ There is a belated recognition that speed
is of the essence in transferring technology into the marketplace. And as
the pace of technological change speeds up, so the value of research
depreciates more rapidly. ‘The things we do ‘rot’ at a rate of about 30 to
40 per cent a year,’ says McGroddy. ‘It’s like fresh strawberries – you
can’t freeze them.’

The research division is trying to become smarter by managing its projects
more effectively. Much of the responsibility for this management lies with
Mark Bregman, vice-president of technical plans and controls. He joined the
division in 1984 as a researcher on a project to measure the mass of the
neutrino, one of the few speculative research projects undertaken by the
division. In 1989, Bregman moved to a semiconductor research project at an
IBM laboratory in Japan, eventually becoming director of the laboratory and
also of the manufacturing plant to which it was attached. There, he says,
he learnt the importance of disciplined management and the rigorous
execution of tasks – an attitude he is applying to the research division
since he moved back to Yorktown Heights in January.

Bregman rejects the idea that tight management controls smother scientific
creativity. ‘Applying discipline to the management of research doesn’t mean
telling the researchers what to do or how to do it. It really means having
the mechanisms in place to ensure that they are achieving what they want and
are able to achieve.’ This can include providing extra resources or
management help when projects run into trouble, or help with mundane tasks.
‘When you have a researcher spending time rewiring their lab instead of
doing research, that’s not very productive,’ says Bregman.

One of the things Bregman is trying to do is to ‘minimise the boundaries
between disciplines’. Internally the division continues to be organised into
disciplines, much like a university with its physical sciences, computer
science and mathematics departments. But the approach now favoured is the
one taken with the PVS project – bringing together people from different
disciplines to achieve a specific business objective.

These kinds of project take 95 per cent of the research budget. The
remaining 5 per cent (which amounts to $25 million in this year’s
budget) goes towards blue-sky projects that have no obvious links with IBM’s
business. Although they make up a small proportion of its overall effort,
the blue-sky projects have an impressive track record. One produced the
scanning tunnelling microscope (STM) – a device so sensitive that it can
produce images of surfaces showing the arrangement of individual atoms. The
STM was invented by IBM researchers Gerd Binnig and Heinrich Rohrer at the
division’s Zurich laboratories – and it earned them a share of the 1986
Nobel Prize for Physics.

The STM uses a probe with an extremely fine metal tip which, when it is
brought very close to an object, interacts in a fine but measurable way. The
electron ‘clouds’ of the tip and the object overlap and the amount of
overlap determines how much current can flow between the tip and the object
when a small voltage is applied. Just two atoms can dominate the process –
the front atom on the tip and an atom in the region being probed on the
object. As the tip is scanned over the object, the measurements of current
are used to create an image of the object’s surface. As it turns out, the
invention has proved useful to IBM. In semiconductor manufacturing, for
example, the STM has revealed the surface structure of silicon, allowing
more complex chips to be designed.

Despite the changes and cuts in funding and people, IBM still has one of the
biggest corporate research divisions in the world. In the five years to
1992, the company might have spent a slightly lower percentage of its
revenues on research and development than its nearest rivals in the computer
industry – around 10 per cent, compared with about 12 per cent spent by
Fujitsu of Japan and Digital Equipment of the US (according to estimates by
The New York Times). But thanks to greater revenues, IBM’s R&D budget was
still three times larger than that of Fujitsu or Digital.

Bregman says the division is ‘trying to ensure our researchers are more
worldly’. There are many signs of this new contact with the outside world.
One of the researchers developing programs for the PVS is Clifford
Pickover, well known in computing circles for what he calls ‘the creative
use of computers – simulating, visualising, speculating, inventing and
exploring’. Pickover continues to work mainly with other scientists in the
division, but in the future expects to work more directly with outside
customers. In the meantime, while the visualisation group was getting ready
for Las Vegas, Pickover was preparing for lunch with the manager of Robert
Plant, the heavy metal singer formerly with Led Zeppelin. Plant wants to use
Pickover’s images as stage backdrops on his forthcoming tour.

It is no longer docile deer in quiet woods that researchers at the Thomas J.
Watson Center are encouraged to see from the windows – but white spaces,
customers and competitors.

Clive Davidson is a freelance journalist specialising in computing.

* * *

Changing times

IN 1990, IBM was still America’s most valuable corporation, worth Dollars
106 billion. By the end of 1992, its value had tumbled to under $28
billion. Ironically, the factors which brought the company its massive
success were those that eventually brought about the reversal of its
fortunes.

IBM has always dominated the market for mainframe computers, the large
central computers which earned large profits. The gross margin on a
mainframe is estimated to be up to 80 per cent on machines costing $1
million or more. Until very recently, all mainframes were proprietary, using
technology exclusive to each manufacturer. Software or peripherals such as
storage devices from one manufacturer would not work with any other system.
This encouraged many customers to buy exclusively from IBM.

During the 1980s, small but powerful computers appeared, including the
personal computer (PC) and the workstation, which took work away from
mainframes. Consequently the sales of mainframes have fallen – by around 10
per cent for IBM in 1992. Although IBM brought out its own PC in the early
1980s, it did not bring the same profits as mainframes.

With the PC, IBM did not use proprietary technology, and ‘clone-makers’ like
Compaq and Dell produced almost identical machines that could run all the
same software as the IBM machine. Although IBM remains the largest PC
supplier it has had to compete with hundreds of clone-makers for a share
of the market, with profit margins on machines less than half those of
mainframes.

Meanwhile, the 1980s also saw the beginning of the ‘open systems’ movement
in which customers began to demand that computers conform to international
standards so that hardware and software from different suppliers could work
together. This allows customers more choice of supplier. Many rival
companies, such as ICL and Hewlett-Packard, adopted the standards far more
quickly than IBM and so attracted customers who might formerly have bought
exclusively from IBM.

IBM also has huge running costs. The scale of the problem is highlighted by
the fact that IBM employs 300 000 people; Microsoft, the PC software
supplier which now has roughly the same stock market value as IBM, employs
only 12 000 people. To reduce costs, IBM cut its workforce in 1992 by 40
000, with at least 25 000 staff to go this year.

In an effort to improve performance, the company has been undergoing major
restructuring, with a plan to break up the organisation into 13 autonomous
business units, each with its own financial targets and balance sheet.