IBM's Eight Wonders

The Sun never sets
in...." Well, this is no advertisement pro-moting tourism in some

country, but a serious message from one of IBM's divisions-IBM Research.

With eight labs-the eight wonders-all over the world, one would guess

that the sun really never sets over IBM Research. But mind you, just

research and not the traditional R&D.






But then, what's this big deal about 'research'? Although the term 'research'
was applied frequently to the activities of IBM labs beginning in 1920s,

the mission was to improve and develop products. However, in February

1945, 'research' found a new meaning with the coming up of Watson Scientific

Computing Laboratory at Columbia University. With this lab, the mission

of an IBM lab was clearly weighed toward pure science. 'Research for

Research's sake' was to be the new motto. This fitted perfectly with

the company's overall strategy of maintaining its lead in the IT segment.

As Caroline Kovac, VP (Services, Apps and Solutions), puts it, "Our

mission is to build a technology that can become the future technology

for IBM and the IT segment." However, the big bang came only in

November 1963, when 'research' gained divisional status, even though

a formal announcement was made a decade earlier in January 1953.






Of course, it will be naive to assume that IBM Research is only into
pure science research or exploratory projects. The division is very

much into product-oriented and industry-based research. According to

Dr Alok Aggarwal, Director, Solution Research Center (SRC) India, "Though

the ratio between very directed focused formal projects (new products,

solutions or very focused on developing new technologies like Java)

and exploratory or informal projects (copper chip, deep blue) varies

from year to year, however, on an average it stands at around 2:1."








Structure and projects


Though an independent division, IBM Research falls under the overall
supervision of Nicholas Donofrio, Sr VP, Technical (includes R&D)

and Manufacturing. The rationale is simple. Any new industry-led research

should be quickly incorporated into development and manufacturing to

exploit the potentials of such research.



Most of the division's labs focus on only few segments (see table) and
are headed by VPs. Says Dr Aggrawal, "Though each lab focuses on

few areas, the beauty of IBM Research is its transparency. Because of

this, any technology developed in India or USA is available to other

labs." Currently, it is focusing its energies on four broad areas

(see chart). To replenish its energies as well as its kitty, the division

is dependent on budget allocation from the Big Blue. Of the $2 billion

budgeted for R&D, the `R' division gets an allocation of around

$500 million for itself. This is the money which the division uses for

all its projects. Since, most of its industry researches are of a prototype

nature, usually the division funds the cost of the whole research, excluding

the hardware, which are usually borne by other IBM divisions.






In the past few years, the division has been setting up new centers
around the world to give its slogan "the sun never sets at IBM

Research" the final shape. Over the last three-and-a-half years,

the division has set up new centers in Austin, USA, Beijing, China and

India. And of course, to decide in favor of one country against another

is a difficult thing for the people at the helm of IBM Research. As

Dr Aggrawal puts it, "We consider a few criteria before the decision

is taken."






The first criteria is whether there is really a need for a lab. For
example, the lab in Austin was set up due to the need to focus on microprocessors

in and around the Austin region as the area is known to be the place

for microprocessor-based research. Second is the issue of ample availability

of local talent. So chances are that it will not consider a country

or location, even if there exists a need, because of non availability

of world-class local talent. Finally, how does the lab fits in the overall

picture for the big blue. Issues like, is IBM planning to expand in

the particular market, its future presence in a country etc, need to

be answered before the decision can be taken. As Dr Aggarwal says, "As

because these centers not only attract research talent but also from

a publicity and mindshare point of view, it is important to consider

them."






Zeroing in on India


India fitted the bill on all these criteria and the opening up of the
eighth IBM Research lab at the IIT Delhi campus in early 1998 did not

surprise many. But why IIT Delhi only? According to Kovac, "When

the decision to open a lab in India was taken, it was made clear that

the lab would be in one of the IITs or the IIScs in the country."

Delhi scored over other places because of the location factor and, importantly,

the kind of research in the campus goes with the kind of research planned

for India, like weather forecasting.






Though SRC is only about eight to nine months old, it's already in the
midst of activity, both, focused and pure-science based projects. In

the former, the focus is very clear-ecommerce and supply chain. A very

significant portion of the lab is working on these activities. In the

latter, weather forecasting or rather tropical weather forecasting fits

the bill. Though there is nothing wrong with the conventional weather

forecasting methodology, but new algorithms needs to be worked out for

tropical conditions. Compared to others, convection currents play mischief,

creating turbulence, cyclones and other patterns in the tropics. The

SRC is trying to incorporate these into a new weather forecasting architecture.

It is a collaborative effort with SRC India joining hands with institutes

like IIT Delhi, Indian Meteorological Department and State Universities

of Colorado and Florida, among others.






Research is also being carried on various aspects of media. Speech recognition
under noisy environment is one such aspect. This would be an extension

to the IBM's 'Via Voice' speech recognition software which, as per the

company claims, gives up to 80% accuracy. However, it assumes a clean

environment, and so in a noisy environment the accuracy rate falls to

about 40%. The SRC is trying to incorporate a human way of interaction

into the software. For example, in a one-to-one conversation, apart

from listening, humans also pay a lot of attention to the facial movements,

including lip, eyes and cheek movements. As Dr Aggrawal says, "We

are trying to incorporate these facial movements into speech recognition

to get a higher degree of accuracy, irrespective of the environment."

Another thing feature that SRC is planning to incorporate into the speech

recognition system is media mining. "We are trying to put intelligence

into media search," says Dr Aggrawal. Most of the current systems

incorporating media mining usually try to figure out on the basis of

color combination. For example, if one searches for something like,

lets say a plant, the system will look at the color combination of the

plant and display all the pictures matching the green leaves and brown

trunk. And don't be surprised if a carpet with green and brown combination

is also displayed. Though the project is still at an exploratory stage,

SRC is trying to incorporate intelligence like, as in the above example,

concept of plants, leaves, trunks into the speech recognition system.








Comparison mechanism


In spite of the short span, how would IBM Research rate its lab in the
country? "It is hard, nearly impossible, to rate the eight labs

on an one to eight scale," concedes Dr Aggrawal. Every Research

lab would have its own idiosyncrasies and any rating would definitely

be detrimental to the morale of the labs. Moreover, since the focus

of the labs is different, it again becomes hard to quantify the results.

For example, the Chinese Lab is doing a lot of China-specific projects

like Chinese speech and handwriting recognition. The same can be replicated

in India too, but successful deployment of any Chinese-specific project

would cover nearly 75% of the population. However, in India for the

same work effort, the result would compare poorly with the Chinese counterpart

as no single language would cover a majority. Nevertheless, it would

be naive to say that there is no mechanism to evaluate the labs. The

measures are slightly different than the regular evaluation techniques.

Issues like, how well the projects are going, what is the quality of

hirers, are the customers happy with the projects and what has been

the number of research papers and patents filed, in terms of quality

and quantity?






There is an operations group which looks after these things in terms
of quality of hirers, projects and other above mentioned issues. The

big day for the labs will be when the top management, including Lou

Gerstner, evaluates on the above issues and gives the due credit to

the labs.






So what's IBM Researchs' hall to fame? How about high-temperature superconductivity,
relational databases and RISC to begin with? Among the thousands of

innovations, the above three lists among the 'Top 12' innovations as

put by the Research division's rating. The others include Scalable Parallel

Systems, Fractals, Random Access Method of Accounting and Control (RAMAC),

speech recognition, Fortran, one-device memory cell, token rings, scanning

tunneling microscope and the thin-film magnetic recording heads. Other

popular technologies on display at the labs include, datamining, supply

chain management and internet solutions.






IBM owns more than 30,000 patents worldwide and tens of thousand are
pending. Another achievement for the Big Blue is that 1998 was the sixth

year in the row when it has been the largest patent producer in the

US. And IBM Research's share in the number game-about 30% of the total

number. So the next time you are talking of an organization with such

a qualitative and quantitative record, chances are its IBM Research.

How Deep Is Big Blue Categorization

of IBM's research activities

Yograj Varma,


in New Delhi.

Back To Home

| Back

To Content

Copyright Cyber Media India Ltd.



You must read the terms and conditions before
using this service