The eighth IBM Research
Laboratory was set up in New Delhi during the period 1998-99. The New
Delhi lab is unique because it is the only one in the south Asian region,
and incorporates IBM’s global shared university program. As part of the
ramp-up process, the lab was visited by IBM’s worldwide head of
research, Paul Horn, who has been working with IBM for the last 19 years.
He directs the company’s worldwide research program across the Yorktown,
Almaden, Austin, Zurich, Haifa, Tokyo, Beijing and New Delhi laboratories.
Horn is a specialist in the areas of solid state physics, semiconductors
and storage. He has a doctorate degree from the University of Rochester.
Excerpts of an exclusive interview with him across a wide range of topics:
How
are research areas decided between the various global centers? What are the main
research areas?
We look at the laboratories and
see how best we can do the work, based on the skill-base in the local regions.
Electronic commerce and electronic business are the big focal items that we have
here. Work on little pervasive devices that allow you to
connect to the network and do business is another area. There is an outstanding
effort in what we call deep computing spirit with projects like weather
forecasting.
If you look across all the
laboratories just about 15% is what we call white space research...you have no
idea what the product is. Then about 25% is future work, which we are pretty
sure will impact servers or semi conductor storage. And the remainder is on next
generation products.
Research
laboratories are generally well known for conducting activities with lack of
industry relevance. How is this discontinuity tackled in IBM?
This disconnect is rapidly going
away and in fact laboratories like ours are part of the reason why it is going
away. The industry has a connection with
individual faculties in the labs, and they work
on projects that they mutually agree upon, and are
both excited about. They purposely define the programs in a way that may well
lead to rapid
commercialization.
In general, the key thing is that
you rapidly flow technology into the marketplace. Then it is irrelevant. You can
try to create all sorts of barriers
but in the end the best protection is the speed in utilizing it.
How
are research areas distributed across the IBM labs?
There are laboratories that are
closely linked to the development or manufacturing facilities. For example–the
storage work–most of the time it is manufactured and developed in San Hose and
research work on storage is elsewhere. The semiconductor work tends to go in the
Watson laboratory. For solutions, the work is in many laboratories around the
world. Since IBM is so broad there are a lot of things going on.
How
do you encourage interaction with the academic community?
We try to have exploratory
programs in all the laboratories, because often exploratory programs are the
ones that make the best contacts, for example, with the university students. We
use exploratory programs to flow in new knowledge and new ideas, and make
partnerships that allow flow of new knowledge and new ideas from outside.
It is very important to think
about research in the context of how the economies around the world are
changing. Let me give you some thoughts. We are rapidly moving into what many
futurists have called the information age. We have gone from an agricultural
economy to an industrial economy to an information economy. If you are in an
information age, the information that is available free is enormous. Success for
a company, for an individual,
or for a country is to find ways to flow in key information that one needs, and
to use that to distinguish oneself.
How
does that apply to the research center at IIT Delhi?
In many ways the reason why we
created this laboratory was because we thought we could build great university
programs in connection with, first the IIT here in Delhi and then the other IITs.
IBM
files a large number of patents every year. Is there a planned program for
patent generation and technology licensing?
We work very hard at it. We have
valuable
technology, and one way to get a return for that
technology is to license know-how, patents and the technology.
Are
there any particular areas where patents are proliferating from IBM...
It is in all areas...
...What
about patents for the deep computing algorithm?
It is conceivable that you can
patent algorithms and sometimes we do patent algorithms–for example the
algorithms we have used to look at and understand the human intelligence.
The key thing–I do not want you
to think that when we go in we try to create patents. What we try to create is
products and solutions. Sometimes there will be patents as well and they may be
valuable. If you can understand a better way to get the underlying meaning of a
sequence that could be valuable to a lot of pharmaceutical companies, for
example, we would like to use that algorithm to
create a set of solutions for pharmaceutical companies that are interested.
Maybe there is a patent, maybe not.
Do
you see the deep computing algorithms being used in the commercial and business
space?
I think because there is so much
data freely available on the internet, there is going to be a very big problem.
As an example, we have in IBM, within one company, one and a half million pages
on the internet. We had to do a research project to find out how many pages we
had on the internet and the intranet. Now there are probably three or four pages
that you have got to read, but you never know how to find the way to those three
or four. Deep computing will allow you take this massive amount of data that is
getting collected–for example about customers doing work on pervasive devices–and
make some meaning out of it. Datamining technology, algorithms for patent
matching–these kinds of things are going to be absolutely critical for
business.
Will
the advances in algorithms help in the development of artificial intelligence?
I do not think artificial
intelligence is the right way to think about it. It will be a set of tools–business
intelligence tools among other things, and algorithms–that will allow meaning
out of massive amounts of data domains. You are starting to see those tools.
What
about research in the area of internet and ecommerce?
We are looking at everything from
understanding economies to digital economies, understanding how they operate to
actually building portals and websites. For example, there is a paper company
that auctions off paper from a variety of suppliers. Now imagine you can do that
online–so if you want to buy paper from a mill you go to their website. On
their website, all suppliers that have paper available in their mills list what
they have. You could pick and you could bid. We are helping them do that.
Are
your e-com models sufficient to predict changing consumer preferences?
That is a hard question. I do not
think anyone can really predict the changes in the future very well. The changes
are so great. We have some thoughts but I think you have got to believe that the
changes will be so phenomenal that it is hard to predict how consumers and
society will react to them.Â
What
is the most critical edge for organizations in the information age?
For any company in the internet
space, it is going to be to find ways to make sense out of all its data.
What
are the activities in the area of consumer information appliances?
Lots of neat things. I will give
you an example of a neat project. There is a pilot project we built with a
supermarket called Safeway in United Kingdom. This project deals with giving a
palm pilot to a bunch of their customers. This is a standard palm pilot, except
the palm pilot lists what is available from Safeway. What they do at home is
whenever they want something, instead of writing it on the refrigerator, they
just check it on the palm pilot. When they are ready they put the palm pilot in
the cradle saying when they want to pick it up. It gets transmitted to the store
and is already in a bag for you when you get there. So this becomes a consumer
appliance that allows you to connect over the internet and order your groceries.
Is
IBM looking at moving into the consumer appliance area?
Well, we may or may not actually
build a palm pilot–we now have what is called an IBM workpad. In order to
operate in a pervasive space you may not need a fundamentally new client but it
would be good to have the software focused on some solution.
Around
what OS is IBM looking at building its appliances?
Essentially you have got Windows
CE, which is what the IT vendors would prefer. But it is not going to be like
the Windows dominance in the desktop. You are not going to have Win CE down on
smaller devices–it is just too big.
How
are graduates from IITs recruited into the IBM India lab?
We send a letter to the heads of
departments of computer science and electrical engineering of all the six IITs.
Essentially, we take the top three people by their grade point average from each
of the departments. Typically, what happens is that IIT Delhi may get a couple
of more students in because they may be working on projects during the year, and
automatically get absorbed. Starting next year we are hoping to get five top
guys from all computer science departments of all the six IITs, and two each
from the electrical engineering departments. We will assign a researcher from
the lab who will work with them, advise them, and mentor them as they go along
because they have had no experience whatsoever in doing research.
Is
it correct that Indians form or are likely to form the largest share of
employees recruited in IBM laboratories?
I do not know how you define the
largest number. Right now this is a tremendous environment for bringing in new
talent. You can just say that the calibre of people we have been able to get has
really been outstanding.
Does
China have the ability to compete with us in terms of infotech professionals?
Yes. We are able to get wonderful
people there. There is more competition in information technology in Beijing
because there are other research organizations. Here
nothing. Ours is really a unique opportunity. I expect eventually that other
companies will enter.