 Expanding Areas of Research
––– Please tell us the details of how CSL was
founded.
Tokoro: The year before CSL was founded (1987), I
was working as an assistant professor at a university
and received an unexpected visit from Dr. Toshitada
Doi, who was then an executive director at Sony.
He asked if there was something interesting that
we could do. How about creating a laboratory for
computer related research? I proposed that we
create a world-class research laboratory so good
that researchers from around the world would notice
it and come to participate. A few days later, Dr. Doi
came back and said “It's been decided. Will you
accept the position of director of the laboratory?”
Although I had previously had no experience with
Sony, I thought it must be an interesting company.
––– The range of subject areas is certainly wide:
it includes brain science, systems biology, and
econophysics.
Tokoro: CSL’s history can be roughly grouped as
follows: during the first ten years we mostly worked
in pure computer science, for example, distributed
operating systems. For the next ten years, we shifted
our focus to asking how computers could be useful
to society and mankind. Dr. Kitano and Dr. Takayasu,
who are also here today, are doing world-class work.
Kitano: When I joined CSL in 1993, AI was the
main focus. I worked on an entertainment robot
(which was later commercialized as AIBO*1) and
managed RoboCup*2. Starting in 1994, however,
much of my time was allocated to biological research. I did not expect to see any breakthroughs
in AI for the next five years. I thought that it would not
be too late if we immersed ourselves in biological
areas for the near future and then returned to
AI. However, what happened was that we simply
remained focused on biological research.
Takayasu: It turns out that the vast majority of
economic laws that we learn in economics have
not been proved. Most are just hypotheses, and it is
only recently that it has become possible to check,
using real data, whether or not they actually hold.
Nowadays, the “footprints” from economic activities
are all retained in computers. For example, there
are about one million products that are handled
by supermarkets, but it is now possible to analyze
changes in demand in detail using POS data*3. The
amount of data that can be handled in economics
has increased a million times due to computer
advances.
––– Does that mean that you are analyzing
actual data using methods from physics and
explicating the behavior of complex economic
phenomenon?
Tokoro: In sciences up to now, the basic approach
has been to break down the subject finely and
drill down to simpler and simpler parts. However,
the problems we are facing now are horrendously
complex. These problems cannot be solved by just
looking at the parts. Another issue is that some
action thought to be helpful for a particular part turns
out to be detrimental to another part of the problem.
Environmental problems are truly of this nature,
as are many problems in biology and economics.
Developing approaches that use scientific
standpoints and methodologies to handle complex
problems has become an important theme at CSL.
Selection and Evaluation of
Research Topics
––– While the research targets are established
by the researchers themselves, it seems that
while there is freedom, the direction is firmly
maintained.
Tokoro: We all share a common understanding
that the problems we face immediately cannot be
solved with traditional approaches. I think it can be said that the
concept of “open
systems science”
became firmly
established in this
context in which
such researchers
can argue again
and again. When
we meet a new
researcher, we
hold thorough
discussions and
hire people who
are sympathetic
to our concepts. Thus we are highly accepting of a
various desires within that range.
Kitano: Research which proceeds by looking
at individual phenomena can be carried out at
universities and academic conferences. One
strategy is for us to focus on research that can have
a great impact because we are CSL. Although this
may occasionally take 10 or 20 years, it can lead
to research that is truly useful to humanity and the
world.
Tokoro: This is because we do not look at things
with an excessively short span.
––– However, employment is on a one-year
contract. When an annual contract is up
for renewal, what standards do you use for
evaluation?
Tokoro: By the sparkle in their eyes. If we can see
that they put their reputation as a researcher on the
line and adopt an approach that is different from
that of ordinary people or if they have goals that are
different from those of ordinary people, we ask them
to continue that research. Also, CSL is an open lab,
and all researchers take part in a variety of projects,
both within Japan and overseas, and automatically
become the leading player in their area. If that
happens, we know. The level of a person's research
becomes self-evident from the amount of external
activity the researcher is involved in, the people they
associate with, what conferences they are invited to
deliver lectures at.
Hints Arise at the Points were
Different Fields Come in Contact
––– While I understand that there are
opportunities to present results to other
researchers within CSL itself, in what way to
researchers listen to discussions in other fields?
Takayasu: Despite myself, I sometimes find myself
thinking “That’s it!”. For example, recently I have
been analyzing the network of trading relationships
between the one million or so companies in Japan,
but there is a great similarity between this work and
Dr. Kitano’s work on mutual interactions between
biochemical substances in immune systems.
Especially when one makes graphs and considers
quantities.
––– Networks often have a hub.
Kitano: That is one idea. I am also considering
introducing economic models, such as portfolio
theory and game theory, into the theory of biological
evolution. There is also something I'd like to ask Dr.
Takayasu about, namely, that when studying the
economic crises of the past, there are a variety of
stabilizers that are built in to in economic systems.
But despite these, events such as Black Monday
(the stock crash of 1987) and the subprime problem
still occur. I think that the reason these occur is
that the stabilizers inversely cause problems when
they move in unexpected directions. It seems that
derivatives, which were developed to hedge risks
of market fluctuations, have become a source
of instability in
the financial
markets. From
my standpoint as
someone who
studies the fragility
of living organisms,
this looks very
similar to the state
where the excretion
of cytokines*4,
which are excreted
to kill viruses,
cannot be stopped
and the excess
cytokines cause pneumonia.
Tokoro: This is the fragility of living organisms
and the fragility of economics. If we consider this
from the perspective of systems structures, there
are aspects that are similar. Of course, there are
differences as well. It is because these relationships
are subtle that we can get hints from comparisons.
Kitano: Dr. Takayasu is always mentioning the point
that all economic data is now stored on computers.
Unfortunately, that is not possible in biology. But
perhaps the theories I have can be proven within
economic systems. I think that this is a theme that
will hold much interest in the future.
Takayasu: The
problem that
I have been
concerned with
recently is malice.
In the subprime
problem, there
have been people
who thought up
schemes in which
only they would
profit. Similarly,
in the biological
field, there are
organisms that
think that it is good if only their descendents survive.
This is a difficult concept for physicists, since it is
inconceivable that there would be malice between
pairs of physical substances. However it may be
possible to establish completely new intellectual
issues if we approach this from the standpoint that
even though the phenomenon itself is impossible,
there may still be similarities.
Tokoro: Hmm, can we really say that physical
substances have no malice? Anyway, note that
we have used almost no technical terms in this
discussion. This is the essence of “open systems
science”. I believe that there will be no progress in
science if we do not step back to this level.
Science and Engineering
Tokoro: One thing I’d like to emphasize here is that
the reason CSL has succeeded so far is that it is
an environment in which researchers with a strong science orientation often discuss and debate with
researchers who are closer to engineering. Dr. Jun
Rekimoto, who worked as director of the Interaction
Laboratory, has always maintained his lead in terms
of number of patents acquired at CSL, and is also
by far the top in the Sony Group as a whole. The “cybernetic earth” concept that he advocates is
both close to science fiction and at the same time
firmly positioned close to Sony's business. The point
that these current business areas and fundamental
research that could create new business
domains are tightly intertwined is both a defining
characteristic and a strength of CSL.
––– Dr. Kitano was appointed as the new
director this July.
Kitano: I’m am often asked how I will impress my
personality on the lab, but since CSL is currently
doing well and producing results, I am not planning
on making any major changes. One area that is,
however, always on my mind is the energy and
environmental problems that will be the major issues
of the 21st century. What sort of societal systems will
be required to respond to these issues and what sort
of technological systems will be suitable? And, in
particular, what sort of approach should CSL adopt?
I hope to take action early in areas where we need
corrections to our approach.
––– Do you have opinions or ideas for Sony itself
or for the semiconductor business?
Tokoro: I think the semiconductor business should
take new and drastic actions. Including, for example,
considering giving up silicon. One example would be
business based on generating solar energy just by
painting dyes on window glass. Furthermore, LCD
shutters could be used for blinds.
Kitano: We should make all home and office
building windows be solar panels. If necessary,
we could even collaborate with universities, rival
companies, and major construction companies.
Tokoro: If there’s anyone reading this who would
like to work on these problems, and feels in any way
constrained in their current position, please, come
to CSL. We welcome people who argue with their
superiors.
*1 AIBO: An entertainment robot in the form of a puppy sold by Sony. “AIBO” is a trademark of Sony Corporation.
*2 RoboCup (Robot World Cup): An international joint research project for soccer played by robots proposed in 1993 by Japanese researchers led by Dr. Kitano. RoboCup is intended
to promote and advance AI and robotics engineering research.
*3 POS data: Point of sale information management is a method for collecting product sales results in individual item units.
*4 Cytokine: The collective designation for biological activation factors that are produced by cells for all biological structures and that provide mutual interactions between cells.
Systems Biology
A field that aims at understanding biological phenomena
not as individual parts but globally as a dynamic system.
Systems biology considers, for example, biological
robustness as a systems level principle. Based on this
concept systems biology attempts to understand and
modify cells at a basic level. It is furthermore applied in
areas such as research on cancer treatments.
Econophysics
Econophysics attempts to explicate economic phenomena created
from the complex entanglements of human actors using approaches
from physics, for example chaos theory and fractals. Since financial
market transactions are currently completely computerized, much
as police investigators determine a criminal’s activity from footprints,
econophysics researchers can search for underlying principles lying
hidden in economic phenomena from this transaction data. Econophysics
methods are also used to analyze production yields in semiconductor
manufacturing.
Cybernetic Earth
In addition to the increasing penetration of the Internet, when a
wide variety of sensor data becomes available over the Internet,
the real and the cyber (or virtual) become indivisible, and we
reach a state where the earth must be seen as a cyborg. Since
the elemental technologies already exist, the problem of how to
construct that state is now at hand. |
Sony Corporation Global Headquarters
Vol.54