Matsuoka Satoshi oversees the operations of Fugaku, a Japanese supercomputer named for Mt. Fuji. It made global headlines when it succeeded in visualizing the aerosol dispersal of COVID-19.
"Direct Talk"
Our guest today is Matsuoka Satoshi.
He oversees operations of the supercomputer Fugaku.
Fugaku is currently the world's fastest supercomputer in four major rankings,
including things like calculation speed and big data analysis.
Fugaku was also used to study the transmission of COVID-19,
running detailed simulations of how droplets move
as they are expelled from the nose or mouth.
This research helped guide COVID policies around the world.
We spoke with Matsuoka Satoshi about the development of Fugaku,
and the incredible potential of supercomputing.
We named the supercomputer Fugaku after Mount Fuji,
which of course is our national symbol of Japan.
It's the tallest mountain in Japan.
Mount Fuji has a very broad base because it's a stand-alone mountain.
And the shape of Mount Fuji basically signifies our idealism
towards what supercomputers should be.
It should have a very high peak,
a very high performance.
But at the same time, it should have a very, very wide broad base.
Kobe, the capital of Hyogo Prefecture.
The city is home to the RIKEN Center for Computational Science.
Matsuoka is the center's director.
Inside, we find Fugaku,
the world's fastest supercomputer.
Composed of 432 high-performance computer racks linked together,
it's capable of performing massive computational tasks in an instant.
Fugaku is so fast.
How many smartphones does it take to match Fugaku?
And the answer is, it's equivalent to about 20 million smartphones.
So 20 million smartphones is about the number of smartphones
that are sold in Japan over the course of a year.
So that's how I would say large or how fast Fugaku can be.
Fugaku can also run simulations of real-world phenomena,
another category in which it is top in the world.
Matsuoka was involved in studying the transmission of COVID-19 through droplets
which can carry the virus.
Precise simulations of how droplets spread
were run by Fugaku and made public.
One thing Fugaku revealed was the importance of masking.
Here is a comparison of how droplets spread
while not wearing a mask versus when wearing a mask.
Their simulations found that, without a mask,
droplets travel more than two meters.
Wearing a mask reduces the amount of escaping droplets
by about two thirds.
Fugaku's findings changed workplace arrangements.
If you cough at a table across from someone,
the spread of droplets puts the person at significant risk.
Partitions make a big difference.
If a partition is 140 centimeters tall,
enough to block the head,
people in the surroundings will mostly avoid the droplets,
and infection can be prevented.
Basically, aerosols are invisible.
We were able to visualize it using computer graphics.
You're sitting across but, you know... This is fairly complicated stuff.
Sandwiched between, and you have different sizes of droplets.
The heavy ones would drop immediately,
while the light-weight ones would become aerosols and kind of stay in the air for hours.
And this was broadcast all over Japan,
and in fact, abroad.
And so for the first time in history,
people were able to visualize what exactly happens
when somebody would cough.
And then, you know, really feel the threat
and the need to maintain social distance,
the need to wear masks and so forth.
Matsuoka and his fellow researchers won a special Gordon Bell Prize
for their work with Fugaku on the transmission of COVID-19.
The Gordon Bell Prize is one of the most prestigious awards
in supercomputing given out each year.
It's synonymous to like the Academy Awards in the movies.
Fortunately, we won.
I was one of the authors by the way.
The committee and the whole field recognized
that the impact of our simulations on COVID-19
had impacted the society in a rather global fashion to,
at least early on in the pandemic,
basically overturn some of the naysayers' suspicion.
Things like masking or ventilation and so forth. The basics.
There were a lot of doubts early on as to their effectiveness,
and we had proven that they are really effective
under many societal situations.
And that really helped
to save lives, which of course was probably the most important thing.
Matsuoka Satoshi began studying computer science
on his own in junior high school.
His passion for programming continued through his high school and college days.
During that time, he met another young programmer named Iwata Satoru.
He was a very bright, young, handsome person,
and we got to be very good friends.
And later on, he became the CEO of Nintendo.
He was also a pretty good programmer.
Matsuoka and Iwata were part of a group that founded a video game company.
One of their titles was "Pinball,"
which they released on Nintendo's Family Computer console.
That was one of the programs I also co-authored with Satoru.
So I wrote most of the core,
the so called physics part the game, the main part of the game.
And Satoru basically wrote some of the more peripheral parts of the game
like some of the graphics stuff.
I think it sold something like two million copies or something.
So that was a very interesting piece of history.
Iwata continued to develop games,
but Matsuoka was torn.
Should he keep making games,
or should he try and enter the field of academic research?
I thought I would go to a company to do development
because I enjoy doing development.
But then Iwata said, "I think you should go to research
because I think you're pretty good there."
Matsuoka followed Iwata's advice and pursued a career in research.
Today, he oversees the operations of the world's fastest supercomputer.
Iwata blazed his own trail of success,
becoming president of Nintendo
and overseeing the creation of numerous hit games.
But in 2015, he passed away at the age of 55.
It's a life lesson saying that you can never truly judge yourself.
It's always a third party that has that right vision of what you really are.
So never over or underestimate your own capabilities.
I'm really thankful.
In 2001, at age 38,
Matsuoka became a professor at the Tokyo Institute of Technology
and he forged ahead with his research.
Matsuoka started creating a supercomputer,
one that could be used by scientists, companies, and students alike,
a "supercomputer for everyone."
In 2006, he completed the Tsubame series supercomputer,
which emphasized usability and versatility.
When it began operations,
it was the fastest supercomputer in Japan.
Around the same time,
development began on a supercomputer called "K,"
a national project led by RIKEN
with a budget exceeding 100 billion yen.
But then, in 2009,
Japan elected a new ruling party
and they undertook a review of the national budget.
One lawmaker from the party criticized
the huge cost of creating the world's fastest supercomputer.
Why does it need to be the world's best?
Is second place so bad?
The whole community felt that we had to basically sustain it.
Otherwise, if we canceled it, the whole Japanese supercomputing industry
would be dead, dead in the water.
They weathered the criticism, and K was completed in 2012.
For a time, it was considered the world's fastest supercomputer.
But it lost its number-one ranking almost immediately.
And so much effort had gone into computational speed
that K was not user-friendly.
Adoption was limited.
In 2014,
a project led by RIKEN started work on a supercomputer
that would be a successor to K.
In 2018, Matsuoka became the director
of the RIKEN Center for Computational Science.
He was tapped for his widely respected work on Tsubame.
We had a technical challenge.
We also had challenges to make the machine
easy to use and also very broad-based.
Matsuoka returned to the concept he had for Tsubame,
a user-friendly "supercomputer for everyone."
His goal was to put applications first.
And so, software developers
took a major role alongside the hardware developers.
The goal was to develop a supercomputer
that had good compatibility with a wide range of software.
There were various types of meetings,
both private and public, and also involving the vendors and ministry.
So, many, many meetings.
So there must have been hundreds of meetings.
So it was really important for us,
Fugaku's idealism of sporting a broad class of applications
that we would be number-one across the board and not just on one benchmark,
or at least do very well across the board.
A supercomputer that was to be exceptional across the board.
Fugaku.
In 2020, it began operations.
In June of that year,
Fugaku placed first in all four major supercomputer rankings
which measure computational speed,
real-world applications,
artificial intelligence applications,
and graph analytic performance.
Fugaku has held the top spot in all four categories
for four straight judging periods,
an unprecedented feat.
Having a number-one machine from a country perspective,
at least having the ability to produce these number-one machines
really makes people proud.
It' like winning a gold medal in the Olympics.
Because supercomputing is an extremely challenging field.
And countries and even companies are spending
millions, hundreds of millions, or even billions of dollars to compete.
So to win there, you know...
We're under-budgeted compared to other countries, but we still win.
And that's, you know... That's a proud moment.
The highly versatile Fugaku has found uses in a wide range of fields.
Searches for specific compounds within existing drugs
that are effective against COVID-19.
Simulations of aerodynamic and fluid-dynamic phenomena
that can be run before building new planes,
improving aircraft design and safety.
And apps that can predict and observe torrential rains
in a way that wasn't possible before.
Fugaku truly is a "supercomputer for everyone."
Right now, vast sums of money are being spent
in the US, China and Europe to create supercomputers
that can catch up to and surpass Fugaku.
The competition is fierce.
Now there's a recognition that
this big computing is really important for things like AI,
for big data processing, for metaverse and everything
related to advanced IT.
These big computing powers are an absolute necessity
for doing auto driving and things like that.
So there's a recognition that advancing this technology
will give you an inherent advantage
in the country or the industry and so forth.
So that's why there's big investment.
(Do you have any words to live by?)
With great... Well, it's a variation of,
"With great power, comes great responsibility."
With great supercomputing power comes great responsibilities.
Of course, as a scientist/engineer,
we have the ambition to advance computing.
But we also have to be cognizant of the implications
of what this great power brings to the world
and, you know, what kind of societal problems we can solve,
and these kinds of power not being put to bad use.
So it really gives us a metric
as to how we would behave
in terms of having this great power
which had never been experienced before by mankind.