When the Supercomputer Saves the World

Did you know that Japan has the world-fastest supercomputer as of now?  – I didn’t.  When I saw Conor’s tweet about “Fugaku” a couple of weeks ago, I felt a little bit ashamed of not having known this great news as a Japanese. (I bet every international student has had such an embarrassing moment, though.) This has brought me the subject of the blog post in which I want to do some research. I’m not a technical person, however. After touching on the basic information of the supercomputer, I will deliver more general learnings and insights for social impacts.

What is Supercomputer and Fugaku?

A supercomputer is a computer with greatest speed and memory. It is mostly used to solve problems in the fields of scientific computing and engineering, and usually, the problems they are targeting require billions or trillions of calculations. U.S. develops supercomputers called “Summit,” “Sierra” and China holds “Sunway TaihuLight,” “Tianhe-2” while Japan has “Fugaku(1). I imagined a supercomputer would be one gigantic computer, but it’s not. For instance, Fugaku is composed of 432 racks with each having a size of 220 cm height and 1.73 tons weight. The speed is 442 quadrillion calculations per second, or 442,000,000,000,000,000, which is equal to the speed that the whole world population would take 714 days to calculate (2). Someone would say, “OK cool, but my iPhone looks more capable.” Yet this is a good reminder that machines are far more efficient for some specialized purposes whereas humans use tools in a more flexible way. As compared to general purpose computers, supercomputers are supposed to deal with a broad range of complex challenges that require enormous data sets and massive calculations.

Fugaku is another name of Mt.Fuji, which is the highest mountain with elegant broad base in Japan, suggesting the highest performance and wide range of application.

Figure: Supercomputer Fugaku and Mt.Fuji

Learning from Failures

It was not built from scratch, but there was a previous model called “Computer K,” also one of the highest performance computers years ago. Japanese organizations are familiar with the concept of continuous improvement – or “Kaizen,” which is famous in the quality management area. Here in this context, I want to point out that Kaizen does not mean the process method per se, but the attitude of growth mindset, which we’ve found essential in the digital era. Further, Computer K’s failure is somewhat relevant to every businessperson. The officer sincerely reflected that they were too obsessed with pursuing the speed to make it more versatile. What resonates with me most in Prof. Kane’s book is the insight that “leaders often become so focused on the technological aspects that they forget why they are engaged in these efforts in the first place.” The fastest computer. Great. So what? This may sound obvious, but technology becomes valuable ONLY when it is used. Although I do not understand the technical problems, Computer K seemed to fall into the common but critical mistake. Now that Fugaku has been developed on this predecessor’s bitter experience, the main goal is not excelling in numerical benchmarks anymore.

Benchmark excellence is not our priority. Its success is assessed based on how much we can accelerate the applications that are important in society.

by the director of Riken’s Center for Computational Science (3).

Why needs supercomputers?

Thus, Fugaku has the “application-first philosophy,” meaning that its exclusive purpose is to dedicate its computational excellence to tackling some of the world’s biggest challenges, such as climate change.

The application with Fugaku will offer the following anticipated outcomes:

  • Acceleration of new drug development by performing high-speed, high-precision drug discovery simulation.
  • Early detection of diseases and advancement of preventive medicine through medical big data analysis and biological simulations.
  • Accurate prediction of tornadoes and heavy rain through weather big data analysis.
  • Simulation of earthquake, tsunami, and evacuation routes at the meter-level.
  • Accelerating the creation of new devices and materials to for next-generation industries.
  • Dramatic reductions in development time and costs by partially replacing the actual testing of airplanes and automobiles.
  • Increased insight into fundamental questions of science, such as when and how matter was created in the universe.

In fact, the system launched early in order to combat a disaster threatening the world: COVID-19. There are various projects from exploring new drug candidates to analyzing economic impacts (4). It is also tasked with another disaster that regularly threatens the country: tsunamis. It is innovative that a regular PC could function once Fugaku has trained the AI model, which represents the embedded belief in wider applications (5).

Fig. 1 Overview of tsunami prediction with AI
Figure: Overview of Tsunami Prediction with AI

Dark side

As we’ve seen in the class, technology is always a double-edged sword. As for supercomputers, they are said to have been used for nuclear research as well. Weapon designers could benefit from better simulation tools and computers which can run highly detailed calculations. They say actual experiments on the ground are not required since supercomputers with big data enable complex simulations. In 2015 for example, Intel and Nvidia were blocked from selling chips to Chinese supercomputer centers because the U.S. government had severe concerns about the possible consequences (6). I will not judge its validity or any political issues, but this reminds me of Albert Einstein’s famous E=mc2 and the Manhattan Project (7). Having through the atomic bombings of Hiroshima and Nagasaki, Japan can and should ring an alarming bell on the vicious use of technology from its unique standpoint. It’s all up to humans whether to make or break the blessing of technology.

Sources

The supercomputer Fugaku, jointly developed by RIKEN and Fujitsu: https://www.fujitsu.com/global/about/innovation/fugaku/

1. https://sciencenode.org/feature/the-5-fastest-supercomputers-in-the-world.php

2. https://www.youtube.com/watch?v=f3cx4PGDGmg

3. https://www.japantimes.co.jp/news/2021/01/07/business/tech/japans-fugaku-supercomputer/

4. https://blog.global.fujitsu.com/fgb/2020-12-02/combating-covid-19-with-supercomputer-fugaku-part-1/

5. https://www.fujitsu.com/global/about/resources/news/press-releases/2021/0216-01.html

https://www.techtimes.com/articles/257167/20210218/ai-predict-tsunami-real-time-through-worlds-fastest-supercomputer.htm

6. https://www.bbc.com/news/technology-32247532

7. https://www.businessinsider.com/albert-einstein-wrote-letter-us-roosvelt-atomic-bomb-2019-8

6 comments

  1. Nice post. One of the things thats striking about supercomputers is the ability to open up new fields of research with the increased computing power. It’s hard for me to imagine that there are still things we can’t do with current mainstream computing, but there are.

  2. shaneriley88 · ·

    Fascinating post, Sayo. I ran into some of the fantastic things supercomputers are doing while researching digital twins. It’s great to see how Fujitsu viewed Fugaku and its wide range of capabilities. It would be great if a place like the United Nations High Commissioner for Refugees or other agencies could use similar machines to help combat some of the effects of climate change, i.e., mass migration, drought, food shortages, etc.

  3. williammooremba · ·

    Great post, Sayo. I ended up digging a bit more into Fugaku and for me another exciting thing is the Fujitsu A64fx processor used in its production. This processor can now be used in the creation of more supercomputers. I personally am interested both in Fugaku’s impact on solving societal problems, but also platforms that use the underlying technology to do the same. As an example, Stony Brook University has a testbed called Ookami supported by the National Science Foundation. This provides an access point for US researchers using the A64fx. Definitely a very exciting area from both a business, and technology perspective.

    Link for Ookami: https://www.stonybrook.edu/ookami/

  4. Jie Zhao · ·

    Great post, Sayo! That’s crazy how many machines make up this supercomputer! As you said, to me my iPhone is already so powerful but I can only imagine when it comes to engineering and science, it requires that much more processing power. It is also scary to think about its powerful capabilities that can be used for malicious reasons, I wonder has there been even more regulations globally imposed to prevent from one government building a “too powerful” machine.

  5. AndraeAllen · ·

    Super computers are a fascinating topic. I appreciated how the engineers of Fugaku shifted away from benchmarks and more towards excelling in practical application. This shift reminds me of when Apple released the M1 processor but didn’t release any hardcore benchmarking specifications. After reading your post, I am curious about two more things. Q1: Was this computer designed for research or profit? Q2:How fast can this beast mine some crypto?

  6. Chuyong Liu · ·

    Wow, Sayo you are not alone with that embarrassing moment, I did not know China has supercomputers such as “Sunway TaihuLight” and “Tianhe-2” until I read this blog! The video showing the supercomputer reminded me that as a kid I have seen such a supercomputer in a smaller way in the place my mother works, which is the Industrial and Commercial Bank of China (the largest bank in China). They called it the “data center” and also used tofu interconnect. I used to think that is what a supercomputer should look like. After reading your blog, now I know that the CPU is more important and the hard part to develop.

    I also really love that you mentioned the fact that building the supercomputer with an aim of solving certain social problems. I found many firms lost in the process of digital transformation to keep finding the cutting-edge technology that looks cool but may not really contribute to solving strategic problems.

%d bloggers like this: