Accelerating Discovery: The Future of Scientific Computing at the Department of Energy (Hearing, 2021)
Summary
Transcript of the hearing "Accelerating Discovery: The Future of Scientific Computing at the Department of Energy" before the Subcommittee on Energy of the Committee on Science, Space, and Technology, May 19, 2021, Serial No. 117-16, with Chairman Jamaal Bowman presiding virtually. Mr. Bowman said the hearing would examine the status and needs of DOE's scientific computing programs and the research, development and workforce training they require. Opening statements also came from Randy Weber, Eddie Bernice Johnson and Frank Lucas. Witnesses were J. Stephen Binkley of the DOE Office of Science, Georgia Tourassi of Oak Ridge National Laboratory, Karen Willcox of The University of Texas at Austin, Christopher Monroe of IonQ, Inc. and Seny Kamara of Brown University.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Accelerating Discovery: The Future of Scientific Computing at the Department of Energy (Hearing, 2021).”
Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/house-hearing-117th-congress-accelerating-discovery-the-future-of-scie-acba7fe762/
, SHA-256 cf579533ced82392cd086db9a24a8e2983b490d515d9eb40c559beb616d725ab.Full text
[House Hearing, 117 Congress]
[From the U.S. Government Publishing Office]
ACCELERATING DISCOVERY:
THE FUTURE OF SCIENTIFIC COMPUTING
AT THE DEPARTMENT OF ENERGY
=======================================================================
HEARING
BEFORE THE
SUBCOMMITTEE ON ENERGY
OF THE
COMMITTEE ON SCIENCE, SPACE,
AND TECHNOLOGY
HOUSE OF REPRESENTATIVES
ONE HUNDRED SEVENTEENTH CONGRESS
FIRST SESSION
__________
MAY 19, 2021
__________
Serial No. 117-16
__________
Printed for the use of the Committee on Science, Space, and Technology
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
Available via the World Wide Web: http://science.house.gov
__________
U.S. GOVERNMENT PUBLISHING OFFICE
44-540PDF WASHINGTON : 2021
-----------------------------------------------------------------------------------
COMMITTEE ON SCIENCE, SPACE, AND TECHNOLOGY
HON. EDDIE BERNICE JOHNSON, Texas, Chairwoman
ZOE LOFGREN, California FRANK LUCAS, Oklahoma,
SUZANNE BONAMICI, Oregon Ranking Member
AMI BERA, California MO BROOKS, Alabama
HALEY STEVENS, Michigan, BILL POSEY, Florida
Vice Chair RANDY WEBER, Texas
MIKIE SHERRILL, New Jersey BRIAN BABIN, Texas
JAMAAL BOWMAN, New York ANTHONY GONZALEZ, Ohio
BRAD SHERMAN, California MICHAEL WALTZ, Florida
ED PERLMUTTER, Colorado JAMES R. BAIRD, Indiana
JERRY McNERNEY, California PETE SESSIONS, Texas
PAUL TONKO, New York DANIEL WEBSTER, Florida
BILL FOSTER, Illinois MIKE GARCIA, California
DONALD NORCROSS, New Jersey STEPHANIE I. BICE, Oklahoma
DON BEYER, Virginia YOUNG KIM, California
CHARLIE CRIST, Florida RANDY FEENSTRA, Iowa
SEAN CASTEN, Illinois JAKE LaTURNER, Kansas
CONOR LAMB, Pennsylvania CARLOS A. GIMENEZ, Florida
DEBORAH ROSS, North Carolina JAY OBERNOLTE, California
GWEN MOORE, Wisconsin PETER MEIJER, Michigan
DAN KILDEE, Michigan VACANCY
SUSAN WILD, Pennsylvania
LIZZIE FLETCHER, Texas
VACANCY
------
Subcommittee on Energy
HON. JAMAAL BOWMAN, New York, Chairman
SUZANNE BONAMICI, Oregon RANDY WEBER, Texas,
HALEY STEVENS, Michigan Ranking Member
JERRY McNERNEY, California JIM BAIRD, Indiana
DONALD NORCROSS, New Jersey MIKE GARCIA, California
SEAN CASTEN, Illinois RANDY FEENSTRA, Iowa
CONOR LAMB, Pennsylvania CARLOS A. GIMENEZ, Florida
DEBORAH ROSS, North Carolina PETER MEIJER, Michigan
C O N T E N T S
May 19, 2021
Page
Hearing Charter.................................................. 2
Opening Statements
Statement by Representative Jamaal Bowman, Chairman, Subcommittee
on Energy, Committee on Science, Space, and Technology, U.S.
House of Representatives....................................... 9
Written Statement............................................ 10
Statement by Representative Randy Weber, Ranking Member,
Subcommittee on Energy, Committee on Science, Space, and
Technology, U.S. House of Representatives...................... 11
Written Statement............................................ 12
Statement by Representative Eddie Bernice Johnson, Chairwoman,
Committee on Science, Space, and Technology, U.S. House of
Representatives................................................ 13
Written Statement............................................ 14
Statement by Representative Frank Lucas, Ranking Member,
Committee on Science, Space, and Technology, U.S. House of
Representatives................................................ 14
Written Statement............................................ 16
Witnesses:
Dr. J. Stephen Binkley, Principal Deputy Director, Office of
Science at the Department of Energy
Oral Statement............................................... 18
Written Statement............................................ 20
Dr. Georgia Tourassi, Director, National Center for Computational
Sciences at Oak Ridge National Laboratory
Oral Statement............................................... 29
Written Statement............................................ 31
Dr. Karen Willcox, Director, Oden Institute for Computational
Engineering and Sciences at The University of Texas at Austin
Oral Statement............................................... 44
Written Statement............................................ 46
Dr. Christopher Monroe, Co-Founder and Chief Scientist, IonQ,
Inc.
Oral Statement............................................... 56
Written Statement............................................ 59
Dr. Seny Kamara, Associate Professor, Brown University
Oral Statement............................................... 63
Written Statement............................................ 65
Discussion....................................................... 70
Appendix I: Answers to Post-Hearing Questions
Dr. J. Stephen Binkley, Principal Deputy Director, Office of
Science at the Department of Energy............................ 94
Dr. Georgia Tourassi, Director, National Center for Computational
Sciences at Oak Ridge National Laboratory...................... 100
Dr. Karen Willcox, Director, Oden Institute for Computational
Engineering and Sciences at The University of Texas at Austin.. 101
Appendix II: Additional Material for the Record
Letter submitted by Representative Frank Lucas, Ranking Member,
Committee on Science, Space, and Technology, U.S. House of
Representatives................................................ 108
ACCELERATING DISCOVERY:
THE FUTURE OF SCIENTIFIC COMPUTING
AT THE DEPARTMENT OF ENERGY
----------
WEDNESDAY, MAY 19, 2021
House of Representatives,
Subcommittee on Energy,
Committee on Science, Space, and Technology,
Washington, D.C.
The Subcommittee met, pursuant to notice, at 11:03 a.m.,
via Zoom, Hon. Jamaal Bowman [Chairman of the Subcommittee]
presiding.
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
Chairman Bowman. This hearing will now come to order.
Without objection, the Chairman is authorized to declare a
recess at any time.
Before I deliver my opening remarks, I wanted to note
that, today, the Committee is meeting virtually. I want to
announce a couple of reminders to the Members about the conduct
of this hearing. First, Members should keep their video feed on
as long as they are present in the hearing. Members are
responsible for their own microphones. Please also keep your
microphones muted unless you are speaking. Finally, if Members
have documents they want to submit for the record, please email
them to the Committee Clerk, whose email address was circulated
prior to the hearing.
Good morning, and thank you to all of our witnesses who
are joining us virtually to discuss the importance of
scientific computing at the Department of Energy (DOE). This
hearing is one of a series on research and development (R&D)
activities sponsored by the Department of Energy's Office of
Science. Today, we will be examining the current status and
needs of DOE's scientific computing programs, as well as the
research, development, and workforce training required to
ensure that DOE and the Nation maintains its leadership in this
crucial area.
Stewardship of DOE's scientific computing ecosystem is led
by the Office of Science's Advanced Scientific Computing
Research program, or ASCR. ASCR is also DOE's main sponsor of
research in foundational areas such as applied mathematics and
computer science. This year, ASCR was funded at just over $1
billion, about 1/7 of the total Office of Science budget.
DOE possesses some of the most powerful supercomputers in
existence. It will deploy the Nation's first exascale system
this year, signaling an exciting new era in the field of
scientific computing. Housed at several national laboratories,
DOE's supercomputers help researchers analyze huge data sets
and test complex computational models, greatly accelerating the
pace of discovery in the design of life-saving medical
treatments, advanced manufacturing, and the prediction of
climate systems, among many other fields of research. DOE's
supercomputing ecosystem serves as a critical resource for
academic and industry users from the U.S. and around the world.
I am looking forward to discussing with our witnesses the real-
world applications of these incredible systems, and how
Congress can ensure that they are continuously maintained and
improved.
It is also critically important for DOE to support
research that will lay the groundwork for future computing
capabilities. We are fast approaching the point at which the
computing architectures we have relied upon for decades will
reach their physical and economic limitations. Therefore, ASCR
must continue to invest in the applied mathematics, computer
science, and the game-changing technology development
activities that will enable powerful new paradigms like quantum
computing.
As we craft a forward-looking Office of Science
authorization bill, I will be looking to our witnesses for
insights into how we in Congress can ensure that these
activities are robustly supported. As we will explore today,
scientific computing holds tremendous promise for accelerating
scientific discovery. But we need to use these capabilities
responsibly, ethically, and to advance the public good. For
example, as computing and artificial intelligence become more
powerful, we must ensure that algorithms are designed to
protect people's privacy and eradicate bias. We must also stop
these tools from fortifying the structures of systematic
racism, as we have seen happen with things like predictive
policing and facial recognition technology. This will only
become more important as DOE's supercomputing capabilities are
used to process, analyze, and store sensitive information, such
as biomedical datasets.
Let's also discuss how to retain a strong role for the
public sector here, to fully tap into computing's potential to
help solve humanity's most pressing problems, from curing
diseases to addressing the climate emergency. And let's involve
the public, especially marginalized communities, in shaping the
development and aims of new technologies like these so that all
can share in the benefits equally. As you will hear from our
witnesses today, we need to pursue an agenda of scientific
computing for the people.
Finally, as I have said before, research and
infrastructure funding represent just one piece of the puzzle.
We need a skilled and diverse workforce to maintain the
vitality of DOE's scientific computing ecosystem long into the
future. I am particularly interested in leveraging programs
such as the Computational Science Graduate Fellowship to forge
closer connections between the Department and minority-serving
institutions. We can all agree on the need for greater
diversity, equity, and inclusion across our research
enterprise.
I want to again thank our excellent panel of witnesses
assembled today, and I look forward to hearing your testimony.
[The prepared statement of Chairman Bowman follows:]
Good morning, and thank you to all of our witnesses who are
joining us virtually today to discuss the importance of
scientific computing at the Department of Energy.
This hearing is one of a series on research and development
activities sponsored by the DOE's Office of Science. Today, we
will be examining the current status and needs of DOE's
scientific computing programs as well as the research,
development, and workforce training required to ensure that
DOE, and the nation, maintains its leadership in this crucial
area.
Stewardship of DOE's scientific computing ecosystem is led
by the Office of Science's Advanced Scientific Computing
Research program, or ASCR. ASCR is also DOE's main sponsor of
research in foundational areas such as applied mathematics and
computer science. This year, ASCR was funded at just over a
billion dollars, about one-seventh of the total Office of
Science budget.
DOE possesses some of the most powerful supercomputers in
existence. It will deploy the nation's first exascale system
this year, signaling an exciting new era in the field of
scientific computing. Housed at several national laboratories,
DOE's supercomputers help researchers analyze huge data sets
and test complex computational models, greatly accelerating the
pace of discovery in the design of life-saving medical
treatments, advanced manufacturing, and the prediction of
climate systems, among many other fields of research. DOE's
supercomputing ecosystem serves as a critical resource for
academic and industry users from the U.S. and around the world.
I am looking forward to discussing with our witnesses the real-
world applications of these incredible systems, and how
Congress can ensure that they are continuously maintained and
improved.
It is also critically important for DOE to support research
that will lay the groundwork for future computing capabilities.
We are fast approaching the point at which the computing
architectures we have relied upon for decades will reach their
physical and economic limitations. Therefore, ASCR must
continue to invest in the applied mathematics, computer
science, and the game-changing technology development
activities that will enable powerful new paradigms like quantum
computing. As we craft a forward-looking Office of Science
authorization bill, I will be looking to our witnesses for
insights into how we in Congress can ensure that these
activities are robustly supported.
As we will explore today, scientific computing holds
tremendous promise for accelerating scientific discovery. But
we need to use these capabilities responsibly, ethically, and
to advance the public good. For example, as computing and
artificial intelligence become more powerful, we must ensure
that algorithms are designed to protect people's privacy and
eradicate bias. We must also stop these tools from fortifying
the structures of systemic racism, as we have seen happen with
things like predictive policing and facial recognition
technology. This will only become more important as DOE's
supercomputing capabilities are used to process, analyze, and
store sensitive information, such as biomedical datasets.
Let's also discuss how to retain a strong role for the
public sector here, to fully tap into computing's potential to
help solve humanity's most pressing problems - from curing
diseases to addressing the climate emergency. And let's involve
the public, especially marginalized communities, in shaping the
development and aims of new technologies like these - so that
all can share in the benefits equally. As you will hear from
one our witnesses today, we need to pursue an agenda of
scientific computing for the people.
Finally, as I have said before, research and infrastructure
funding represent just one piece of the puzzle. We need a
skilled and diverse workforce to maintain the vitality of DOE's
scientific computing ecosystem long into the future. I am
particularly interested in leveraging programs such as the
Computational Science Graduate Fellowship to forge closer
connections between the Department and Minority-Serving
Institutions. We can all agree on the need for greater
diversity, equity, and inclusion across our research
enterprise.
I want to again thank our excellent panel of witnesses
assembled today, and I look forward to hearing your testimony.
With that, I yield back.
Chairman Bowman. Finally, I want to note that it is a busy
day on the Hill, and I may have to step out briefly to ask
questions on another Committee.
With that, I now recognize Mr. Weber for an opening
statement.
Mr. Weber. Well, thank you, Mr. Chair, and I'll be glad
to, you know, conduct the hearing while you're gone.
I do want to thank you for hosting this hearing and to our
esteemed witness panel for being here this afternoon or
technically, I guess, this morning. I'm excited to hear about
the critical advanced scientific computing research and
development activities being carried out through the DOE,
Department of Energy's Office of Science.
The Advanced Scientific Computing Research program, ASCR
as you referred to it, is one that enjoys bipartisan support as
a priority within the Office of Science. For the past 30 years,
researchers within this program have led advances in
mathematics and computing that form the foundation for those
complex models and simulations. These developments, in turn,
have translated to increased knowledge and understanding of
everything from bioenergy and climate change to Alzheimer's
disease and health models.
Today, ASCR hosts some of the world's most powerful
supercomputers and a high-speed network that moves enormous
volumes of scientific data at light speed. In the rapidly
evolving fields of quantum computing and artificial
intelligence, ASCR is dedicated to maintaining U.S.
competitiveness and leadership. The program also supports DOE's
goal of completing the world's first exascale computing system
this year and a second system within the next year. As our
competitors race to develop exascale systems on their own,
DOE's strong support of advanced computing research within ASCR
is essential to maintaining U.S. leadership in this field.
And it's more than just hardware that needs additional
focus, I might add. We need significant modifications to
today's tools and techniques to deliver on the promise of high-
performance computing. Researchers are in need of a new suite
of software tools, programming models, and applications to
enable effective use of exascale systems. Without software and
application R&D, we will simply have high-powered machines
collecting dust.
Additionally, in order to fully and effectively support
innovation in next-generation science, DOE must also support
and also encourage cross-cutting research initiatives within
the Department, as well as other Federal agencies. Within the
Office of Science alone, ASCR resources and capabilities can be
used to drive innovation in computational chemistry and
nanomaterials for energy applications, improve simulations of
fusion energy reactors, and enhance our ability to predict
changes in the global climate with next-generation Earth system
models.
Other Federal agencies could also capitalize on these
unique, world-leading resources. As authorized by the Energy
Act of 2020, the Department of Veterans Affairs (VA) is
partnering with DOE to use high-performance computing in
analyzing massive amounts of health data. This data analysis
will help the VA better understand diseases and improve
veterans' overall quality of life.
We should seek to build upon and expand partnerships like
this so that the entire Federal Government benefits from ASCR's
tools, as well as its technologies. At the end of the day,
we're all supporting one thing: U.S. leadership in science,
technology, and innovation. There is no Federal entity in a
better position to lead this change than DOE's Office of
Science.
That's why I am pleased that we are very close to
finalizing legislation that provides strong support and long-
term guidance for the Office of Science. We're making sure that
rubber meets the road, and that the U.S. research enterprise is
well-equipped with all available resources to successfully
overcome the generational challenges they face.
I want to again thank my colleagues for their bipartisan
outreach and collaboration. And I want to thank the witnesses
for offering and bringing their effort, their input.
Thank you, Mr. Chairman, and I yield back.
[The prepared statement of Mr. Weber follows:]
Thank you, Chairman Bowman, for hosting this hearing, and
thank you to our esteemed witness panel for being here this
afternoon. I am excited to hear about the critical advanced
scientific computing research and development activities being
carried out through the Department of Energy's (DOE) Office of
Science.
The Advanced Scientific Computing Research Program, or ASCR
program, is one that enjoys bipartisan support as a priority
within the Office of Science. For the past thirty years,
researchers within this program have led advances in
mathematics and computing that form the foundation for complex
models and simulations. These developments, in turn, have
translated to increased knowledge and understanding of
everything from bioenergy and climate change to Alzheimer's
disease and health models.
Today, ASCR hosts some of the world's most powerful
supercomputers and a high-speed network that moves enormous
volumes of scientific data at light speed. In the rapidly
evolving fields of quantum computing and artificial
intelligence, ASCR is dedicated to maintaining U.S.
competitiveness and leadership. The program also supports DOE's
goal of completing the world's first exascale computing system
this year and a second system within the next year.
As our competitors race to develop exascale systems of
their own, DOE's strong support of advanced computing research
within ASCR is essential to maintaining U.S. leadership in this
field. And it's more than just hardware that needs additional
focus.
We need significant modifications to today's tools and
techniques to deliver on the promise of high-performance
computing. Researchers are in need of a new suite of software
tools, programming models, and applications to enable effective
use of exascale systems. Without software and application R&D,
we will simply have high-powered machines collecting dust.
Additionally, in order to fully and effectively support
innovation in next- generation science, DOE must also encourage
cross-cutting research initiatives within the Department and
with other Federal agencies. Within the Office of Science
alone, ASCR resources and capabilities can be used to drive
innovation in computational chemistry and nanomaterials for
energy applications, improve simulations of fusion energy
reactors, and enhance our ability to predict changes in the
global climate with next generation Earth System Models.
Other federal agencies could also capitalize on these
unique, world-leading resources. As authorized by the Energy
Act of 2020, the Department of Veterans Affairs is partnering
with DOE to use high performance computing in analyzing massive
amounts of health data. This data analysis will help the VA
better understand diseases and improve veterans' overall
quality of life.
We should seek to build upon and expand partnerships like
this so that the entire federal government benefits from ASCR's
tools and technologies. At the end of the day, we are all
supporting one thing: U.S. leadership in science, technology,
and innovation. There is no federal entity in a better position
to lead this charge than DOE's Office of Science.
That is why I am pleased we are very close to finalizing
legislation that provides strong support and long-term guidance
for the Office of Science. We are making sure rubber meets the
road, and that the U.S. research enterprise is equipped with
all available resources to successfully overcome the
generational challenges they face.
I want to again thank my colleagues for their bipartisan
outreach and collaboration. And I want to thank the witnesses
for offering their input on our efforts. Thank you, Mr.
Chairman, and I yield back the balance of my time.
Chairman Bowman. Thank you so much, Mr. Weber.
The Chair now recognizes the Chairwoman of the Full
Committee, Ms. Johnson, for an opening statement.
Chairwoman Johnson. Thank you very much, Mr. Bowman, for
holding this hearing, and Ranking Member Weber. And I also want
to thank our witnesses for your participation, and I have
enjoyed reading your thoughtful written testimony.
The Department of Energy has long been a leader in
advancing new energy technologies, as well as the fundamental
and the foundational sciences of physics, chemistry,
engineering and math and computational science that support
energy innovation. High-performance computing, or
supercomputing, is one area the Department has led for decades,
and DOE shows no signs of slowing down. The Department
currently stewards two of the top three fastest supercomputers
in the world. And as we will learn more about from our
witnesses here today, the United States is on track to finish
building the first exascale computer in the world this year.
These systems serve as critical resources for academic and
industrial users and are a key component of our economic
competitiveness, scientific leadership, and national security.
In the past, high-performance computers were needed almost
solely for specialized scientific and engineering applications.
Now, as we enter the world where thousands of devices all
around us are generating millions of bytes of data every
minute, high-performance computers can be used to fundamentally
improve our quality of life. Public policies play a critical
role in supporting the advancement of these capabilities and
enabling our society and economy to directly benefit from them.
Additional Federal investments in high-performance computing
will enable the development of new industries, grow our
technology economy, and advance our technological leadership
internationally.
All that said, as we continue to support the development
and use of these breakthrough technologies, we almost--we must
also do everything we can to ensure that we are doing this in a
responsible and ethical manner even in the face of competition
from our adversaries.
I thank you again for being here, and I look forward to
this important discussion today, and I yield back.
[The prepared statement of Chairwoman Johnson follows:]
Thank you, Chairman Bowman, for holding this hearing, and I
also want to thank this excellent panel of witnesses for your
participation and thoughtful written testimony.
The Department of Energy has long been a leader in
advancing new energy technologies, as well as the foundational
sciences of physics, chemistry, engineering, mathematics, and
computational science that support energy innovation.
High performance computing, or supercomputing, is one area
the Department has led in for decades, and DOE shows no signs
of slowing down. The Department currently stewards two of the
top three fastest supercomputers in the world. And as we will
learn more about from our witnesses here today, the United
States is on track to finish building the first exascale
computer in the world this year.
These systems serve as critical resources for academic and
industrial users, and are a key component of our economic
competitiveness, scientific leadership, and national security.
In the past, high performance computers were needed almost
solely for specialized scientific and engineering applications.
Now, as we enter a world where thousands of devices all around
us are generating millions of bytes of data every minute, high
performance computing can be used to fundamentally improve our
quality of life.
Public policies play a critical role in supporting the
advancement of these capabilities, and in enabling our society
and economy to directly benefit from them. Additional federal
investments in high performance computing will enable the
development of new industries, grow our technology economy, and
advance our technological leadership internationally.
All that said, as we continue to support the development
and use of these breakthrough technologies, we must also do
everything we can to ensure that we are doing this in a
responsible and ethical manner. Even in the face of competition
from our adversaries.
Thank you all again for being here, and I look forward to
this important discussion today. With that I yield back.
Chairman Bowman. Thank you so much for your opening
statement, Madam Chairwoman.
The Chair now recognizes the Ranking Member of the Full
Committee, Mr. Lucas, for an opening statement.
Mr. Lucas. Thank you, Chairman Bowman, for hosting this
hearing, and thank you to all our witnesses for being with us
this afternoon.
Earlier this month, the Energy Subcommittee held a hearing
on the Department of Energy's Office of Science, which
emphasized the essential role of DOE in our Federal research
enterprise and highlighted our shared support of these
programs.
Today, we have an opportunity to examine the activities of
another Office of Science program in Advanced Scientific
Computing Research, or ASCR. Advanced computing research and
infrastructure is the backbone of scientific discovery, not
just at the Department of Energy but at U.S. research
institutions nationwide. Through the ASCR program, DOE supports
the development of tools and technologies in high-performance
computing, applied mathematics, advanced networks, data
analytics, and next-generation computing initiatives. It also
hosts some of the most advanced computing resources in the
world at its national laboratories.
There is a great potential for Federal agencies and U.S.
industry partners to leverage ASCR's unique computing
resources. With adequate support, DOE's program will
revolutionize our relationship with advanced technology and our
capacity for scientific progress. This work is vital to our
clean energy economy, our national security, and our leadership
in science and technology.
Yet we know that our international competitors like China
are outpacing us in basic research investment and are closing
the gap in key computing focus areas like artificial
intelligence and quantum sciences. Expanding our capacities in
these fields requires a strategic effort with strong Federal
investment and active public-private partnerships.
That's why in this Congress I've introduced legislation to
address those challenges. My bill, the Securing American
Leadership in Science and Technology Act, SALSTA, roughly
doubles funding for ASCR over the next 10 years.
Another bill I introduced, the Quantum User Expansion for
Science and Technology Act, QUEST, establishes a program at the
Department of Energy to expand public-private partnerships for
quantum resource use and encourage greater participation in the
development of quantum information sciences (QIS).
Mr. Chairman, at this time, I'd like to ask unanimous
consent to submit for the record a letter from the Quantum
Industry Coalition on the need to maximize the value of the
U.S. quantum industry and the role that DOE and its national
laboratories can play in this high-priority work.
Chairman Bowman. Without objection.
Mr. Lucas. Thank you, Mr. Chairman.
I'm also proud to join my colleague and the Ranking Member
of the Investigations and Oversight Subcommittee, Jay
Obernolte, on a bill to strengthen the other high-priority
computing research program carried out at the Department. This
week, Representative Obernolte introduced the Next Generation
Computing Research and Development Act, which authorizes
various DOE advanced scientific computer programs. These will
support beyond excellent energy computing, computing workforce
development, and applied mathematics and software development
activities. This bill, along with the QUEST Act and SALSTA, is
an important step to move forward in improving our Nation's
global standing in science and technology.
We know that maintaining U.S. leadership will require a
shared commitment to prioritize DOE and its Office of Science,
and nowhere is this clearer than in the advanced computing
space. The United States relies on computing capacities that
only the Department of Energy can provide. We know that the
Nation who takes the lead in advanced computing will set the
stage for the next generation of technologies and technology
standards. We cannot afford to fall behind in this race.
Last week, I was encouraged by the progress made by our
friends in the Senate to recognize the important role the
Department of Energy plays in advancing U.S. innovation. But
DOE and the national labs shouldn't be an afterthought when we
consider the U.S. research enterprise. They're integral to our
scientific progress. That's why Chairman--woman Johnson and I
have been working on bipartisan Office of Science legislation
that will make a strong commitment to the Department of Energy
and its work, including successful programs like ASCR.
This legislation to support research at the Department of
Energy will go hand-in-glove with the NSF (National Science
Foundation) For the Future Act, which supports basic research,
STEM education, and technology transfer at the National Science
Foundation. Together, these research bills will solidify the
long-term stability of our international leadership in science.
I once again want to thank our witnesses for being here
today. I look forward to a productive discussion. Thank you,
Chairman Bowman, and I yield back the balance of my time.
[The prepared statement of Mr. Lucas follows:]
Thank you, Chairman Bowman for hosting this hearing, and
thank you to all our witnesses for being with us this
afternoon.
Earlier this month, the Energy Subcommittee held a hearing
on the Department of Energy's Office of Science which
emphasized the essential role of DOE in our federal research
enterprise and highlighted our shared support of its programs.
Today, we have an opportunity to examine the activities of
another Office of Science program in Advanced Scientific
Computing Research, or ASCR. Advanced computing research and
infrastructure is the backbone of scientific discovery, not
just at the Department of Energy but at U.S. research
institutions nationwide. Through the ASCR (``Oscar'') program,
DOE supports the development of tools and technologies in high
performance computing, applied mathematics, advanced networks,
data analytics, and next-generation computing initiatives. It
also hosts some of the most advanced computing resources in the
world at its national laboratories.
There is great potential for federal agencies and U.S.
industry partners to leverage ASCR's unique computing
resources. With adequate support, DOE's program will
revolutionize our relationship with advanced technology and our
capacity for scientific progress. This work is vital to our
clean energy economy, our national security, and our leadership
in science and technology.
Yet we know that our international competitors like China
are outpacing us in basic research investment and are closing
the gap in key computing focus areas like artificial
intelligence and quantum sciences. Expanding our capacities in
these fields requires a strategic effort with strong federal
investment and active public-private partnerships.
That's why, this Congress, I've introduced legislation to
address these challenges. My bill, the Securing American
Leadership in Science and Technology Act (SALSTA), roughly
doubles funding for ASCR over ten years. Another bill I
introduced, the Quantum User Expansion for Science and
Technology Act (QUEST) Act, establishes a program at the
Department of Energy to expand public-private partnerships for
quantum resource use and encourage greater participation in the
development of quantum information sciences.
Mr. Chairman, at this time I'd like to ask unanimous
consent to submit for the record, a letter from the Quantum
Industry Coalition, on the need to maximize the value of the
U.S. quantum industry, and the role that DOE and its
national laboratories can play in this high-priority work.
I'm also proud to join my colleague and Ranking Member of
the Investigations and Oversight Subcommittee, Jay Obernolte,
on a bill to strengthen other high-priority computing research
carried out by the Department.
This week, Representative Obernolte introduced the Next
Generation Computing Research and Development Act, which
authorizes various DOE advanced scientific computing programs.
These will support beyond-exascale and energy efficient
computing, computing workforce development, and applied
mathematics and software development activities. This bill,
along with the QUEST Act and SALSTA, is an important step
forward in improving our nation's global standing in science
and technology.
We know that maintaining U.S. leadership will require a
shared commitment to prioritize DOE and its Office of Science.
And nowhere is this clearer than in the advanced computing
space.The U.S. relies on computing capabilities that only the
Department of Energy can provide. We know that the nation who
takes the lead in advanced computing will set the stage for the
next generation of technologies and technology standards. We
cannot afford to fall behind in this race.
Last week, I was encouraged by the progress made by my
friends in the Senate to recognize the important role the
Department of Energy plays in advancing U.S. innovation. But
DOE and the National Labs shouldn't be an afterthought when we
consider the U.S. research enterprise. They're integral to our
scientific progress. That's why Chairwoman Johnson and I have
been working on bipartisan Office of Science legislation that
will make a strong commitment to the Department of Energy and
its work-including successful programs like ASCR.
This legislation to support research at the Department of
Energy will go hand-in-glove with the NSF For the Future Act,
which supports basic research, STEM education, and technology
transfer at the National Science Foundation. Together, these
research bills will solidify the long-term stability of our
international leadership in science.
I once again want to thank our witnesses for being here
today. I look forward to a productive discussion. Thank you
Chairman Bowman and I yield back the balance of my time.
Chairman Bowman. Thank you, Mr. Lucas, for your opening
statement.
If there are Members who wish to submit additional opening
statements, your statements will be added to the record at this
point.
At this time, I would like to introduce our witnesses.
First, Dr. J. Stephen Binkley is the Acting Director and
Principal Deputy Director in the Office of Science at the U.S.
Department of Energy. Prior to his experience in various
leadership positions in DOE, Dr. Binkley has held senior
positions at DOE's Sandia National Laboratories and the
Department of Homeland Security. He has conducted research in
theoretical chemistry, materials science, computer science,
applied mathematics, and microelectronics.
Next, Dr. Georgia Tourassi is the Director of the National
Center for Computational Sciences at the Oak Ridge National
Laboratory. She also holds appointments as an Adjunct Professor
of Radiology at Duke University and as a Professor of the
Bredesen Center Data Science Program at the University of
Tennessee at Knoxville.
Next, Dr. Karen Willcox is Director of the Oden Institute
of Computational Engineering and Sciences, Associate Vice
President for Research, and Professor of Aerospace Engineering
and Engineering Mechanics at the University of Texas at Austin.
She holds the W.A. ``Tex'' Moncrief, Jr. Chair in simulation-
based engineering and sciences and the Peter O'Donnell, Jr.
Centennial Chair in Computing Systems.
Dr. Christopher Monroe is Co-Founder and Chief Scientist
at IonQ Inc. and the Gilhuly Family Distinguished Presidential
Professor of Electrical and Computer Engineering and Physics at
Duke University. He is an atomic physicist and quantum engineer
specializing in the isolation of individual atoms as the core
of a quantum computer.
Last but certainly not least, Dr. Seny Kamara is an
Associate Professor of Computer Science at Brown University
where he codirects Brown's Computing for the People Project and
the Encrypted Systems Lab. He is also affiliated with Brown's
Center for Human Rights and Humanitarian Studies, the Data
Science Initiative, and the Policy Lab. Kamara is a principal
scientist at MongoDB, a company that provides one of the most
widely used platforms to store and process data.
Thank you all for joining us today. As our witnesses
should know, you will each have 5 minutes for your spoken
testimony. Your written testimony will be included in the
record for the hearing. When you all have completed your spoken
testimony, we will begin with questions. Each Member will have
5 minutes to question the panel. We will start with Dr.
Blinky--Blinkley, excuse me. Dr. Binkley, please begin.
TESTIMONY OF DR. J. STEPHEN BINKLEY,
PRINCIPAL DEPUTY DIRECTOR,
OFFICE OF SCIENCE AT THE DEPARTMENT OF ENERGY
Dr. Binkley. OK. Thank you, Chairman Bowman and Ranking
Member Weber. I'm pleased to come here before you today to
discuss the scientific computing capabilities of the Department
of Energy, including the forthcoming exascale systems.
DOE computing traces its roots back to the Manhattan
Project where extensive use was made of computers. During the
1950's, John von Neumann, the pioneer in computing, advocated
for a program that would advance computer development. Over the
years, ever more powerful computing capabilities were developed
at the national laboratories beginning with the Lawrence
Livermore and Los Alamos National Laboratories.
DOE and its predecessor agencies have supported applied
mathematics and computer science, along with major investments
in computer hardware and computational science that have been a
major driver of progress in high-performance computing,
spurring the U.S. computing industry forward. DOE computing
applications have expanded from their original national defense
focus to a broad portfolio of scientific research and
significant use by industry beginning with the establishment of
the leadership computing facilities at Argonne and Oak Ridge
National Laboratories in 2004.
Today, DOE computing is a partnership between the National
Nuclear Security Administration (NNSA) and the Department's
Office of Science. Our two organizations are working hand-in-
hand to advance high-performance computing, including the
exascale computing project.
The strategic importance of high-performance computing has
grown enormously. High-performance computing has become an
essential pillar not just of America's national security but
also of our leadership in science. DOE's supercomputing has
brought major computational and driven advances in a wide range
of fields such as climate science, fusion energy, and high
energy and nuclear physics, materials science, chemistry,
particle accelerator design, and biology, to name a few.
Over the last 6 years, we have been very focused on
achieving exascale computing. The first exascale computing
system is scheduled for delivery at the Oak Ridge National
Laboratory to be complete by October of this year. The second
system will go to Argonne National Laboratory in 2022, and a
third system to Lawrence Livermore National Laboratory in 2023.
Exascale has the capability to deepen our understanding of
climate change and hasten the development of clean energy.
Partnerships between the Office of Science and NNSA with major
computing and microelectronic vendors have been key in the
development of exascale. A series of five partnership programs
have brought DOE-supported researchers to work hand-in-glove
with U.S. high-performance computing vendors, including AMD
(Advanced Micro Devices), Cray, IBM (International Business
Machines), Intel, Nvidia, and HPE (Hewlett Packard Enterprise)
to overcome the key technical hurdles in exascale. In total,
DOE has invested $460 million in this effort alone, matched by
at least an additional $307 million contributed by industry.
Current and planned upgrades to Office of Science
scientific user facilities, including light sources, neutron
scattering sources, nanoscale, and genomic facilities will
bring more sophisticated and precise observations and vastly
larger data outputs. Artificial intelligence and machine
learning will play a key role in this.
AI also holds the promise of more sophisticated and
autonomous facility operations. It has the potential to monitor
observations and adjust instrument operations in real time to
further enhance efficiency and utilization of the facilities.
DOE's ESnet provides ultrahigh broadband connectivity across
the DOE laboratories as connectivity will be increasingly vital
as facility operations are controlled computationally.
We are looking forward and beyond exascale to new
frontiers such as quantum information science. Leadership in
science remains indispensable to our high-performance--to the
country's prosperity, and high-performance computing is key.
Continued stewardship and development of the skilled HPC
workforce is essential. Our Computational Science Graduate
Fellowship program is one such activity. Since its
establishment in 1991, the program has sponsored over 450
fellows from more than 60 universities. DOE's response to the
COVID-19 pandemic demonstrates the enormous value of DOE's
high-performance computational research resources.
In summary, opportunities for accelerated scientific
discovery will be enabled by current--the current era of high-
performance computing marked by the advent of exascale systems
and the rapid development of AI and machine learning.
And I'll end there.
[The prepared statement of Dr. Binkley follows:]
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
Chairman Bowman. Thank you so much, Dr. Binkley.
Dr. Tourassi, you're now recognized.
TESTIMONY OF DR. GEORGIA TOURASSI,
DIRECTOR, NATIONAL CENTER
FOR COMPUTATIONAL SCIENCES
AT OAK RIDGE NATIONAL LABORATORY
Dr. Tourassi. Chairman Bowman, Ranking Member Weber,
Chairwoman Johnson, Ranking Member Lucas, and distinguished
Members of the Committee, thank you for the opportunity to
appear before you today. My name is Georgia Tourassi. I lead
the Department of Energy's Oak Ridge Leadership Computing
Facility, OLCF, at the Oak Ridge National Laboratory in
Tennessee. I'm a biomedical engineer and a computational
scientist by education and training.
High-performance computing has been the cornerstone for
the Nation's scientific advancement, technology innovation,
competitive advantage, and economic prosperity. Its impact on
global competitiveness has long been embodied by the saying,
``You must out-compute to outcompete.'' With advances in data
technologies, machine learning, and AI, this saying can be
amended to, ``You must learn faster to outcompete.''
OLCF has been a global leader in high-performance
computing for nearly 30 years. Currently, OLCF hosts Summit,
the Nation's most powerful supercomputer for open science.
Summit is in high demand for modeling and simulation, data
analytics, and AI to better understand climate change, develop
new ways to produce clean energy, design advanced materials,
advance public health, and overall push the frontiers of
science. The request for time on Summit is up to five times
more than the hours available.
Our facility is both deliberate and responsive to national
needs. In the past year, I have experienced firsthand our
staff's Herculean efforts to be both fast and offer world-
leading computing resources and computational and data
expertise in the fight against the coronavirus. Through the
COVID-19 High-Performance Computing Consortium, Summit and our
competent staff, using world-leading AI, helped accelerate
discovery, understand the virus, and inform management of the
pandemic response. I would like to thank Congress for the CARES
Act funds OLCF received to augment Summit and help support the
COVID-19 research community.
Now, in 2021, OLCF is at the brink of delivering the first
exascale computer in the United States called Frontier. This
supercomputer will perform calculations up to eight times
faster than Summit and will keep the United States at the
forefront on high-performance computing. To prepare, the DOE's
Exascale Computing Project is developing critical applications
across many scientific and technical disciplines to run on
Frontier on day 1.
In addition, exascale will offer training opportunities to
grow a more high-tech and computationally savvy workforce in
our Nation. It is imperative for the United States to expand
and enhance the national research computing ecosystem.
DOE has asked us to deliver Frontier 1 year earlier than
planned, and we're focusing our efforts on meeting that
schedule. Once the system is delivered, we will need to
properly fund the operation and applications to solve complex
real-world problems in partnership with leading research
institutions, industry, and other Federal agencies.
We need to continuously invest in new technologies such as
AI and accelerated computing methods to maintain our
competitive advantage and ensure our global leadership. We need
to make investments in a national data infrastructure that
makes the most of our high-performance computing and national
data assets. The COVID-19 pandemic demonstrated the vital
importance of having established interagency programs and data
integration ahead of these anticipated crises and the utility
of high-performance computing and AI for rapid, complex, real-
world data analysis.
The DOE leadership computing facilities are uniquely
positioned to support and integrate its research
infrastructure, combining our leadership computing with
national experimental facilities and Federal data assets to
deliver unprecedented technological, scientific, and economic
advantages to the Nation.
We know high-performance computing is high on other
nations' priorities. We know that China, Japan, and the
European Union are all investing heavily in exascale computing,
and this has implications for both our national security and
our overall global competitiveness. We cannot afford to be left
behind.
Thank you again for the opportunity to testify. I welcome
your questions on this important topic.
[The prepared statement of Dr. Tourassi follows:]
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
Chairman Bowman. Thank you so much, Dr. Tourassi.
Dr. Willcox, you are now recognized.
TESTIMONY OF DR. KAREN WILLCOX,
DIRECTOR, ODEN INSTITUTE FOR
COMPUTATIONAL ENGINEERING AND SCIENCES
AT THE UNIVERSITY OF TEXAS AT AUSTIN
Dr. Willcox. Thank you, Chair Bowman, Ranking Member
Weber, Chair Johnson, Ranking Member Lucas, and Members of the
Subcommittee.
Today, I have three main points. First, the future of
scientific computing must be interdisciplinary. Second, the DOE
ecosystem that supports mission-driven basic research in
scientific computing is a national scientific treasure. And
third, the future of scientific computing hinges critically on
the availability of a highly skilled workforce passionate about
addressing the Nation's challenges in science, security, and
sustainability.
So first, on the interdisciplinary future of scientific
computing, the pace at which scientific computing can
accelerate discovery and innovation will be limited by the rate
at which we address foundational challenges that currently
limit the complexity, scale, and trustworthiness of
computational tools. This requires scientific computing
research that draws on many fields, including computer science,
computational science, the mathematical sciences, the domain
sciences, and engineering.
Particularly important is the role of the field of
computational science. Computational science differs from
computer science because at its core, computational science
involves developing mathematical models and simulations rooted
in physical and mechanistic principles.
As we look to the future of scientific computing, the
boundaries between computational science and computer science
are becoming increasingly blurred. The future of scientific
computing will involve new approaches that span the two fields
such as AI and machine learning, and indeed the DOE has been at
the forefront of defining notions such as AI for science and
scientific machine learning.
However, when it comes to AI approaches in science and
engineering, we must be careful not to chart our course based
entirely on the successes of data science and machine learning
in vastly different domains such as social media and online
retail. We must instead recognize that energy, environmental,
and nuclear challenges by their very nature require predictions
that go well beyond the available data. There's a critical need
to quantify uncertainty and to make informed decisions that
account for risk. The future of scientific computing will only
address these needs through a balanced investment in the
foundational mathematical sciences and in computational
science, along with data science and computer science. And we
must also not underestimate the criticality of continuing to
invest in experimental research and development since advancing
discoveries through computational models really requires
validation.
That brings me to my second point on the value of the
DOE's mission-driven basic research ecosystem and its role in
addressing these challenges. DOE supports for basic research at
the national labs and at the Nation's universities has fostered
interdisciplinary computing research in a way that community-
driven basic research has struggled to achieve. And as one
example, I highlight the Mathematical Multifaceted Integrated
Capabilities Centers, or MMICCs, of the DOE applied math
program. These centers focus on applied math basic research but
strongly driven by application needs. For example, our AEOLUS
(Advances in Experimental Design, Optimization and Learning for
Uncertain Complex Systems) MMICC is addressing the basic
mathematical research needs for advanced materials and additive
manufacturing.
The MMICC program has been transformational in how it has
shaped my own basic research portfolio, and one of the critical
elements, first, the size of the center is large enough to
bring together a diverse team that includes mathematicians,
computer scientists, computational scientists, engineers, and
domain experts spanning universities and national labs. This in
turn enables a much-needed holistic approach for a complex
system.
Second, the long funding horizon provides the stability to
invest in challenging, high-payoff basic research ideas.
And third, the mission-driven nature challenges my
mathematical research to target problems that are of high
relevance to practitioners if the focus on basic research
permits us to lay long-lasting foundations.
My final point is that achieving this future vision for
scientific computing hinges critically on the availability of a
highly skilled workforce. The challenges in front of us are
twofold. First is training the workforce with the
interdisciplinary skills that cut across the mathematical
sciences, computing, and domain sciences, and second is
ensuring a strong, diverse pipeline of highly trained
professionals who remain committed to scientific and
engineering domains rather than being lured away by more
lucrative positions in commercial and business sectors. A
critical part of this training is the immersive research
experiences enabled by basic research grants such as the MMICC
program I described earlier. Maintaining a strong investment in
DOE basic research funding for universities while also
continuing to support the collaborative and academic alliance
programs at the national labs is absolutely critical to
addressing the Nation's future workforce needs.
Thank you, and I look forward to your questions.
[The prepared statement of Dr. Willcox follows:]
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
Chairman Bowman. Thank you so much, Dr. Willcox.
Dr. Monroe, you are now recognized.
TESTIMONY OF DR. CHRISTOPHER MONROE,
CO-FOUNDER AND CHIEF SCIENTIST, IONQ, INC.
Dr. Monroe. Good morning. Thank you, Mr. Chairman, Members
of the Subcommittee, for this opportunity to testify before you
today. I'm here on behalf of IonQ, a company that builds
quantum computers. IonQ is headquartered in College Park,
Maryland, and was spun out of the University of Maryland and
Duke University about 5 years ago. I'm also a Professor of
Electrical and Computer Engineering and Physics at Duke
University. I have over 2 decades of experience in the field of
quantum computing technology from both academic and industrial
perspectives, and I'm here to talk about the future of
computing in terms of quantum information.
Quantum computers, as you may have heard, are--they're as
revolutionary as they are challenging to grasp and build. Their
might, given these challenges, demand special attention. As you
know, the 2018 National Quantum Initiative, or NQI, was
initiated by the House Committee on Science, Space, and
Technology to ensure that the United States remains at the
forefront of this technology. The NQI endowed the Department of
Energy, the National Science Foundation, and the National
Institute of Standards and Technology (NIST) with coordination
from the Departments of Defense (DOD) and the intelligence
community to stimulate foundational research in quantum
computing and other quantum technologies and translate this
technology from laboratory to industry.
So how does a quantum computer work? It's essentially not
hard. It's just that quantum computers follow laws of physics
that have no analogy in everyday life, so it's confounding.
It's not exactly that it's hard. Information in quantum
computers can exist in superposition; that is, multiple values
can be stored and processed simultaneously in a single memory
device. But each time you expand a quantum computer by just a
single bit--we call them quantum bits or qubits--its power
essentially doubles. So with just 300 quantum bits, that's a
pretty small chunk of matter, a quantum computer can process
more possibilities than there are atoms in the entire universe.
This massive parallelism in quantum computers allows certain
computations to be performed that could never be accomplished
using regular computers.
So a few far-reaching applications for this new mode of
computing include optimization of complex problems dealing with
huge amounts of data, including logistics and things like
pattern recognition; secondly, molecular and materials design
for energy, medical and defense applications; and finally,
security, including secure communication, encryption, and
decryption or code-breaking.
IonQ has collaborative projects in all of these areas, and
one thing that's interesting in this field is it's such an
early stage of this technology that it's----
Chairman Bowman. Mr. Monroe, time has run out. Can you
just finish that last point?
Dr. Monroe. Yes. I think somehow the clock never started,
I noticed.
Chairman Bowman. Yes, that's probably right. That's weird.
Dr. Monroe. I think I was talking for about 2 minutes
but----
Chairman Bowman. Yes, let's put 2 minutes on the clock if
we can.
Dr. Monroe. OK. I'll go quickly. So it's critical that
quantum computer builders co-design applications with the
systems they build. It's really not physics anymore, but all of
the physical and chemical sciences, all of the engineering
fields, computer science, algorithm design, economics, and even
social sciences. So it's no surprise that one of the most
important applications in quantum computers is energy and that
the Department of Energy is an important player in advancing
this field.
So IonQ machines and those built by others are still too
small to beat regular computers in these types of problems, but
we're just at the beginning of this commercial phase, and this
situation will change very soon.
The core of a quantum computer is exotic, and its key
attribute is isolation. It involves devices either cooled to
nearly absolute zero temperature--that's negative 460 degrees--
or in the case of our technology at IonQ, we use individual
atoms suspended in a small vacuum chamber and poke with laser
beams. I should also note that this technology was developed at
the National Institute of Standards and Technology in the
1990's where I worked with David Wineland in developing the
first quantum logic gate.
So, as exotic as this is, the core technology is not
necessarily the main challenge. The real challenge as I see it
is creating the workforce to understand how to deploy quantum
systems. I like to say at the universities we typically don't
build components for people to use, but industry so far has
been a little bit slow to develop this technology because they
don't have a basis in quantum. And so this is, I think,
historically where government laboratories can play a role. At
IonQ, our systems are now available on cloud servers, and at
Duke University we're setting up a quantum--Duke Quantum Center
that will be a scientific user facility that will serve the
scientific use cases. And this is very important for the field.
So I want to conclude. I think I--it was a lot faster than
I thought it would be. But now is a critical time for DOE, NSF,
NIST, DOD, and the intelligence community to redouble and
coordinate their efforts in translating quantum computers to
the real world. One example is the QUEST program that
Congressman Lucas mentioned that would subsidize access to
industrial quantum computers. Another is endowing the NSF with
a technologically driven division mandate. This is a
particularly good way to ensure that this emerging technology
gets used in the field.
I'm a member of many advisory boards in Europe and in
Asia, and I'm well aware of the coordinated investments
overseas. The United States must lead the race to build quantum
computers and other quantum technologies, and I think that the
programs and continued stewardship of the National Quantum
Initiative by DOE and the other agencies I mentioned are
critical to continued leadership.
I again think the Committee and Chairman for his
leadership and for the opportunity to testify today even though
I took, I think, 4 minutes. Thank you.
[The prepared statement of Dr. Monroe follows:]
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
Chairman Bowman. Thank you, Dr. Monroe. Apologies about
the issues with the clock.
Dr. Kamara, you are now recognized.
TESTIMONY OF DR. SENY KAMARA,
ASSOCIATE PROFESSOR, BROWN UNIVERSITY
Dr. Kamara. Thank you. Chairman Bowman, Ranking Member
Weber, Chairwoman Johnson, Ranking Member Lucas, and
distinguished Members of the Committee, I appreciate the
opportunity to testify at today's hearing on the future of
scientific computing at the Department of Energy.
By the end of the year, the Oak Ridge National Lab will
receive the world's first exascale supercomputer. This computer
will be able to process 10 to the 18 or one quintillion
operations per second. It is hard to overstate how difficult
this is to achieve and what an accomplishment it is. This
considerable leap in computing power will open the doors to new
discoveries and significantly impact a multitude of fields,
including medicine, meteorology, cosmology, and artificial
intelligence.
It is clear that the world-class research and high-
performance computing that has been conducted by U.S.
universities, national labs, and industry in order to achieve
exascale computing will affect our lives for the foreseeable
future. But as we enter the era of exascale computing, I would
like to provide a word of caution. I'm sure we can all agree
that computing and the technologies it enables have had a
tremendous impact on society. Because of this, it is easy to
assume that technological progress always leads to positive
outcomes and that new technologies benefit everyone equally.
But this is not the case. Technology, like policy, can have
disparate impact. It can enable positive outcomes for some and
cause great harms to others.
Consider, for example, advances in facial recognition
which allow us to log into our smartphones faster but also
enables suspicion-less mass surveillance with the progress in
computer vision and robotics that enables new drones that can
deliver medicine to hard-to-reach rural areas or missiles at
the push of a button by somebody sitting in a room thousands of
miles away. We must always remind ourselves that technology is
not inherently good and does not always benefit everyone
equally by default. In fact, we need to think hard about the
harms technology can cause and work even harder to mitigate
those harms.
One of the many important applications of exascale
computing is artificial intelligence and machine learning, for
example, to predict how a cancer patient might respond to a
particular treatment. But as we know, thanks to the work of
scholars like Cathy O'Neil, Joy Buolamwini, and Timnit Gebru
and to outlets like Pro Publica, machine learning algorithms
can be biased and can exhibit different behaviors on different
populations. And as has been widely documented, these biases in
machine learning most often harm people of color and those from
marginalized communities.
So while we should appreciate that thousands of world-
class scientists and engineers across the country are
diligently working toward making exascale machine learning for
cancer a reality, we also have to ask how many are working on
ensuring that these cancer treatment prediction models work for
people of all genders and of all races?
The investments we are making in exascale computing will
improve national security, the U.S. economy, and industry, but
will everyone benefit equally from this investment? Will the
13-year-old girl from Washington Heights, New York, benefit
from this investment as much as the tech, energy, and
pharmaceutical industries? Will there be as much effort to use
these supercomputers in the fight against sickle-cell anemia as
other diseases?
Exascale computing is not only an incredible achievement
but it's an incredible resource with the power to shape our
lives and those of future generations. As such, we must be
careful and thoughtful about how we make use of it. In
particular, it is incumbent upon us to make sure that we deploy
and use this resource in a manner that is fair and inclusive
that benefits not only the powerful but those who have
historically been marginalized by society and by technology.
Thank you, and I look forward to answering your questions.
[The prepared statement of Dr. Kamara follows:]
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
Chairman Bowman. Thank you, Dr. Kamara.
At this point, we will begin our first round of questions.
The Chairman now recognizes himself for 5 minutes.
Dr. Kamara, I'm going to start with you. Thank you so much
for your testimony and for your attention to making sure that
all people benefit equally from investments in scientific
computing and other new technologies. And I understand that you
are cofounding a new research institute at Brown University
called Computing for the People. Can you talk more about why we
should incorporate these kinds of questions into the R&D agenda
from the very beginning? What can we do as Congress to design
research programs that will prevent harmful applications down
the road, keep the needs of marginalized communities and all
people in mind, and prepare the computing workforce to engage
with these issues as well?
Dr. Kamara. Yes, so, as I said in my statement, it's clear
that computer science and technology have had a huge impact,
but the reality is that, as a field, we haven't really centered
the problems of marginalized groups. It's just not something
that comes naturally to the field. And there's many reasons for
that. Some include the lack of diversity in computing, which is
something that is well-documented. And so there's no natural
way for--you know, for computer science research and technology
research to really address those problems. And so this is why
we're building this institute at Brown, and the motivation is
to really make it a priority, right, understanding the problems
that marginalized communities face and how technology can help
and really focusing on that as our main motivation. So that's
what we're doing.
The way that Congress can help is both in funding this
kind of research and in asking questions, right, making sure
that the needs of all people are addressed by technology and by
computing research, just as we're doing in this hearing today.
Chairman Bowman. I was muted. Sorry about that. How much
is a deep dive understanding and analysis of implicit bias in
algorithms and computer science a part of this work that you're
referring to?
Dr. Kamara. Yes, it's crucial. So when we design--and
it's--and I also want to highlight that it's not only at the
level of research. It's also the level of education. So when we
teach our students computer science and we teach them how to
design algorithms, we're not teaching them how to think about
bias and how to address it. We're not teaching them how to
think about energy efficiency and how to design algorithms that
are not only fast but also that minimize the amount of energy
consumption. So these are all things that we want to do in our
institute. And basically we want to integrate what we call
responsible computing into computer science education and into
computer science research as well.
Chairman Bowman. OK.
Dr. Kamara. Learning how to find biases in data, learning
how to find biases in algorithms is a crucial part of that.
Chairman Bowman. Thank you very much.
Dr. Tourassi, thanks for your testimony as well. I was
intrigued by your suggestion for investing in a national data
infrastructure that could be housed at DOE labs and that could
play an important role in democratizing AI. Can you say more
about what this would look like?
Dr. Tourassi. Absolutely. We know that the explosive
growth of AI is based on the three pillars, the supercomputing,
the algorithms, and data. We talk a lot about investments on
the supercomputing side and algorithms, but data is the fuel
that will make the engine--the airplane fly. And, as I pointed
out in my written and oral testimony, the past few years have
taught us the importance of having a data infrastructure that
makes the most of our Federal data assets, examples with the
partnership with the National Cancer Institute, as well as with
the Veterans Administration.
Building on that thread, because that infrastructure was
in place and the interagency partnership was in place with the
VA, we were able to pivot fast to address challenges with the
COVID-19 pandemic because the infrastructure was in place to
accept data, new data related to COVID-19 cases in the veteran
population to start doing within 48 hours large-scale
epidemiological studies and observational studies.
So this is what I meant in my statement that this was an
important lesson learned that we need to be proactive and to
put all resources necessary to support that infrastructure. DOE
has a long history of building and sustaining successfully data
infrastructures and enable--enabling the broader scientific
community to make the most of it.
Chairman Bowman. Thank you so much, Dr. Tourassi.
I now recognize Mr. Weber for 5 minutes.
Mr. Weber. Thank you, Chairman. I appreciate that. And I'm
going to go back to you, Dr. Tourassi. We're going to keep you
on the hot seat for a little longer.
First of all, I'd like to congratulate you on all your
hard work at Oak Ridge as Director of its Leadership Computing
Facility. I'm sure you just didn't wake up one day and decide
you were going to get into something like that. You had
probably been studying and working at that a long time. And I
know I speak for all of us when I say I cannot wait to see
Frontier in action this fall, so congratulations.
You've noted, Doctor, that Frontier will on day one of its
operation no less, be ready to run applications across two
dozen scientific and technical disciplines. Can you take a
moment for us not-so-technical people maybe to explain why,
when it comes to the race to the world's fastest supercomputer
that the hardware development can't be our only focus? Talk to
us, please.
Dr. Tourassi. Absolutely. As you said, the tool by itself
is not the enabler. We need to see the scientific impact. And
scientific impact is measured in many different ways. First of
all, by the breadth of application domains that benefit by the
tool, how much faster we can do scientific discovery across
that breadth of applications, and also how efficiently we use
the tool.
So in partnership with the Exascale Computing Project,
there is a portfolio of applications that have many critical
mission areas, some of which I highlighted in my report. And we
are working very closely with the exascale computing team to
make sure that the necessary software tools will be in place to
enable the scientific community that will need them, as I said,
from day one. Some of these applications in climate change, in
renewable energy, in advanced manufacturing, in biology and
public health because we know that these are pressing
application areas with great societal impact.
Mr. Weber. Absolutely, thank you. How does the utility or
usability of DOE's supercomputer like Summit, for example,
we're talking about competitors. How does it compare to some of
our competitors' supercomputers?
Dr. Tourassi. So, clearly, these metrics are not monitored
and publicly known, but we can say for sure for the United
States and the leadership computing facilities not only we are
oversubscribed, we operate in the high 90's efficiency.
Anecdotally, some of our competitors are certainly behind, so
they have the tool, not necessarily the most effective use of
the tool. But, as I said, this is a race. We cannot just relax
in that race. We need to keep moving forward and making sure
that we are spearheading the specific domain.
Mr. Weber. Well, thank you for that. I'm quite sure China
has all the information on everybody's capabilities, so we know
that.
So we do want to keep up with our competitors. We'd love
to have that information. And point out to us why it really
matters. I know we want to be first in the race, but what other
examples would you use other than just time and being first?
Why does it matter that we get ahead of our competitors? Give
us some examples that tie to our country. Can you do that?
Dr. Tourassi. Well, certainly the issues of national
security are extremely important, and we know that
supercomputers have played a very important role in addressing
applications in the national security space, and we expect that
space to grow even further.
As the other panelists and you all mentioned the issues of
adversarial use of the technologies we develop, so we'd need to
be very much aware that this is not only in our horizon, it is
present.
Mr. Weber. Well, perfect. That's kind of where I hoped you
were going.
I'm going to jump over to you, Mr. Binkley--or Dr.
Binkley, I'm sorry. As I mentioned in my opening statement, I
believe that in order to fully realize the potential of the
world-leading computing capabilities stewarded by our national
laboratories, the Office of Science must prioritize expanding
ASCR research partnership with other Federal agencies. Very
quickly, Department of Energy's position to conduct scientific
computing research to resolve diverse challenges, how so in
your opinion? I'm going to be over time a little bit.
Dr. Binkley. So we have in fact over the last half a dozen
years really put priority on growing the ASCR program. It was
clearly in recognition of the need to develop and then deploy
exascale. And, as has been elaborated on here in the last 5 or
6 minutes, concurrent with the development of the exascale
computer systems has been a very concerted effort to develop
applications that are ready--that will be ready on the day the
machine is first turned on.
Mr. Weber. Yes.
Dr. Binkley. The other thing that we've done is we have
worked through the National Science and Technology Council
processes to coordinate closely with other Federal agencies,
including the National Science Foundation, NASA (National
Aeronautics and Space Administration), NOAA (National Oceanic
and Atmospheric Administration), et cetera, and have tried to
encourage the other agencies to also make investments in the
high-performance computing capabilities for their specific
applications. And in the case of the National Institutes of
Health, as Dr. Tourassi has pointed out, we--one of the
exascale applications is focused on cancer problems stemming
from the National Institutes of Health. And I'll stop there.
Mr. Weber. All right. Well, thank you. I appreciate that.
And then, Mr. Chairman, thank you for the indulgence.
Staff. Ms. Bonamici is next.
Ms. Bonamici. Thank you so much to Chair Bowman and
Ranking Member Weber, and thank you to the witnesses.
In the district I represent in northwest Oregon,
researchers at Intel are developing the foundation for exascale
computers, commercially viable quantum systems, and also
partnering with the Department of Energy to advance other high-
performance computing technologies. And I know these efforts
will help us transition to clean energy economy, better predict
extreme weather events, strengthen preventative medicine,
improve emergency response, and more.
So, Dr. Binkley, it's sometimes challenging to
conceptualize the benefits of the Department of Energy's work
on scientific computing. I am the Co-Chair of the House Oceans
Caucus, so I especially appreciated your specific example of
the work Lawrence Berkeley National Laboratory (LBNL) and Los
Alamos and the University of Bristol are doing to provide more
accurate pictures of retreating ice sheets contributing to sea
level rise. How will future exascale capabilities strengthen
our understanding and response to the climate crisis, and how
can Congress better support this work expanding SciDAC
(Scientific Discovery through Advanced Computing) partnerships
across the DOE?
Dr. Binkley. So let me take your last question first.
The--with the incoming Biden-Harris Administration, there has
been a reorganization within the Department of Energy that
brings the applied energy programs back into the Under
Secretary for Science organization, and that gives us very
close coupling with the applied energy programs. That is
between the Office of Science and the applied energy programs.
And I'm hopeful that that will set the stage where we can
expand SciDAC to reach the other parts of the Department.
Going back to the question about climate simulations, our
Earth systems model, which has been in development now since
about 5 or 6 years and is slated to be, you know, up and
running on the exascale computer in the fall, we are
systematically increasing the resolution of that model by use
of the--you know, the power of the exascale computers to have
higher resolution, more predictive models of climate effects
and then, you know, we will be able to predict areas that are
going to be problematic in the future. You know, that will
become, I think, a fairly standard tool for predictive earth
systems modeling. I'll stop there.
Voice. You are muted.
Ms. Bonamici. Sorry about that. Dr. Binkley, thank you. I
do want to get in another question.
Dr. Kamara, we know that technology is not developed or
used in a vacuum. The growing body of evidence suggests that,
left unchecked, digital tools can absorb and replicate systemic
biases that are ingrained in the environment in which they are
designed. For example, one hiring algorithm often referenced
for its apparent biases identified high-performers as anyone
named Derek who played lacrosse even though those features had
no connection for the actual jobs which the firm was screening.
Unfortunately, digital tools are opaque in their design and
operation, and it's hard to hold it accountable. So what steps
can Congress take to prevent harmful bias in the research,
development, and commercialization of supercomputing
technologies? And will advances in neuromorphic computing help
to end/or minimize bias or simply make it harder to detect?
Dr. Kamara. Yes, so there are a lot of ways that bias can
creep into algorithms, and there's a lot of research going on
on how to mitigate that, so a lot of people are thinking hard
about that. And that's great, but what we also have to be
careful of is that just because we can automate something
doesn't mean that we should, right? So it's not just--it's not
enough to just say, OK, well, let's design unbiased algorithms.
We can make some effort, you know, toward that, but we also
have to ask ourselves should we be using algorithms, you know,
in these particular cases, right? So should we be using
algorithms for--you know, for estimating the probability that
someone is going to commit a crime? Is that the right use of
this technology? OK. So there's a lot of different sort of
aspects of this that we have to think about.
And the way Congress can help is in funding more research
on fair algorithms and also on providing some structure for
auditing algorithms. That is also an important component of
this is that we need to be able to say, OK, well, you know, if
an algorithm is going to be deployed, right, if we are going to
make the decision to deploy some kind of algorithm in a
particular use case, then we really should have rules in place
for how we're going to audit those algorithms, how we're going
to make sure that they don't discriminate and that they are not
biased.
Ms. Bonamici. Great. And your thoughts on neuromorphic
computing and what that means to bias and minimizing bias?
Dr. Kamara. Well, that's a little bit outside of my scope
of expertise, but I would definitely say that, you know, these
different types of sort of--as we--as we diversify the kinds of
algorithms that we're using, it's going to be harder and harder
to detect bias, right? So we also have to really be really,
really mindful of this. It's easy to think that, well, just
because an algorithm is running or some kind of algorithm--
there's many different types of--that, you know, there's sort
of this veneer of objectivity, but this really isn't the case,
right, and so we have to be really vigilant against that.
Ms. Bonamici. Thank you, Mr. Chairman. I yield back.
Staff. Mr. Baird is next.
Mr. Baird. Thank you, Mr. Chairman, and thank you, Ranking
Member Weber, for holding this session. It's very interesting
and exciting to me. Of course, my background is agriculture,
and so I'm going to relate agriculture to this quantum
computing.
I recently introduced a bill, 2961, that really says that
the Department of Energy Biological Innovation Opportunities
Act, and that's making sure that DOE has sufficient
infrastructure to be able to provide access to quantum
computing to researchers and university people, as well as the
industry. So I was trying to mention that because it'll be a
part of Ranking Member Lucas's SALSTA bill and keeping us in
the leadership around the world.
But, Dr. Monroe, you noted in your written testimony that
translating quantum computing to practical applications will
create opportunities for the workforce. I would also suggest
that it will provide students and researchers the ability to
analyze large-scale, complex, practical situations. And an
example of this is the National Institute of Food and
Agriculture's Genome to Phenome Initiative. It's designed to
look at plant materials and give plant researchers the ability
to analyze those plants and look for the visual characteristics
and tie that to the genome. And that all is designed to improve
our ability to raise food and feed a hungry world in the
future.
So, Dr. Monroe, would you care to elaborate--you were
talking about practical applications and the need for that.
Would you care to elaborate on what quantum computing can do
for us in agriculture and how you think that fits into the
progress of our world?
Dr. Monroe. Yes, thank you for that question. So, yes,
sorry for my garbled statement. I think I had to kind of jump
around a little bit on the hot seat there.
So quantum computers appear to be good at solving generic
optimization problems, that is, problems that have way too many
inputs that we can't sample all of the configurations on a
conventional computer. Those are called combinatorial
optimization problems, and we take guesses with even high-
performance computers. And as I hinted, there are problems out
there that we'll never be able to solve because the number of
configurations is just way too big.
And so you mentioned agriculture, but I would maybe
broaden in it to pharmaceutical, energy, and, you know, gas and
oil industry. The need to understand the structure of
molecules, this is something computers are also very poor at
because even a small molecule, if it has more than a few
hundred electrons--and this is--you know, even a caffeine
molecule has 100 electrons--we can't easily model that molecule
to see how it interacts with others to form catalyzers, better
fuels, to form better drugs even.
But at the same time there's a logistics problem even in
big pharma if you have 10,000 compounds, which combination of
10 of them make a good drug for something. We may have models
of that, but, once again, we can't optimize that.
And on materials science side--and this is closely related
to what you had mentioned in plant science and agriculture--can
we develop new materials that harvest sunlight much more
efficiently than known materials today? It's--this--for the
same reason. We can't compute or simulate the behavior of these
very fundamental things.
So, you know, quantum computing is just starting, but
that--those are the types of problems that are naturally
attackable by a quantum computer. And I'll put it this way.
Those problems will never be solved using conventional
computers. If they are to be solved, they will demand a quantum
computer. So it is a very researchy field now, but we're
starting to launch into building devices. And the Department of
Energy over the last half-dozen years has played a big role in
their laboratories in trying to translate that research, basic
research to product. And I think industry is starting to play a
role as well, very big-named industries that you've heard of.
Mr. Baird. You're exactly right. You know, this genome to
phenome, they take these spray rigs that are 60 foot wide with
cameras down into look at each and every plant, and there's no
way in my background in the past that we could have had the
ability to analyze that much data, so thank you for that.
And I see I've got about 13 seconds left, and I have other
questions I would like to ask the other witnesses, but I thank
all of them for being here and appreciate this opportunity, and
I yield back.
Staff. Mr. McNerney is next.
Mr. McNerney. Well, first of all, I want to thank the
witnesses. It's great to hear your viewpoints and share the
excitement about what's going on with these exascale computers.
Dr. Monroe, you mentioned some of the potential
applications, including pattern recognition in huge data sets,
molecular and material design, and secure communications. The
DOE and several other agencies are engaged in the National
Quantum Initiative, which was authorized by Congress in
December 2018. This initiative includes basic research
activities and establishes large-scale Quantum Information
Science Research Centers. How have these activities and the DOE
research ecosystem in general helped strengthen the Nation's
quantum computing industry?
Dr. Monroe. Thanks for the question. Indeed, I think DOE
has shown in force their establishment of five very highly
scoped centers distributed throughout the country that each
have a separate mission and aim in the general field of quantum
information, not just quantum computing but quantum sensing,
things like quantum simulation. And these--I think the DOE
laboratories are a great place, think about this, because the
DOE labs in a sense are that ideal combination of having
seasoned engineers, device people, but also not being a
stranger to the weird laws of quantum physics that underlie
these devices. So things are just getting started.
I--as a disclaimer, I am a key contributor to a DOE
quantum center headquartered at Lawrence Berkeley National
Laboratories in your State, and it's a very exciting
collaboration that will--our consortium will build systems, and
we're interested in sort of the system engineering of--when you
build a big system, it's a--sort of breathing thing that's not
just a sum of the individual parts. And with the DOE laboratory
at LBNL and also Sandia National Laboratory, New Mexico is
involved, we hope to really further the field.
But I should also mention the National Science Foundation
and NIST are big players. They have historically been at the
field since the beginning in quantum, and they--you know, one
beauty of our national system in funding science is that we
have many agencies, all with different missions, and, you know,
DOE obviously I think the centerpiece are the laboratories.
NSF, they are more vertically organized in a certain sense.
They can bring together physicists with computer scientists and
everything in between to come to bear on this. So I think, you
know, collaboration between those agencies is really what's
going to keep the American in the lead in this field.
Mr. McNerney. Excellent. Dr. Binkley, you first discussed
the importance of strengthening the core research that feeds
into technologies like quantum sensors and networks. Would you
discuss that a little bit, please?
Dr. Binkley. Yes, the--one of the major premises of the
DOE activities in quantum information science is to be really
focused on the fundamental science first and then follow that
through with technology developments to get practical
applications. And so, you know, there--if you look across the
Office of Science portfolio of activities, you know, it's
chemical and material sciences--well, know, those are systems
that are governed by the laws of quantum mechanics. It's
nuclear physics and high-energy physics governed by the laws of
the standard model. And, essentially, quantum mechanics
permeates through all of the physical sciences activities that
are supported by the Office of Science. And, you know, that is,
I think, a major focus of how we're organizing our QIS
activities.
Mr. McNerney. Well, thank you. Moving on, Dr. Monroe, in
your written testimony you note the proposed QUEST program. How
much annual funding is needed for this program? Do you think it
would make sense to ramp up the program from a relatively
smaller amount over the first few years?
Dr. Monroe. I think there's some amount of sense there. I
think quantum computing systems that are capable, that itself
is ramping up right now. I think things started maybe 4, 5
years ago, very small systems, and these are provided by
companies you've heard of, some maybe you haven't like my
company IonQ but also, you know, Google, IBM, other companies
are putting these devices out there. They're very expensive
and, you know, I speak from experience that industry, the bet
that quantum computing will have a commercial payoff, it's--you
know, it's going to be a long-term bet, and it's not easy for
companies to play that risk, and I think this will allow the
Department of Energy and the U.S. Government to help connect
these systems to users and help subsidize the use of those
machines.
And I'll also say--and one thing that--the wildcard to me
is that the killer application for quantum computing, I don't--
I'm not sure when it happens. I'm not--I can't predict what it
will be, but I'm pretty sure that it will happen out of left
field from somebody that thought of a problem that I don't know
about or none of us makers know about, and getting them
connected to the system will just hasten progress in the field
because as soon as we hit that killer application, I think it's
just going to explode.
So, you know, your mentioning of ramping up the access
program. I certainly think that's maybe not a bad play, but I
think it needs to get going. We need more industrial
involvement in this field, less risk. Thank you.
Mr. McNerney. Well, I've run over my time, so I'm going to
yield back at this point.
Staff. Ranking Member Lucas is next.
Mr. Lucas. Thank you. As I've made clear, I believe that
ensuring strong support for the Department of Energy and its
world-leading national lab system is essential for our global
leadership in science and technology. And this is one of my
highest priorities as Ranking Member of the Science Committee.
So I ask my first question to Dr. Monroe and then I'd like
the rest of the panel's comments, too. And bear in mind that
part of our responsibility as Ranking Member and my colleague
as Chairman of this Committee is not only to try and create the
right policy here in the Science Committee but also we have to
be able to persuade our colleagues in the body as a whole and
ultimately the American taxpayers that we are on the right
track.
So I ask the following question in that vein. In your
respective areas of expertise, what would it mean to U.S.
leadership in advanced scientific computing if we fail to
provide adequate support for DOE Office of Science? Start with
you, Dr. Monroe, and whoever in the panel would care to touch
on that.
Dr. Monroe. Sure, I'll----
Mr. Lucas. And this is a message from my colleagues who
may not be spending time listening to this Committee hearing,
which flabbergasts me, I acknowledge.
Dr. Monroe. Sure, I'll answer very briefly. In my own
field of quantum computing that, again, we're at the early
stages of this field, and it is absolutely critical that
industry eventually take it over. And I think the DOE labs in
particular but also university laboratories supported by NSF
and NIST under the NQI is essential that they are able to make
the investments necessary to keep the United States ahead of
the world in this field. We're ahead now. We have mighty
industry just waiting in the wings to take it over. We have to
translate it. It's absolutely critical given the coordinated
investments from folks around the world.
Mr. Lucas. Anyone else care to take a stab?
Dr. Tourassi. So I could go next. Certainly if we----
Mr. Lucas. Please.
Dr. Tourassi [continuing]. Start with investments,
essentially, we are going to stagnate scientific innovation. We
will stop innovating not only across basic sciences but also
across applied sciences. Since we are using quantum computing
as an example, as Dr. Monroe said, this is the beginning of a
promising and disruptive technology, but it will take more than
a decade for that technology to become reality for all of us.
What happens in between? And it is classical computing that
will lead to those innovations of materials, new materials that
are needed to advance quantum computing. So it is a relay, and
we cannot just pause and wait for the next big thing to come
out from another nation.
Dr. Binkley. I could go next. There's another dimension to
this that I think is really important to keep in mind, and that
is that to be first in innovation, to be first in economic
security, and so on, national security, you know, it's really a
race for getting the best and brightest people into the United
States to do new research in areas like quantum information
science that have not been a topic of heavy investment before.
You know, it's essentially--the best tools attract the best
people, and that works in our advantage in an international
context. And so having, you know, the best workforce ultimately
is the--something that we need to strive for.
Dr. Willcox. I echo Dr. Tourassi's comments and to say
that under this the Office of Science would absolutely stagnate
and in fact set back our ability to tackle the critical
challenges in energy security and environment particularly, and
the Nation, the world just can't afford that.
Mr. Lucas. With that, I thank you. You make very
compelling cases. My time is winding down. I would observe one
other the thing. We are very sensitive in Congress these days
about not only international competition but how we protect
U.S. research from theft, at the same time encouraging
transparency and a cooperative environment. Anyone in a few
seconds who'd like to touch on that, I'm more than pleased to
hear what you have to say. But that's the struggle we're also
facing, how to give you the tools, create the environment, but
at the same time preserve your good work and the work being
supported by the American taxpayers.
Dr. Binkley. Well, I would add, Congressman Lucas, that's
something that we are really focused on. In fact, the meeting I
was in just prior to this hearing was a National Science and
Technology Council meeting focused on research security. And
the--I can attest to the fact that a number of--in fact, all of
the major science funding agencies of the Federal Government
are really focused on this problem these days, and it's a very
difficult problem. You know, we're trying to develop policies
that will protect the U.S. interest. It's a very thorny issue.
And in the Department of Energy, we've already--because we have
such a large research establishment in our national labs, you
know, we have over the last 3 or 4 years begun implementation
of policies to help protect the results of research from being,
you know, taken illicitly.
Mr. Lucas. Thank you, Doctor, and thank you to the entire
panel. My time is expired. I yield back, Mr. Chairman.
Staff. Ms. Stevens is next.
Ms. Stevens. Great, thank you so much. Congresswoman Haley
Stevens from southeastern Michigan and a major fan of
supercomputers and supercomputer technology and today's
hearing. I couldn't be more grateful.
In particular, I've had the privilege of working on a
supercomputer program called the National Digital and
Engineering Manufacturing Consortium working in partnership is
a public-private partnership between Purdue, OSU (Ohio State
University), and a handful of small businesses in the Great
Lakes area funded through the Economic Development
Administration.
Also, I think I'm getting some feedback, Weber. That might
be you. I think I'm getting Weber's chit chat, which I love
hearing, but, you know, I want----
Chairman Bowman. Mr. Weber, please mute your mic.
Ms. Stevens. OK, great, thanks. Thank you, Mr. Chair.
Thanks.
So we get this NDEMC (National Digital Engineering &
Manufacturing Consortium) program. We saw it with Jaco Plastics
was able to develop a new product line resulting in 100 new
jobs.
And, Dr. Tourassi, first of all, it's always a privilege
to have somebody from Oak Ridge at a science hearing, and we
have a tremendous amount of respect for your capabilities and
partnerships out of Oak Ridge and all that you have
represented. And your hearing touched a little bit on this, but
I was just wondering if you could shed some additional light--
and this could also be open to anybody, but I'm just feeling
like Oak Ridge might have some insight into this with
supercomputer technologies, public-private partnerships, and
also abilities to lead to the creation of jobs either at small
enterprises or large and what else we could be doing in
Congress to advance these opportunities.
Dr. Tourassi. So, as you know, building the supercomputers
represent the strong partnership with our vendors, and from
that point on, it's what kind of science we enable, right?
These are the scientific innovations that will lead to
technology transfer, and they will lead to also job growth
opportunities.
We talk a lot about artificial intelligence these days and
about what artificial intelligence will do to our workforce,
both positive and negative effects. And secondly, when I'm--I'm
an optimist and a proponent of developing a computational
savvy, aware, and ready workforce. And I see that all of the
leadership computing facilities being pivotal in that space as
well. It is not only how to use the computers, it is not only
how to do the science, but how exactly we develop the workforce
that can support the data infrastructure that we mentioned
earlier. And that is a different flavor of data--of the
workforce that is absolutely needed for our Nation. So I would
encourage all of you to keep thinking along those lines as
well. It takes a lot of effort to collect, curate, and manage
data. These are technical expertise. And I know that we don't
have enough in the Nation, so we should be building that
workforce and create opportunities for smaller companies to
play a role in that.
Ms. Stevens. Yes, and I--you know, we could go on here,
too, because part of why we had this NDEMC program, the
National Digital Engineering Manufacturing Consortium, was
particularly because of cost. And that was about 10 years ago
now, so I don't know if any of our other panelists while I have
about a minute left could shed some light on--in addition to
Dr. Tourassi's really great statements around workforce and
some of the barriers to entry because of workforce and human
capital, but any barriers to entry that we need to be
considering for small to midsize enterprises with supercomputer
technology and costs in the year 2021? I don't know if Dr.
Willcox or Dr. Monroe have any insights into that.
Go ahead, Dr. Monroe.
Dr. Willcox. Go ahead, Dr. Monroe.
Dr. Monroe. OK, yes. Yes.
Ms. Stevens. You unmuted first.
Dr. Monroe. OK. Yes, thank you for the question. Indeed,
so IonQ is a sub-100-person company notably about to go public
in the next month, and of course we're still building machines
and looking for use cases. And, as a small company, I think
it--you know, while this hearing is predominantly about high-
performance conventional computing, as it turns out, I think
was mentioned that you do need high-performance computers to
not only optimize and run but use quantum computers. They have
to be developed in parallel. And this is why actually we have--
we do have a relationship with Oak Ridge, a working
relationship for--they can run small algorithms on our
machines, and we in exchange can maybe use some of the advanced
computing machinery located in Tennessee.
So, indeed, it's important. These are capital-intensive
purchases to--you know, to be able to have your own array of
GPUs (graphics processing units) on your own site. We don't
have to do that all the time, but it's very important in my own
field that we're able to get access to very high-performance
computing machines, and we do rely on DOE for some of the----
Ms. Stevens. Right, you're relying on DOE and our labs and
maybe universities. Well, I'm out of time, but this was great.
Thank you, Mr. Chair, great hearing.
Staff. Mr. Feenstra----
Ms. Stevens. I yield back.
Staff. Oh, forgive me. Forgive me, Ms. Stevens. Mr.
Feenstra is next.
Mr. Feenstra. Well, thank you. Thank you, Chairman Bowman
and Ranking Member Weber. I want to thank you to the witnesses
for their testimony and in sharing their extensive research and
experience. It is truly outstanding, all the leaps that we're
getting into on the cutting-edge of this technology.
I want to direct this question to Dr. Tourassi, and thank
you for being here. I mean, it's just a pleasure to see you.
Accurate weather forecasting is incredibly important in my
district not only for the farmers in Iowa as they predict the
growing seasons and the harvest but also for covering severe
weather events for the public. Dr. Tourassi, in your testimony
you highlighted the use of Summit to achieve a milestone in
global weather forecasting simulations. Can you elaborate on
the milestone and how it will help in the future of weather
forecasting and how the advent of exascale supercomputing will
also be able to build on this?
Dr. Tourassi. Absolutely. And this is something that Dr.
Binkley addressed earlier on. Simulating modeling climate and
weather forecasting is one of the most complicated problems
because we deal with many different physical systems
interacting with each other. So--and this is something very
similar that happens with the biomedical and the biological
space. If you try to understand only one dimension of the
problem, you just miss the big picture. This is what high-
performance computing and supercomputing has done throughout
the years because it enables modeling and simulation and
weather forecasting by increasing the complexity of the
components we add in the equation. We can take into account
atmospheric patterns. We can take into account soil information
or ocean information, and that's how these models, with the
support of Summit and in the future with Frontier, will provide
models that are more detailed, and they are more precise. Now,
of course, it's an oxymoron in some ways to say we will do
weather forecasting in a more precise way because the
atmospheric system itself is a chaotic system. Still, though,
we're decreasing the uncertainty according with which we do
these predictions.
The way the community has been moving--and again, I'm
speaking a little bit at a high level because I'm a biomedical
scientist--we know that when you're dealing with a problem that
is very difficult to predict, effectively, you try to create a
number of models and just throw them in a pile and see what
they do. Think of it as an ensemble approach. And then we
aggregate the results of these models.
As the supercomputers are getting bigger and bigger, what
are we able to do? First of all, run more complex models.
Second, run more of them, therefore increasing the precision of
our predictions. And this is what we are experiencing with
Summit, and this is what we will see with Frontier even more.
Mr. Feenstra. Wow, thank you so much for that answer. I've
got one more question quickly. Dr. Binkley, as part of
Argonne's Early Science Program, a team from Iowa State and
Ames Lab is rewriting a software program called NWChem for the
exascale era. The new program could provide an increase in the
size of chemical systems that can develop new methods for
converting biomass into biofuels. This could be groundbreaking.
Can you talk about how exascale computers can be used for
breakthroughs such as in the many different areas including
biofuels?
Dr. Binkley. Well, certainly, the example that you're
citing, NWChem, which is one of the foremost quantum chemistry
codes, it's been around for a number of years, and it was
developed originally at the Pacific North--yeah, Pacific
Northwest National Laboratory, and, you know, making an effort
to bring that software set into operation on the exascale
computers is a major step, and it's something that our Basic
Energy Sciences Program has been supporting.
There are a number of other codes that I think are going
to have transformational effects. There--we--there are fusion
simulation models that are under development for use on the
exascale computers. There's also software for understanding
seismic events. You know, there is--if you look at the list of
the 24 exascale applications that are slated for first use, a
number of them I think are poised to have breakthroughs.
Mr. Feenstra. Well, thank you so much. And my time is up.
Thanks, everyone, for being on the panel. This is great
information. I yield back.
Staff. Mr. Casten is next.
Mr. Casten. Thank you so much. I really appreciate our
witnesses being here, and I have some questions for you, Mr.
Monroe. My district is just north of Argonne National Lab and
just to the east of Fermi, which means I can't claim any of it
but spent a lot of time at both. And of course that's where a
lot of the really groundbreaking deployment of quantum
computing is happening.
I want to first just be a little bit nerdy with you. When
you were listing some of the types of questions that you can
uniquely answer with quantum computing, I think you mentioned
optimization problems, encryption, decryption, some issues
around protein chemistry. All of these sort of classes--sort of
NP hard problems where the solution space is huge and you can
just jam through a lot of data, as you look at sort of the
opportunities, is it uniquely to those classes of problems that
just you can churn through a lot more data fields a lot more
quickly, or are there other classes of problems that you're
also excited about?
Dr. Monroe. Maybe I'll answer it in two ways. Thanks for
the question. I'm always glad to put the nerd hat on. So the
one disclaimer I'll say is that it's very hard to prove that a
quantum computer can find the absolute optimal configuration of
variables in some complex model. However, it could be a
heuristic, meaning that it could do better than any
conventional computer. You don't really need proof. All you
need to do is show, well, it gave me a better answer. It gave
me a shorter path----
Mr. Casten. That's correct.
Dr. Monroe [continuing]. Of somewhere out subject to
certain constraints. You don't need to prove it. And that's
sort of why quantum computers have to be deployed. You need to
build them and deploy them as soon as possible and sort of see
how well they perform.
So the other way I might answer this question--this might
be too universal--is that I think every application in a
quantum computer can be cast in terms of an optimizer.
Optimization is a very general thing. Even code-breaking
application, which is actually factoring numbers into their
primes, that's a very hard problem. It's easy to multiply, it's
hard to do the inverse, to factor. You can cast that in terms
of an optimization problem. It's optimizing the two numbers
that actually multiply to give the original number, and that
has, you know, revolutionary impacts on security.
But one thing I want to conclude here with is you
shouldn't think--I hesitate to describe quantum computers as
big data machines. They don't process big data in a way that
you can get access to all the big data. You can't compute
massive--you can't model every molecule in the atmosphere and
predict exactly what the weather is going to be. You need
problems that sort of have a very small--a very simple answer
at the end like there's a model of climate I have in the upper
atmosphere. I don't know how to solve it. I don't know what
conditions will validate this model. It's those problems that
quantum computers can do. They take lots of--they sample lots
of data, but you only get very--there's like a winnowing down
to getting only a very small amount of information at the end,
so it's a little bit subtle. And so high-performance computers
are one-to-one machines. They can compute brute force function
evaluation with all these inputs and all these outputs. Quantum
computers do something a little differently.
Mr. Casten. Fascinating. I'd love to follow up on that
with more time than we have, but let me leave with a question
to you but invite all of our witnesses to respond with your
thoughts. The--when I've gone down and toured these facilities
at our labs, I'm always struck by the fact that part of the
excitement is these NP hard huge data set problems for all
classes of supercomputing. And then the other conversation we
always end up getting into is the ethical questions around what
does it mean to have a computer that's capable of asking
questions that our brains aren't capable of thinking of and
what's the appropriate boundaries there? And I'd welcome any of
your thoughts on what we should be doing as a legislative body
to--you know, maybe we're doing enough already to put those
ethical boundaries in place and really understand how do we
make sure that--you know, that we don't--we don't hit whatever
that moment was in Terminator where the computers are smarter
than we are and we can't figure out why.
Dr. Monroe. Well, I'll answer very briefly there and leave
it to the other witnesses. This is a little bit passive, but I
think the only--I think there's an impetus for us to get there
no matter what because if we don't understand some
revolutionary form of computing and we don't get there, others
will. And so I think that that's some type of ethics. At least
we're at the forefront of that new technology. I agree with you
it can be vexing when you get a new technology, how to use it
ethically and so forth, but, you know, it's very important to
get there and not just ignore it and decide not to get there.
And if we do that, others will.
Mr. Casten. Well, thank you. I'm out of time but would
welcome any thoughts that the rest of the panelists have in
writing afterwards. Thank you, I yield back.
Staff. Mr. Obernolte is next.
Mr. Obernolte. Thank you very much, Mr. Chairman, and
thank you to our witnesses for a fascinating hearing. Also
thank you for allowing me to participate. I know I'm not a
regular Member of the Subcommittee, but this is a subject that
I find extremely interesting.
I think it's important to recognize just how much this
concept of beyond exascale computing has the potential to
change computer science and its contribution to humanity in
general. I mean, we're talking about computers that can
actually perform more calculations per second than the human
brain can, we can approach these artificial intelligence
problems in ways that we just hadn't been able to in the past,
so it's a very exciting thing to be part of, certainly
something that we as a Federal Government need to be
stimulating investment and research into.
So a question for Dr. Binkley. When you were talking in
your testimony about beyond exascale computing, you mentioned
that quantum computing is probably going to be the lead in
that. And my question to you is do you think that quantum
computing is the only technology that's capable of beyond
exascale power, or are there other technologies, traditional
technologies that might also be capable of that?
Dr. Binkley. Well, I don't think that quantum computing is
going to lead to the elimination of the type of computing that
can be achieved on current supercomputers, including exascale.
I mean, as Chris Monroe has pointed out, quantum computers are
very specialized in the type of calculations that they can do.
Conventional computers, including exascale, use numerical
methods and follow principles of sort of orderly input and
orderly output. And so I think that what is going to come next
after exascale, if you look at the typical exascale computer,
each node in the computer has both a conventional CPU (central
processing unit) chip and some type of a graphical processor.
And it's the combination of those two that allows one to attain
10 to the 18 floating-point operations per second.
Experiments are beginning now to look at other types of
computing elements that can be included into a supercomputer,
and so other types of GPUs, one could imagine neuromorphic
chips that could be incorporated. I think there's a lot of
research that can be done in those areas, and I think that the
lineage that has led us up to exascale still has more steps to
go through as we go forward.
Mr. Obernolte. Well, thank you, Dr. Binkley. As a computer
scientist myself, I find this extremely interesting and really
inspirational.
Before I go, if I could just make everyone aware of a
piece of legislation that I've introduced. It's H.R. 3284 that
would direct the Department of Energy to establish a program
for capabilities beyond exascale. And one thing that I think we
need to do more talking about is the need to research energy-
efficient computing because if you look at the current
generation of exaflop-capable supercomputers, one thing that
strikes me is the pure amount of power that they consume. The
El Capitan supercomputer that's being installed at Lawrence
Livermore Laboratories I believe is going to have a power
consumption on the order of 40 megawatts, so the amount of
energy they consume and the heat that they consume is going to
quickly become a barrier to our ability to employ these kinds
of technologies. So as part of this bill I'm suggesting that we
also establish a program to stimulate research into energy-
efficient computing.
And then of course I know the testimony was mentioned in
support for the Computational Science Graduate Fellowship
Program. That's a program that I am very supportive of. I think
that at the same time that we are stimulating research into
these technologies, we also need to make sure that our academic
population and our workforce is prepared to employ these
technologies. It's going to be no good to make the tools if
we're not also stimulating the kind of workforce that's going
to be able to help us take that to the next level. And this
bill would also do that. So I would certainly invite everyone's
support and participation on that bill.
And I'll yield back, Mr. Chair, but thank you very much
for letting me be part of this discussion.
Staff. Mr. Lamb is next.
Mr. Lamb. Thank you, Mr. Chairman. And I apologize up
front if I have any internet connection problems here. It
hasn't been the best day for me on that.
The first thing that I wanted to ask was for Mr. Monroe.
There's a statement on the Department of Energy's website that
says--that right now we're at the same point in quantum
computing that scientists in the 1950's were with computers. Do
you consider that an accurate statement as far as kind of
placing in context where we might be in the development of
quantum computing?
Dr. Monroe. Yes, thanks for the question. It's an
interesting comparison, and I buy into that with the caveat
that we're sort of on this experiential curve on progress and
so, you know, 10 years back in the 1950's is maybe like 2 years
now, so, you know, all of us students of history, it's
wonderful to see the progression from vacuum tubes to germanium
transistors to silicon and then silicon we built silicon out.
It took 20 or 30 years for that to happen. We don't expect that
to be the same with quantum. I think we are on the cusp of
several different technologies, superconducting circuits. At
IonQ we work with individual atoms. We know how to scale these
things. We have to put the, you know, engineering effort into
it.
So the comparison is a very good one because we are at
very low levels now. We're not--we don't have high--we don't
have Windows for quantum yet. You know, we're working with
individual gates, very low-level stuff that they were doing in
the 1950's with silicon transistors. But we also know that
software exists for classic computers, conventional computers
that we can also deploy and accelerate in quantum. So, yes,
it's 1950's in quantum, but, you know, we're sort of moving
much faster.
Mr. Lamb. Good, thank you. And you put in your testimony
that you thought the Endless Frontiers Act would be helpful to
our efforts to go even further in quantum computing. Could you
maybe just say a little bit more concretely why that is?
Because I'm a supporter of that bill. I think it's incredibly
important. I do think there is some hesitation for those
familiar with the NSF that it could somehow maybe detract from
the core basic science mission of the NSF, so would you be able
to address that and assuage those concerns?
Dr. Monroe. Yes, I--my understanding is there are many
different approaches to re-tasking the NSF to have a technology
edge to them. My understanding is also it's not a zero-sum
game, that the NSF will still have in its core mission the
ability to do research, blue-skies research, research for
research sake.
You know, I want to link that to technology, especially
quantum technology and why NSF is very well-suited to play a
big role in the development of the tech, not just the science,
and that is the future of quantum computing, we need
applications, and right now they're coming in the name of
science, mainly in universities. At Duke, in the University of
Maryland, we're deploying our laboratory systems to do models
of black holes and wormholes believe it or not, things that
maybe a company would never do. And it's those scientific
applications that are happening right now. They're also
happening at Department of Energy laboratories across the
country.
Companies won't pay for this. They're not going to build a
device to do black holes. So having an agency like the NSF that
has an eye on the blue-skies fundamental science research,
building and using quantum computers, making a user facility,
for instance, is a really good idea because it's going to tide
us over until industry really controls the building of these
devices. And then we can use those devices for new science just
like we're using exascale and high-performance computers at Oak
Ridge and other DOE labs for current science.
Mr. Lamb. I agree. I agree. Thank you.
I do want to just sneak in one more question. Dr.
Tourassi, thank you for everything you're doing, very excited
for the launch of Frontier this year. Is there in somewhat
simple layman's terms a way to describe a problem that we
cannot solve until we get Frontier? I understand that it'll
probably solve other problems faster. Are there new actual
types of problems that this platform will allow us to solve
that we couldn't before?
Dr. Tourassi. That's a good question. Actually, what we
are hoping that the next machines will enable is that pure
integration of modeling and simulation with large-scale AI,
bringing together models and observational data that we can get
from our different experimental facilities and from the
different Federal agencies. That level of computing has not
been--has not happened yet.
It goes back to the example that was given earlier about
the phenome-genome association studies, which actually are a
very challenging and attractive problem for exascale computing
when we're looking at large populations of, let's say, humans
when we want to do the--at the population level phenome-genome
associations, that is which will drive new treatments,
specialized treatments, precision medicine.
Mr. Lamb. Great, thank you. I'm out of time. Mr. Chairman,
I yield back.
Staff. Ms. Ross is next.
Ms. Ross. Thank you, Mr. Chairman, and thank you for
holding this hearing and to all the panelists for joining us
today.
The Committee Members have heard this before, but to the
panelists, I represent the Research Triangle area of North
Carolina. I represent Wake County. And the Research Triangle
area has a growing ecosystem of hundreds of innovative and
collaborative companies, including science and technology
firms, government agencies, academic institutions, including
Duke, startups and nonprofits. The IBM Quantum Hub at NC State
University is a cross-disciplinary center of quantum computing
education, research development, and implementation. The IBM
Quantum Hub at NC State works with researchers from Duke and
UNC (University of North Carolina) Chapel Hill to help partners
develop quantum teams and explore promising use cases and
promotion of quantum computing in real-world applications.
And so, Dr. Monroe, congratulations on the quantum
computing user facility at Duke. And I want to see how these
partnerships work. But there's a real-world application that
we've been dealing with these past couple weeks on
cybersecurity and how we might be able to collaborate between
research institutions and the private sector and use quantum
computing to maybe prevent some of the kinds of things that
we're seeing, particularly when we're dealing with 20th century
technology that hasn't moved into the 21st century. And I don't
know if the Committee has dealt with this earlier. I was in
another Committee meeting. But I'd love to hear from you about
how quantum computing might help us with cybersecurity.
Dr. Monroe. OK. Thank you for the question, Congresswoman.
So indeed the known killer application of quantum computers is
code-breaking, cracking the most popular data encryption
schemes we have. It looks like that application is one of the
hardest out there. We need much bigger machines to do that.
That said, we also are--this has been said before. We're at
such an early stage in the game. We know we can shrink the
problem, we can make it more efficient through software by
being more clever on how we structure not only the instructions
that we run the quantum computer but the quantum computer
itself, how we control it. And this is why at Duke, for
instance, we're collaborating quite--we're starting a very
large collaboration with the colleagues you mentioned at NC
State, and they're experts in computer architecture. We make
machines, and, you know, this is sort of the marriage that we
really like the direction of that.
Now, making code-breaking problems easier is something
that won't happen in isolation. It won't happen by me or, you
know, it won't happen by somebody that doesn't have a machine,
so I think they have to work in unison. And, you know, I've--at
IonQ we've worked closely with IBM. IBM also writes software
that supports our system, the expression of our particular
hardware. And, you know, IBM I think has really, you know,
taken a lead at educating the public on using systems. They
were the first to put their system on the cloud, and we've sort
of followed a few years later. So it's a wonderful ecosystem
down here in North Carolina, but, you know, also at the company
IonQ, which is a Duke-Maryland startup inside the beltway in
Washington, they--they're very close to the National Security
Agency and, you know, we--we're--we understand that community's
needs in this field.
So it's--I don't know what else to say. It's a very
exciting time for the field. We'll take problems anywhere they
come from, but security is a big one, and I think one--one last
comment, what's not mentioned so much is the energy grid.
That's a huge logistics problem, and I was horrified to find
out that apparently I think the way energy is coordinated
across the country, it's not--maybe it's not done very smartly.
It's not my field, I shouldn't comment on that, but we do know
that's one of the most vexing optimization problems we have,
and the news in the last few weeks, especially here in North
Carolina, is that, boy, it would be nice if we made that a
little smarter in the future. I'm not sure exactly how quantum
will play a role there, but if it's an optimization problem, we
want to take a crack at it.
Ms. Ross. Well, you anticipated my second question, and
that is how open are, you know, some of our big energy
companies to your help?
Dr. Monroe. Well, you know, at IonQ, we are--you know,
we're not a company that can afford to talk to 50 different
companies in parallel, but we do have conversations with a few
energy companies that have teams on the ground that are ready,
they want to deploy quantum for their uses. These could be oil
and gas or, you know, related to big pharma or, you know,
developing new fuels. It is a researchy time in the field right
now, but they're very open to working with companies like ours,
like IBM, and so forth and going in the future. So in terms of
the energy grid, that's definitely not my area of expertise,
yes, so, thank you.
Ms. Ross. Thank you, Mr. Chair. I've exceeded my time, and
I yield back.
Staff. Dr. Foster is next.
Mr. Foster. Thank you, and thanks to all of our witnesses
here.
I--well, first off, for those of you that don't know me,
I'm Congress's Ph.D. physicist and also the--Congress's AI
programmer, although I only have really dabbled in TensorFlow
recently.
The question I have was a more general one about where you
see this whole field is going of advanced computing. It seems
to me that it's likely to fragment, that traditionally we've
had a whole big line of, you know, megaflops, big floating
point pipelines to--for local partial differential equations,
and that will continue for all the usual reasons from, you
know, weather maps to nuclear weapons and everything in
between. And so that will continue perhaps with an overlay of
AI to do the grid scale optimization on the fly but largely
just, you know, a similar thing.
And then if you look at what's happening in AI, the state-
of-the-art AI engines are not in the Federal realm any more.
They are--you know, if you're just high-speed execution of GPT-
3 and other AI interesting algorithms, you know, the best
engines are from places like Google for that. And I don't
believe that the Federal Government is likely to be able to
keep up with that, that industry will continue to pull ahead.
In that case, the Federal investment will largely be in paying
for university groups to have access to these very advanced AI
engines, and that's probably as ambitious as we should be on
that front.
And then, you know, farther down the chain of, I guess,
narrower calculation of widths, you have things like
neuromorphic engines, which may end up as not floating-point or
even low-precision integer but single-bit computing engines for
neuromorphic thing--and so those will be experimental for a
while and may or may not become, you know, really important.
And so from a Federal point of view we have to decide
where to put--you know, there's a bipartisan agreement that we
should something like double the Federal budget, but we have to
understand, you know, whether to write, you know, a big new--a
check for a big new initiative wholly in National Science
Foundation, whether we should sort of scale up all of the
existing enterprises and then see what they're able to produce
in a few years. And a big part of that decision is whether we
think we have to--whether we're going to--what the mixture of
small research and big facilities is going to be because that--
anyway, so I'd be--I guess, Steve, we can start with you and
what you see as DOE's role in that.
Dr. Binkley. Well, clearly, Bill, the--our programs are
really intended to advance the state-of-the-art in scientific
computing writ large, and so that essentially encompasses all
of the technology paths that you just outlined. I mentioned
earlier that the--sort of the next obvious move post-exascale
is to look at heterogeneous computing systems where, in
addition to GPUs, one could have neuromorphic chips and other
computational technologies, and those types of machines would
provide avenues to solve essentially new problems.
But also I think continued investment in quantum computing
is going to be necessary. I don't--I do see a separation
between sort of the classical Turing-type computation and
quantum computation, and so there may--you know, that may
eventually divide--you know, there may be a parting of the
roads there. But at least at this stage in time I think we need
to really look at the diverse portfolio of technologies and not
foreclose any options.
Mr. Foster. Yes, well, are there areas where you sort of
believe we'll not be able to compete with industry? You know,
I'm thinking, as I mentioned, specifically of AI. Are there
areas there where it's clear already that the government is not
going to be leading?
Dr. Binkley. Well, when it--so there are certainly AI
applications that are going to be dominated by the high-tech
companies, Googles and so on. The one area, though, that I
don't see Google moving into is there is--with the advent of
the exascale machines and the type of GPUs they have, they are
actually the most powerful AI engines that exist. And the--you
know, this very, very high end of AI is something that I think
is going to still be the domain of the leadership computing
class facilities. You may disagree, but, I mean, I think----
Mr. Foster. No, I--you know, I'd be interested because
there's a different narrative coming out of Silicon Valley----
Dr. Binkley. Yes.
Mr. Foster [continuing]. And a lot of worry that small
startups in AI simply can't compete because they do not have
access to Googles and so on, Amazon's big AI engines so----
Dr. Binkley. Yes.
Mr. Foster. Now, does anyone else have a--Georgia or----
Dr. Tourassi. Yes, I would like to add a couple of data
points to make the case of why the supercomputers and exascale
computers will actually be the powerful AI engines, and I would
like to draw these experiences from the exascale computing
project and particularly the one that was mentioned earlier to
support cancer. When we started that effort, the specific
application we were focusing was not a traditional application
that anybody would have thought about supercomputers, but it
was a challenge application for us to start building frameworks
to prepare for the scientific community.
Well, in the process, we actually faced the philosophical
debate, do you fit hardware to algorithms or do you build
algorithms, modify algorithms to fit hardware? Typically, we
think of I have a specific application, I will prepare a
certain type of hardware to fit that particular application.
Well, in the case of a supercomputer in transit, we had to do
the opposite. And what we discovered in the end is that for
this particular application, by modifying the algorithm to make
the most of the supercomputer, we exceeded performance, state-
of-the-art performance. So that opened a completely new way of
thinking about using supercomputers for applications that
traditionally were not thought suitable to supercomputing.
The second example I would like to give was the very
recent Gordon Bell award for the COVID-19 pandemic that brought
together lots of modeling and simulation of the complete
envelope, viral envelope of 305, you know, million atoms, but
there was an AI workflow to accelerate that modeling and
simulation, so that coupling of AI and modeling and simulation
that was only possible on Summit. So I do believe that we see
plenty of examples out there to support that these general-
purpose machines will be able to deliver a lot even in the AI
space.
Mr. Foster. All right, fascinating discussion. I've
exceeded my time, so I guess I have to yield back.
Chairman Bowman. Thank you, everyone. Before we bring the
hearing to a close, I want to thank our witnesses for
testifying before the Committee today. The record will remain
open for 2 weeks for additional statements from the Members and
for any additional questions the Committee may ask of the
witnesses.
The witnesses are excused, and the hearing is now
adjourned. Thank you all. Have a good day.
[Whereupon, at 1:06 p.m., the Subcommittee was adjourned.]
Appendix I
----------
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
Appendix II
----------
Additional Material for the Record
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
[all]
File and source
- File
- CHRG_chrg-117hhrg44540_chrg-117hhrg44540.htm
- Size
- 139,403 bytes
- SHA-256
- cf579533ced82392cd086db9a24a8e2983b490d515d9eb40c559beb616d725ab
- Original
- No public link identified.