[Senate Hearing 118-404]
[From the U.S. Government Publishing Office]
S. Hrg. 118-404
THE DEPARTMENT OF ENERGY'S LEAD ROLE
IN CONDUCTING ADVANCED COMPUTING
RESEARCH, APPLICATION, AND SECURITY
=======================================================================
HEARING
BEFORE THE
COMMITTEE ON
ENERGY AND NATURAL RESOURCES
UNITED STATES SENATE
ONE HUNDRED EIGHTEENTH CONGRESS
SECOND SESSION
__________
SEPTEMBER 12, 2024
__________
Printed for the use of the
Committee on Energy and Natural Resources
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
Available via the World Wide Web: http://www.govinfo.gov
__________
U.S. GOVERNMENT PUBLISHING OFFICE
56-727 WASHINGTON : 2025
-----------------------------------------------------------------------------------
COMMITTEE ON ENERGY AND NATURAL RESOURCES
JOE MANCHIN III, West Virginia, Chairman
RON WYDEN, Oregon JOHN BARRASSO, Wyoming
MARIA CANTWELL, Washington JAMES E. RISCH, Idaho
BERNARD SANDERS, Vermont MIKE LEE, Utah
MARTIN HEINRICH, New Mexico STEVE DAINES, Montana
MAZIE K. HIRONO, Hawaii LISA MURKOWSKI, Alaska
ANGUS S. KING, JR., Maine JOHN HOEVEN, North Dakota
CATHERINE CORTEZ MASTO, Nevada BILL CASSIDY, Louisiana
JOHN W. HICKENLOOPER, Colorado CINDY HYDE-SMITH, Mississippi
ALEX PADILLA, California JOSH HAWLEY, Missouri
Renae Black, Staff Director
Sam E. Fowler, Chief Counsel
Sarah Kessel, Professional Staff Member
Alyse Huffman, Professional Staff Member
Justin J. Memmott, Republican Staff Director
Patrick J. McCormick III, Republican Chief Counsel
Derek Fisher, Republican Professional Staff Member
Chad Thorley, Republican Director of Oversight
C O N T E N T S
----------
OPENING STATEMENTS
Page
Manchin III, Hon. Joe, Chairman and a U.S. Senator from West
Virginia....................................................... 1
Barrasso, Hon. John, Ranking Member and a U.S. Senator from
Wyoming........................................................ 9
Durbin, Hon. Richard J., a U.S. Senator from Illinois............ 42
WITNESSES
Fu, Helena, Director, Office of Critical and Emerging
Technologies, U.S. Department of Energy........................ 44
Gleason, Dr. Shaun, Director of Science-Security Initiative
Integration, Office of the Laboratory Director, Oak Ridge
National Laboratory............................................ 53
Kaushik, Dr. Divyansh, Senior Fellow, American Policy Ventures... 63
ALPHABETICAL LISTING AND APPENDIX MATERIAL SUBMITTED
Abrams, Elliott et al.:
Letter for the Record........................................ 293
Barrasso, Hon. John:
Opening Statement............................................ 9
Strider Technologies Report entitled ``The Los Alamos Club:
How the People's Republic of China Recruited Leading
Scientists From Los Alamos National Laboratory To Advance
Its Military Programs'' published in 2022.................. 10
Hoover Institution Report entitled ``Global Engagement:
Rethinking Risk in the Research Enterprise'' Hoover
Institution Press, 2020.................................... 80
Chicago Quantum Exchange:
Letter for the Record........................................ 270
ColdQuanta, Inc. et al.:
Letter for the Record........................................ 272
Durbin, Hon. Richard J.:
Opening Statement............................................ 42
Energy Sciences Coalition:
Statement in support of the DOE Quantum Leadership Act....... 266
Statement in support of the DOE Artificial Intelligence Act.. 300
Fu, Helena:
Opening Statement............................................ 44
Written Testimony............................................ 46
Responses to Questions for the Record........................ 277
Gleason, Dr. Shaun:
Opening Statement............................................ 53
Written Testimony............................................ 55
Responses to Questions for the Record........................ 284
Kaushik, Dr. Divyansh:
Opening Statement............................................ 63
Written Testimony............................................ 65
Responses to Questions for the Record........................ 286
Manchin, Hon. Joe:
Opening Statement............................................ 1
Chart ranking countries by percentage of quality research
papers in the fields of artificial intelligence, advanced
data analytics, quantum computing, and more................ 3
Letter of support for the FASST Act, signed by AMD, Arm,
Hewlett Packard Enterprise, Intel Corporation, and Micron
Technology................................................. 6
Montana Chamber of Commerce:
Letter for the Record........................................ 264
Montana Photonics and Quantum:
Letter for the Record........................................ 263
Montana State University:
Letter for the Record........................................ 265
Quantum Industry Coalition:
Letter for the Record........................................ 269
University of Chicago:
Letter for the Record........................................ 271
THE DEPARTMENT OF ENERGY'S LEAD ROLE
IN CONDUCTING ADVANCED COMPUTING
RESEARCH, APPLICATION, AND SECURITY
----------
THURSDAY, SEPTEMBER, 12, 2024
U.S. Senate,
Committee on Energy and Natural Resources,
Washington, DC.
The Committee met, pursuant to notice, at 10:00 a.m. in
Room SD-366, Dirksen Senate Office Building, Hon. Joe Manchin
III, Chairman of the Committee, presiding.
OPENING STATEMENT OF HON. JOE MANCHIN III,
U.S. SENATOR FROM WEST VIRGINIA
The Chairman. The Committee will come to order.
First of all, we would like to welcome our friend and
colleague, Senator Durbin, for being here in our Committee and
we appreciate very much having you.
Before we begin, I want to acknowledge the terrible loss of
former Director of Los Alamos National Laboratory, Charles
McMillan, who died in a car accident last week. He was going to
be here to testify for us today and we are very sorry for his
family and we express our deepest condolences. Charlie was a
true patriot to this country who made extraordinary
contributions to our nuclear weapons programs and other
critical scientific missions at our national labs for over 40
years.
This morning we are here to discuss where we are and where
we are going in the field of advanced computing, which touches
on a wide range of technologies and applications across every
one of our lives. We will also discuss legislation our members
have introduced in three specific areas: artificial
intelligence, quantum computing, and cybersecurity. Computing
technology is advancing rapidly across the globe, and America
must proceed with both ambition and caution, particularly when
it comes to our national security and public safety. The DOE
has a long legacy in computational science, dating back to the
Manhattan Project in the 1940s, which relied on computational
modeling.
Today, the labs currently operate the fastest two
supercomputers in the entire world. The opportunities that
high-performance computing provides are endless. In my State of
West Virginia, it was recently announced that a new
supercomputer, the Rhea, will be installed in the city of
Fairmont to improve drought, flood, and wildfire predictions
and forecasting. Like during the Manhattan Project, we are now
engaged in a new kind of technological race, one that requires
us to innovate with similar urgencies and vigilance. We are
watching our adversaries ramp up their investments in advanced
computing technologies and even deploy them in battlefields
across the world.
The chart behind me, and I am going to keep this chart up
too, but I want to explain this chart.
[The chart referred to follows:]
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
The Chairman It is really amazing. This is not in dollars
and cents, so forget about the amount of money, because we
don't really have that. What we can tell you is, it has been
scientifically proven that this is the authentic things that
they have done, and we know that China, in artificial
intelligence, has 37 percent more capabilities and authenticity
on what they have been able to report than what we do, at 13.
It shows you how far we are behind. Advanced analytics--31
percent for China, 15 percent U.S. Quantum computing--we are
still ahead 34 to 15. Machine learning--China. Cybersecurity
technology--China. High-performance computing--we are still
ahead, barely, but we are there. And then when it comes down to
quantum communications, it shows you the difference.
This is where we are, and we know this has been proven and
this is where they are at as far as their capacity and
authenticity of what they are doing. Would you all agree on
that? Okay. So we will keep that one up. Put that chart up over
there.
We are using published research paper data in place of
spending because the U.S. Government does not trust how China
is reporting their expenditures. As the chart showed, America
is at real risk of falling behind to China in this race, and it
is becoming increasingly apparent that whoever leads in the
development of these technologies will secure the unequivocal
lead in scientific and technology innovation writ large. For
the United States to maintain our position as a global leader,
we must accelerate our efforts. The stakes are nothing short of
economic prosperity and national security. Our national
laboratories have spent decades building a workforce and
infrastructure to answer the challenging questions about how to
safely deploy these emerging technologies in a way that sets an
example for the rest of the world.
While we are proud that our private sector is making
incredible strides developing and deploying advanced computing,
this does not replace the need for government research and
development. Complex societal challenges, like advanced
manufacturing, nuclear security, and genomics are a few areas
of application where there is not yet an established commercial
market. And this is why Senator Murkowski and I have introduced
bipartisan legislation to reinforce the artificial intelligence
research and development programs at our DOE labs. The bill
authorizes the Frontiers in Artificial Intelligence for
Science, Security and Technology, or the FASST Act, as an
initiative at the Department of Energy which will give the
United States the tools to deliver secure and dependable AI
solutions.
This bill would create AI research and innovation hubs at
our labs that will harness testbeds for the development of AI
platforms, develop foundational models for various applications
for energy and national security, verify the safety of new
large language models, and establish a risk evaluation program
to respond to security risks. It is just common sense to use
our nation's brain trust at our national labs and their state-
of-the-art facilities that we have already built and maintain
to safely advance AI while safeguarding taxpayer dollars.
Industry and the science community alike have publicly
supported our bill. I would like to submit, for the record, the
supportive statement of Hewlett Packard Enterprise, AMD, Intel,
ARM, Micron, and the Energy Sciences Coalition. Without
objection, so be it.
[Letter of support for the FASST ACT follows:]
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
The Chairman. Let me also reiterate that America will need
more energy to meet the growing demand from data centers and
the manufacturing resurgence that has resulted from the
Bipartisan Infrastructure Law, CHIPS and Science Act, and
Inflation Reduction Act. For decades, power demand has been
decreasing, but now we are expecting a rapid turnaround this
decade. But while demand is increasing, we have 2.6 million
megawatts of generation waiting an average of five years to
connect to the grid, and we are also retiring baseload and
dispatchable generation faster than we can replace it. This is
unacceptable. If America can't build the energy infrastructure
needed to support high-tech industries, companies will choose
to take their business elsewhere. We simply must get common-
sense policy like our bipartisan energy permitting bill enacted
or we will have squandered this opportunity and really put
ourselves at risk.
Today, we will be discussing quantum computing, which
processes information much more quickly and efficiently than
even our fastest supercomputers do today. Senator Durbin and
Senator Daines recently introduced bipartisan legislation to
reauthorize many quantum programs throughout the Department of
Energy, and Senator Durbin is going to be joining us to speak
briefly about his bill.
We can't discuss this new era of emerging technology
development without considering cybersecurity and broader
national security implications. These technologies can serve as
assets, but also as threats to the cybersecurity posture of the
United States. We have seen the devastating effects of a
cyberattack on our critical infrastructure, like the Colonial
Pipeline attack in 2021 that forced the shutdown of the
country's most important fuel pipeline.
Last year, I joined Senator Risch in introducing the ETAC
Establishment Act, which establishes the Energy Threat Analysis
Center at DOE. ETAC will serve as the energy sector centralized
hub in the Federal Government for cyber information, sharing
threat response to better defend the U.S. energy sector against
cyber threats. We must also pay attention to how we are
protecting our scientific program from nefarious actors. We
will not outcompete China if they are able to just steal the
technology funded by our taxpayer dollars. The CHIPS and
Science Act authorized improvements to our research security
policies that are already being implemented by the Department.
But as the threat environment becomes more complex and stakes
become higher, much more must be done. During the past several
months, Senator Barrasso and I have been working with our
colleagues on the Intelligence Committee to strengthen our
research security policies to secure the science performed in
our national labs while not stifling their work or closing
ourselves off to the global scientific community.
I am confident this is something we can get done in this
year's National Defense Authorization Act. We have a lot to
cover this morning. I am looking forward to hearing our
witnesses' perspectives on specific steps we can take to ensure
America is advancing all these technologies in a competitive
and a responsible manner. I know we certainly do not need to
start from scratch to achieve this. Instead of duplication, we
can invest smartly in emerging technologies in a cost-effective
way by building upon the resources that we already have at the
DOE and its national labs. This is how we will maintain our
global lead in scientific innovation.
And with that, I am going to turn to my colleague and
friend, Senator Barrasso, for his opening remarks.
OPENING STATEMENT OF HON. JOHN BARRASSO,
U.S. SENATOR FROM WYOMING
Senator Barrasso. Well said, Mr. Chairman. Thank you for
holding today's very important hearing.
Research into advanced computing is critical--critical to
maintaining America's economic growth, our national security,
and our leadership in the world. The Department of Energy,
through its network of 17 national labs, plays a very key role
in all of those innovations. The Department has some of the
most advanced computing systems in the world. In fact, the
Department has the world's two fastest supercomputers and a
third supercomputer among the world's top ten. These systems
have pioneered advances in artificial intelligence and in
quantum computing. These are two fields that the People's
Republic of China does seek to dominate. For this reason, China
is watching nearly every move that our national labs make. Our
labs are under constant surveillance by a branch of China's
intelligence network that focuses on science and technology.
This branch alone consists of about 100,000 agents. Beginning
under Chairman Mao, this intelligence network has supported the
development of China's nuclear weapons and its missile and
satellite programs. And its mission remains the same today--
target foreign technologies useful to the Chinese communist
regime and acquire them by any means possible.
America's open research environment is the envy of the
world. It has fostered our greatest scientific achievements,
yet it is a rich target for China and other adversaries. As
stated by the National Academy of Sciences, ``The integrity of
research is based on the values of objectivity, honesty,
openness, fairness, accountability, and stewardship.'' Contrast
this with the view of China's President, Xi Jinping. He
recently described science and technology development as a
contest to be won. He stated, ``The initiatives of innovation
and development must be secretly kept in our own hands, and
whoever holds the key to innovation makes an offensive move,''
he said, ``in this chess game, and will be able to take the
lead and win the advantage.''
A 2022 report, titled ``Los Alamos Club,'' by Strider
Technologies, is telling. I have a copy of the report here, Mr.
Chairman.
[The report referred to follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Senator Barrasso. It says between 1987 and 2021, the
Chinese Communist Party targeted over 160 Chinese nationals
working at Los Alamos National Laboratory. Upon returning to
China, these researchers helped them advance key military
technologies using knowledge financed by us, by the American
taxpayers. Today, thousands of non-U.S resident Chinese
nationals still work at our national labs, and I believe the
majority of these foreign nationals strive to further
scientific innovation and collaborate in good faith. Make no
mistake, they are beholden to an authoritarian regime, and the
Chinese Communist Party is ruthless. Some of these Chinese
nationals will see no other choice but to support the Chinese
Communist Party through theft of American research and
technology because if they don't comply, their families back in
China may be punished.
Others will be tempted through bribery. Earlier this year,
a Chinese national and former software engineer at Google was
arrested for stealing on behalf of a Chinese firm which was
paying him secretly. The U.S. Justice Department has charged
this individual with stealing software used to orchestrate
Google's supercomputers at the cutting edge of machine learning
and AI technology.
In 2020, Congress required the Department of Energy to
devise a study of counterintelligence efforts at our national
labs. The Department hired MITRE, a government contractor, to
conduct this study. In April 2023, MITRE produced an
unclassified report. Upon receiving the report, the Secretary
of Energy then decided to classify it. The Secretary reassigned
the Director of the Department's Office of Intelligence and
Counterintelligence without an explanation. I have asked the
Department to declassify the MITRE report and for the
Department to come clean with the American people. The U.S.
Department of Energy has refused. One can draw many different
conclusions from the Department's stonewalling. The Department
may simply want to hide its failures from the public, but
whatever the reason, it is clear that the Department of Energy
and our national labs have failed to take the China threat
seriously.
Mr. Chairman, we can't let our research and technology fall
into the hands of China's brutal dictatorship. The Department
must dramatically increase its efforts to protect our research
from our adversaries, and Congress must step in if the
Department fails to do its job.
Thank you again, Mr. Chairman, for calling this important
hearing, and I look forward to today's testimony.
The Chairman. Thank you, Senator.
Now we will turn to our friend and colleague, Senator
Durbin.
OPENING STATEMENT OF HON. RICHARD J. DURBIN,
U.S. SENATOR FROM ILLINOIS
Senator Durbin. Do you want to go to Senator Cortez Masto
first?
The Chairman. No, we will go after you, then we are going
to go to the witnesses.
Senator Durbin. I am going to be very brief.
Thank you, Chairman Manchin and Ranking Member Barrasso for
this opportunity, and to all the members of the Committee.
I won't sit before you today and pretend I am Nobel Prize
material for science or engineering. I am just a liberal arts
lawyer. However, there are brilliant scientists in my home
State of Illinois that have given me a crash course in quantum
computing. I have visited Illinois' two national labs--Argonne
and Fermi--many times and seen their extraordinary work. And
this past summer, Argonne's Aurora supercomputer achieved
exascale computing speeds, landing at the top spot among the
most powerful supercomputers in the world. Achievements like
this are why, more than a decade ago, I founded the Senate
National Labs Caucus, along with Senator Risch. It's why I
worked with so many of you on this Committee to support the
Department of Energy's Office of Science. The fact is, when
America invests in science, we lead the world, and nowhere is
this truer than in advanced computing.
I want to say, parenthetically, thank you to Senators
Manchin and Barrasso for talking about competition with China.
It should be a focal point every single day for all of us. I
think for a moment, and I know Senator Barrasso is an amateur
historian himself--it was in my lifetime that ping-pong
diplomacy took place. This primitive, backward, oxen-driven
economy in China in 50 years emerged as a world competitor to
the United States of America. And that is why we are meeting
today to discuss it.
Senator Daines and I have introduced a bill called the
Department of Energy Quantum Leadership Act, reauthorizing
quantum research and development activities across DOE,
expanding DOE quantum research centers, tackling supply chain
and prototype challenges in the private sector, and expanding
training programs for the quantum workforce. This is timely,
it's important, and it's before this Committee. I urge you to
enact it as quickly as possible. Let's get it passed.
The Chairman. Thank you, Senator, so much. And I would say
to all of our members that we would, as a Committee, share this
and keep updating it to show every legislative member--535--
exactly where we stand. And if we don't start acting, and
acting in unity, we are not going to be able to catch up or
excel. So I think this is a stark reminder of where we are and
this is very accurate, very factual. I think you all would
agree that we pulled these out of what you all have looked at,
and said this is where they are and this is where we are.
So we want people to understand that. So we will be sharing
that without any objections from the Committee, and with that,
Senator, we appreciate you coming in and sharing your thoughts.
Senator Durbin. Thank you.
The Chairman. At this time, we are going to turn to our
panel of witnesses.
We have Ms. Helena Fu, Director of the Office of Critical
and Emerging Technologies at the Department of Energy.
We have Dr. Shaun Gleason, who is Director of Science-
Security Initiative Integration at Oak Ridge National
Laboratory.
We have Dr. Kaushik, Senior Fellow, American Policy
Ventures.
And with that, we will start with Ms. Fu.
STATEMENT OF HELENA FU, DIRECTOR, U.S. DEPARTMENT OF ENERGY,
OFFICE OF CRITICAL AND EMERGING TECHNOLOGIES
Ms. Fu. Thank you so much.
Chairman Manchin, Ranking Member Barrasso, distinguished
members of the Committee, thank you for the opportunity to
testify about the Department of Energy's leadership in advanced
computing research and its application and cybersecurity. I
want to start by thanking this Committee for your strong
support of DOE for many years. My name is Helena Fu. I serve as
the Director of the DOE Office of Critical and Emerging
Technologies. My office coordinates across the Department of
Energy and its 17 national laboratories in artificial
intelligence, biotechnology, microelectronics, and quantum
information science.
Since its origins from the Manhattan Project, the entire
Department of Energy complex has been working at the frontier
of science, driving advances central to America's prosperity
and security. Two of the most critical scientific frontiers
that we currently face are artificial intelligence and quantum
information science. And today, we all see the transformative
potential of AI. DOE and its national labs have long invested
in AI and its applications, and in addition, DOE is the leading
generator of classified and unclassified scientific data
through the world's largest collection of scientific
experimental facilities. We build and operate, in partnership
with industry, the world's fastest and most powerful
supercomputers, that are both strategic assets and serve the
scientific community. Our proposed Frontiers in AI for Science,
Security and Technology initiative, or FASST, seeks to harness
this infrastructure at DOE to deliver a step change in
capability for the nation, to develop AI-ready data, and to
advance the next generation of frontier-scale computing
platforms. Building on this data and compute, we need to
develop models that deeply understand science, math, physics,
and chemistry, and we need to apply these models to solve our
most pressing challenges in discovery science, in applied
energy, and in national security.
Beyond AI, quantum could help unlock new forms of computing
and information processing. And the National Quantum
Information Act authorized DOE's five national QIS research
centers. DOE is making strides in the science that could help
unlock quantum's potential in computing, in simulation, in
networking, and in sensing. We have created a first-of-a-kind
quantum computing user access program, created testbeds and
foundries, and built underground facilities to characterize
devices. These investments are helping to build up the quantum
ecosystem in the United States, where we are working across 115
institutions, 24 states, with trusted international partners
and with other parts of the interagency. DOE is also making
strategic investments in quantum computing to address nuclear
security challenges, and we are exploring potential
applications to our energy mission.
A resilient and secure power grid underpins and enables
U.S. leadership in AI and quantum, and DOE is continuing to
strengthen the energy sector's cyber defenses and invest in new
capabilities, such as the Energy Threat Analysis Center, or
ETAC. The ETAC pilot brings experts from government and from
industry together to address the growing cyber threats to U.S.
electricity, oil, and natural gas systems. The ETAC pilot has
been instrumental in rapidly addressing cyber threats, such as
the PRC-sponsored Volt Typhoon activity.
On a solemn note, I would like to recognize the life of Dr.
Charlie McMillan, who passed away unexpectedly last week. For
35 years, Charlie worked at Lawrence Livermore National
Laboratory and as Director of Los Alamos National Laboratory.
He had recently come out of retirement to work with me and our
labs on our AI initiative because he saw how important this
was. We in the entire DOE community feel this loss keenly, and
our hearts go out to Charlie's family and friends.
We are at an inflection point in AI and in quantum and
cyber. The DOE, with its dedicated scientific workforce of
40,000 strong, our ability to drive mission science through
deep partnerships, and the ability to work across the entire
ecosystem, from discovery science to applied energy to national
security--we stand ready to do our part. I want to thank the
Committee for its ongoing and bipartisan support for the DOE
mission, and we look forward to working with all of you. I am
happy to answer your questions.
[The prepared statement of Ms. Fu follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
The Chairman. Thank you, Ms. Fu.
And now we have Dr. Gleason.
STATEMENT OF DR. SHAUN GLEASON, DIRECTOR OF SCIENCE-SECURITY
INITIATIVE INTEGRATION, OAK RIDGE NATIONAL LABORATORY, OFFICE
OF THE LABORATORY DIRECTOR
Dr. Gleason. Chairman Manchin, Ranking Member Barrasso, and
members of the Committee, thank you for the opportunity to
speak with you today. My name is Shaun Gleason. I am currently
the Director of the Science-Security Initiative Integration at
the U.S. Department of Energy's Oak Ridge National Laboratory
(ORNL) in Oak Ridge, Tennessee, where I have 35 years of
service. I serve as a liaison between the open science and
national security mission communities and I am specifically
focused on emerging technologies, such as artificial
intelligence, quantum science, cybersecurity, and high-
performance computing across the diverse mission areas of ORNL.
I am also an entrepreneur who founded a startup company that
successfully transitioned a medical imaging technology to the
market.
There are grand challenges where we must accelerate
progress to ensure continued U.S. leadership in the emerging
technologies of quantum, cyber, and AI. For example, in AI,
energy efficiency is a grand challenge, as many are predicting
that energy use by AI-driven data centers will approach ten
percent of U.S. energy demand by 2030. While energy-efficient
AI is crucial, we also need our AI systems to be safe, secure,
and trustworthy. For quantum, a primary grand challenge is the
ability to create reliable and affordable quantum devices that
are the building blocks for quantum computers. Another grand
challenge is creating a quantum internet that can reliably and
securely share quantum information with many different devices
over long distances. Cybersecurity is a grand challenge arms
race where every defensive move inspires an adversary's
offensive move and vice versa. Protecting U.S. critical
infrastructure, such as the electric grid, from cyberattacks
requires regional, public-private partnerships and real-world
cyber testbeds that are connected to a national information
coordination network.
Some of the most exciting and critical areas for
revolutionary innovations are where the fields of AI,
cybersecurity, and quantum intersect with one another. In the
fields of cybersecurity and AI, AI is being used to create
dynamic, self-learning cyber-defense tools that can adapt to
the rapidly changing cyberattacks against our nation's
infrastructure. AI systems themselves are uniquely vulnerable
to a variety of cyberattacks that can manipulate AI into making
decisions favorable to an adversary. Adversaries can create
public websites filled with disinformation that are
automatically scraped for data that can bias an AI model during
training. To combat cyberattacks on AI systems, several of the
DOE national labs, including ORNL, Los Alamos, Pacific
Northwest, and Lawrence Livermore, have created internal
organizations to develop cybersecurity methods specifically for
the defense of AI systems.
The discoveries being made at the intersection of AI and
quantum are also powerful, and the scientific community is only
scratching the surface. For example, AI is being leveraged to
accelerate the development of quantum computers by accelerating
the discovery of new quantum materials and generating new types
of error-correcting codes for quantum computers. Conversely,
quantum computers are being used to speed up AI model training
algorithms and to generate realistic simulated data that can be
used to train data-hungry AI models. Additional research is
needed to capitalize on these opportunities.
DOE's national laboratories are also home to powerful
experimental facilities that are made available to
universities, industry, and other government agencies to
conduct cutting-edge research. Thanks to investments by
Congress through DOE's Office of Science and the National
Nuclear Security Administration, the national labs have
deployed the first open science exascale computers--Frontier at
Oak Ridge National Lab, Aurora at Argonne, and later this year,
El Capitan at Lawrence Livermore.
The combination of world-class talent, computing, and
experimental facilities positions DOE to lead in AI, quantum,
and cybersecurity research. Exemplifying the value of
partnerships in quantum, ORNL is currently performing quantum-
based secure communication experiments in collaboration with
the electric power board in Chattanooga and the University of
Tennessee in Chattanooga on their commercial quantum network.
To maintain U.S. leadership in AI, public-private partnerships
are also critical. For example, ORNL recently partnered with
Advanced Micro Devices and Microsoft to develop software that
can train a one trillion parameter AI model on the Frontier
supercomputer.
In summary, the grand challenges before us motivate
partnerships across the government, the DOE national laboratory
system, industry, and academia to accelerate the pace of
innovation. DOE has demonstrated its commitment to advancing
research and technology transitions across all three of these
important areas while trying to balance associated risks. Thank
you once again for the opportunity to testify, and I welcome
any questions you have on these important topics.
[The prepared statement of Dr. Gleason follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
The Chairman. Thank you, Doctor.
And now we have Dr. Kaushik.
STATEMENT OF DR. DIVYANSH KAUSHIK,
SENIOR FELLOW, AMERICAN POLICY VENTURES
Dr. Kaushik. Chairman Manchin, Ranking Member Barrasso, and
distinguished members of the Committee, thank you for the
opportunity to testify today.
We are at a critical juncture in technological history. The
People's Republic of China, the PRC, is in an intense
competition with the United States, aiming to dominate advanced
computing and AI by 2030. Over the past seven years, they have
increased their R&D budgets by ten percent annually while
engaging in sophisticated espionage efforts to acquire
technology. This is not just economic competition, it is a
strategic effort to reshape the global order. The Department of
Energy and its network of 17 national laboratories and 35 user
facilities are our technological vanguard. These institutions
have consistently delivered innovations underpinning American
leadership, from clean rooms to platforms driving key
industries today. The DOE labs' potential to drive AI
innovation is enormous, with applications for national
security, energy, and scientific advancement. These
breakthroughs powered by DOE's supercomputing capabilities
could revolutionize areas like material sciences, molecular
dynamics, and power grid resilience.
This necessitates a multifaceted approach. We must enhance
the security of our national lab facilities. We must maximize
their operational and scientific capabilities and we must
implement strategic initiatives to attract and retain the
talent that wants to come here while removing those who seek to
exploit our system. We must foster an environment where our
scientists and engineers are encouraged to think big, push the
boundaries of innovation, and be confident that their
groundbreaking research is safeguarded against foreign
exploitation.
The PRC has explicitly held up U.S. national laboratories
as models to emulate in their five-year plans. Stating in their
13th five-year plan, and I quote: ``National laboratories have
become key platforms for major developed countries to seize the
high ground in technological innovation. For instance, the
Argonne, Los Alamos, and Lawrence Berkeley National
Laboratories in the United States are all research bases
focused on national missions. It is urgently needed to focus on
national goals and strategic needs, target international
technological frontiers, and establish a group of larger-scale
interdisciplinary and integrative national laboratories.'' This
is a direct quote from their five-year plan. Xi Jinping has
since announced the creation of their own national laboratories
that are explicitly designed to mimic and ultimately surpass
our DOE complex. While imitation may be the sincerest form of
flattery, in this context, it serves as a stark reminder of the
value and the vulnerability of our national laboratories. The
PRC's efforts to surpass our labs and capitalize on their
advancements highlights the pressing need for a comprehensive
approach to research security that preserves our competitive
edge.
The bottom line is that the PRC is pursuing an aggressive
campaign of technological advancement that relies on both
aggressively investing in their own R&D ecosystem and illicitly
acquiring intellectual property. Furthermore, their strategy
integrates civilian research with military applications, as
evidenced by statements from top Chinese academic institutions,
including Tsinghua University, commonly referred to as their
MIT. This goes beyond normal competition. It represents a
coordinated effort to challenge America's innovation leadership
and reshape the global technological landscape in Beijing's
authoritarian image.
The PRC's legal framework further complicates this picture.
Laws such as the 2017 National Intelligence Law compel Chinese
citizens and organizations to ``support, assist, and cooperate
with state intelligence work.'' This means that even well-
intentioned PRC researchers may be legally obligated to share
information with their government, regardless of any
commitments made to U.S. institutions. Let me be clear,
research security is not about stifling innovation or closing
our doors to the world. It is not about--and cannot be about--
targeting individuals based on their ethnicity. It is simply
about ensuring that our openness is not exploited to our
detriment by our adversaries. The DOE's national laboratories
have long been a shining beacon for scientists and researchers
worldwide. This is an asymmetric advantage that we have.
This global appeal is not just a point of pride, it is a
cornerstone of our technological leadership. For instance, 59
percent of top-tier AI researchers work in the United States,
but only about 20 percent of them got their undergraduate
degrees here. Our nation's commitment to freedom, to free
speech, to freedom of inquiry, innovation, and scientific
excellence has made us the destination of choice for the
world's brightest minds. We can and must maintain our
leadership in scientific collaboration, but on terms that
protect our national interests. We passed several laws.
Effective implementation of these remains key, whether it be
the research security provisions in CHIPS or in NSPM-33.
The path that I have laid out today is undoubtedly
challenging and will require sustained commitment and vigilant
oversight. However, the alternative--a world where the PRC
dictates the rules for transformative technologies--is simply
not an option that we can entertain. There is much more that I
could say on these matters, but I trust that we will cover them
more fully over the course of this hearing. Thank you for the
opportunity to testify. I look forward to your questions.
[The prepared statement of Dr. Kaushik follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
The Chairman. Thank you.
Now we are going to begin our questions, and I will begin.
This is for all three of you to think about, but we, as
Senators--Democrats and Republicans--are looking at, whether it
be the CHIPS Act and how we divide our money up and this and
that. I don't want to reinvent the wheel. Okay? And I don't
want to split the baby. The bottom line is, you do what you do,
and NSF should do what they do. And for some reason, whoever
has what in their state--I would like to make sure that we are
not making that mistake and going down a path where you don't
have the ability to make up the differences where we are
lagging behind right now.
So if you can--and this isn't disparaging anything in NSF.
They have so much expertise in certain areas, but in the CHIPS
Act, you know, what we have done there, you already have that
computing expertise, you already have the investment in the
supercomputers and quantum computing and all that. Why would we
basically try to reinvent the baby again--you know, the wheel,
if you will? So would you all speak in comparison of what you
think each one of you all do with our labs versus NSF and how
we can direct more of our attention to make sure that both of
you can meet your full potential?
You want to start, Ms. Fu?
Ms. Fu. Sure, I am happy to speak on this because there has
been a lot of discussion on the AI policy.
The Chairman. Does it make sense, what I am asking you?
Ms. Fu. Of course, yes. NSF obviously plays an extremely
important role in the nation's ecosystem. It is focused on
workforce development. It is focused on training. It is focused
on providing grants to research institutions all across the
country. And that is incredibly important and we need that.
DOE's focus is as a capability organization. We are a mission-
driven R&D agency that is focused on science at scale to solve
complex challenges that only can be done by big-team science.
And so, we also work very closely with universities across the
country, but we are focused on specific problem sets that we
are trying to address.
And so, I think that the roles of NSF and DOE are actually
quite complementary because the NSF funds researchers that then
become part of the AI ecosystem that can become part of the
workforce that is going to help drive the frontier. But
ultimately, we do play very, very different roles in the AI
ecosystem and the innovation ecosystem.
The Chairman. Not to put you on the spot, but did we
direct, through the CHIPS Act, investments toward NSF that are
trying to recreate what you are already doing?
Ms. Fu. Well, they are certainly working on the national AI
research resource. And I will say, on that, because we
recognize the importance of access to compute, I will say this:
DOE's supercomputers have been open to the research community.
We leaned in and extended the life of Summit supercomputer at
Oak Ridge National Lab and also provided access to testbeds
through the NAIRR pilot. So we are working very closely with
NSF in this endeavor.
The Chairman. You don't feel that competition that one is
taking away? We can do even more if it was directed to what we
do and what our expertise is versus trying to duplicate that.
Ms. Fu. I will say, we are playing our part in the
ecosystem. And I firmly believe they are very, very
complementary.
The Chairman. Dr. Gleason, do you want to comment on this?
It's not the easiest. I am putting you in a spot, I know that,
but what we are trying to do--I think all of us are trying to
make sure that we have used our taxpayer dollars in the most
efficient and effective way possible, and have NSF do what they
do best and let you do what you do best, but not trying to
overlap each other.
Dr. Gleason. So this is a very good question. I would echo
a lot of what Helena said about complementary roles and
responsibilities. Maybe saying the same thing just a little bit
differently, from my perspective as a leader of scientists at
Oak Ridge National Laboratory, is that NSF is about access to
software infrastructure tools for the masses--for universities,
students, et cetera, to have access to artificial intelligence
and computational power to try out new methods, new algorithms,
evaluate new software and tools. Oak Ridge National Laboratory
and the lab system really provide world-leading capabilities
and world-leading science and technology input to that. So for
example, we, as Helena mentioned, we provide the Summit
supercomputer as part of the NAIRR effort. We have had lots of
people sign up to use that, as part of the NAIRR program, and
we help them achieve their science and technology objectives
because we have subject matter expertise that we can put
alongside them. How do you use a supercomputer? What is the
scientific area that you are trying to make discoveries in? We
have experts in those spaces that we can line up with them and
help them achieve their science and technology objectives.
So that is a differential. And we rely heavily on NSF for
workforce development, new students, training, that we can--
hopefully, some of those will come to the national labs and
contribute at some point.
The Chairman. Sounds like you all are doing better than we
are at working together.
Dr. Gleason. I agree.
The Chairman. And Dr. Kaushik, do you have a comment on it?
Dr. Kaushik. Yes, I agree with everything that they have
said, but I would point out one other thing. I think NSF's
focus is more on foundational research, on basic research where
we do not have near-term applications or we do not know about
exploring the unknown, where DOE, they are trying to
supercharge the capabilities we have to get things out there to
take an aim at moonshots. And I think that is very critical for
our national security enterprise, and the critical role--one of
the differentiators between DOE and NSF is the role DOE labs
play in furthering our national security--all the NNSA labs--in
furthering our national security work, as compared to the more
fundamental science that NSF tries to invest in.
The Chairman. Thank you all.
We will go for seven minutes, okay, to give you a little
bit more time since other Senators are at so many different
committee meetings. My main thing is on power, too, because I
know it's going to take an awful lot of power to generate this.
I am concerned. We have seen over 90 gigawatts of coal power
retired in the last decade. That is dispatchable, 24/7 power--
gone. And we can see twice as much dispatchable capacity retire
in the next decade on the path that we have been heading down,
historically. It can't be more clearly demonstrated than PJM's
recent assessment that 40 gigawatts of existing generation are
at risk of retirement by 2030, yet only 30 gigawatts of
equivalent capacity are projected to be added. Take that in
comparison, China is bringing on 90 gigawatts online every two
years of dispatchable power, and we are, I mean, this whole
environment, we are all responsible, but we have to understand
what we are dealing with.
I don't know if any of you want to comment on that, and if
you have raised the concern that we are not going to be able to
energize these data centers to compete? Just as quickly as you
can.
Ms. Fu. This is a focus for the Department. We understand
the implications of having enough power to power both
manufacturing that is coming back to the United States,
electrification of the grid, as well as the data centers and
the AI that is going to be needed to train those models in the
United States. The Secretary and the entire Department are very
focused on this issue. We just recently issued a new website--a
new hub--for folks who want to work with us on these issues.
I will say this: there are new technologies and new tools
that we have available--grants, tax credits, loans, technical
assistance--that we are bringing to bear on this particular
issue. Our Lawrence Berkeley National Lab is also working on a
study that is looking at energy efficiency in data centers. And
then, I will also point to the work that we were able to do in
the exascale computing project.
The Chairman. You are all on different grid zones. Are you
concerned about the grid being able to be reliable, the grid
that you are in?
Ms. Fu. We absolutely need to make sure that the grid is
resilient and that it's able to manage the load.
The Chairman. Are you all evaluating that and monitoring?
Ms. Fu. Yes. We are focused on this issue, especially as it
relates to data center energy growth.
The Chairman. How many of you are on PJM? Most of you are
on PJM as far as dispatchable, you know, where our power is--
the PJM system. I am just saying--I just think we are heading
for a real calamity here.
Anybody else want to comment on that?
Dr. Kaushik. I am happy to, Senator. Just one of the things
that we have to understand--the reality is, over the last six
years, the computational needs of AI systems have grown a
millionfold, like ChatGPT-1 to ChatGPT-4, the computational
needs have grown a millionfold. The parameters have grown a
millionfold. Now, since 2010, we have seen that the number of
computations that we are putting toward AI models is doubling
every six months. And there is no reason to believe that that
is going to slow down any time soon.
I think there was a recent Bloomberg article stating that
it is taking about seven years' delay for a completed data
center that would require 100 megawatts or more of energy to be
connected to the grid in Virginia. And I think those statistics
should raise some alarm--and I think to what Helena pointed
out, I think those tools are great, but we would need further
action, probably something, you know, not just about how we are
providing loans and financing to data center companies or
utility companies, but what are we doing in terms of energy-
efficient AI? What are we doing in terms of making sure that
data centers can come online, whether it's, you know, in the
long term, where we are building those data centers in the
United States, but in the near term, how can we make sure that
those data centers are not going to countries which are using--
infrastructure, but are using American hardware.
And so, I think those are important questions for us to
address. And I think the action that this Committee has taken
on permitting will have a big impact on that, but we need more.
The Chairman. Got you.
Senator Barrasso.
Senator Barrasso. Thanks, Mr. Chairman.
Let me stick with you, if I could, Dr. Kaushik. You are an
expert on China's science and technology policy. What unique
threats does China pose to government-funded research and
development of critical and emerging technologies?
Dr. Kaushik. I think, Senator, the answer is very clear.
They have a targeted effort that we call the civil-military
fusion, where even their universities have put out statements
saying that all the basic foundational research that they do is
toward military gain. Every IP theft case that we have seen,
every espionage concern that we have seen, I think, all of that
is ultimately feeding into their military complex. Now, that
plus their legal situation with the National Intelligence Law
or the National Cybersecurity Law and all those things are
making it more complicated for Chinese researchers to be able
to actually act in good faith because no matter what
commitments they made to U.S. institutions, they are required
by law to share all that information with the Chinese
government. And I think when they are trying to go after our
technology, as you mentioned with regards to a recent espionage
case where a researcher was charged regarding data center
plans, I think they are going after that cutting-edge
technology. They want to--they cannot independently replicate
it. And so, they want to steal our IP and build it there in the
short term to be able to catch up.
These numbers are not just a mirror reflection of their
investments in R&D. These are also a reflection of how they
have exploited our open system.
Senator Barrasso. Ms. Fu, he talked about the targeted
researcher for military activities, and if you follow the long
history of Chairman Mao, his efforts when he was on Stalin's
payroll were to try to get the nuclear weapons, and they fast-
forwarded the technology to him. When Stalin was attacked
during World War II, his goal was then to try to work with the
United States to get the nuclear weapons. I mean, all of the
issues have been a militarization of China.
So you know, in 2020, the Hoover Institution released a
169-page report. I know you are familiar with it. It is titled
``Global Engagement: Rethinking Risk in the Research
Enterprise.''
[The report referred to follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Senator Barrasso. The report states that a group of
universities in China, known as the Seven Sons of National
Defense, tying this into the military, operate as prime
pathways for harvesting U.S. research and diverting it to
military applications. Can you please explain why the
Department still collaborates with researchers at the Seven
Sons of National Defense?
Ms. Fu. So I will say research security is an issue that I
have long thought about and the Department has been extremely
focused on since the Manhattan Project. I personally worked on
research security policy in the last Administration at the
Office of Science and Technology Policy. So this is an area
that is critically important for our continued national
competitiveness.
DOE, as a government agency--we do not have any bilateral,
ongoing cooperation with Seven Sons institutions. We are aware
and very clear-eyed about the risks of China's military-civil
fusion policies. And that is why we have a managed research
environment within the fence lines at the DOE National Labs
when it comes to our science and technology risk matrix, the
foreign national access screenings that we do, as well as our
ban on foreign government-sponsored talent recruitment plan
participation.
Senator Barrasso. So, Dr. Kaushik, you mentioned the Seven
Sons of National Defense in your testimony. What is your
opinion? Should the Department just cut all ties along these
lines, or is there further----
Dr. Kaushik. Certainly, I think no researcher being funded
by any taxpayer dollars should be collaborating with any
researcher at Sevens Sons of National Defense.
Senator Barrasso. So what are several things that the
Department of Energy could do better to protect advanced
computing research?
Dr. Kaushik. I am so glad you asked that question because I
think there are a lot of things that the DOE has been doing
that are ahead of the curve compared to other agencies, and in
fact, I have to give them credit for that--the recently
released ``Guidance for Implementing National Security
Presidential Memorandum 33 (NSPM-33) by the OSTP. DOE is
already implementing things that are ahead of that.
Now, what more can we be doing? And I think this doesn't
just apply to DOE, but to the broader research ecosystem. I
think we need to increase transparency here, right? We must be
more forthright with the researchers about the concrete risks
to our national security. Currently, we just often present
warnings or hypothetical scenarios of illicit technological
transfer, which is insufficient to convey the gravity of the
situation. These are researchers and scientists. These are
generally well-intentioned people who are analytical, who want
to see more evidence to be able to see what the stakes are. An
example I will give you is the MD Anderson case, which was
about the NIH. They provided a clear example of a security
breach, where an email explicitly directed the transfer of
privileged information. And that letter was published openly.
So I think such concrete examples are more persuasive than
generalities.
The other thing I would say is, we need to clearly
delineate basic research environments--our research
environments more generally, right? We need to reassess the
classification of basic research. We need to establish clearer
boundaries between open and secure environments. Now, for
example, in the Air Force Research Lab in Rome, they have a
facility outside the gate to interact with universities and
others, whereas, everything behind the gate is doing secure
research. Now, DOE does a lot of that at a lot of labs, but
that needs to be the practice. That needs to be the norm.
It is not clear to me why NSF funded hypersonics research,
for instance, which is considered basic research. I would be
all for having basic research open if it was truly basic
research, but that is not basic. And I think that is something
that we need to clarify.
And then, I would say that very specific to DOE, we need to
strengthen the protection of commercially relevant
technologies. Now, a critical area that requires our attention
is the protection of research that is conducted in partnership
with commercial entities. National labs and universities
frequently engage in projects with significant commercial value
through mechanisms like strategic partnership opportunities or
user facilities agreements. Now, current policies, whether it
be the NSPM-33 or even the CHIPS Act, they do not adequately
address the security concerns associated with these public-
private partnerships. We need to develop more comprehensive
guidelines and authorities to manage these partnerships
effectively to ensure proper safeguards there. And I think
this, again, extends beyond DOE facilities, but also the
universities that are collaborating with major technology and
pharmaceutical companies.
Senator Barrasso. Thank you.
Mr. Chairman, my time is expired. I just wanted to comment
on the last question you asked about the disposable power.
The Chairman. Right. Dispatchable.
Senator Barrasso. And you know--I'm sorry, dispatchable
power. The magazine, the Economist, says that, kind of, since
the Paris Climate Accords, the U.S. has taken out about 150
gigawatts of coal-related power. China and India, 250 more
gigawatts on. Why do they need it? Because of the technology,
the AI, all of those things. The New York Times says five years
from now the amount of energy that we are going to need in this
country is equal to adding an entirely new California to the
grid. Why? Because of this very reason. So we seem to be
woefully behind in our ability to produce the power that we are
going to need and are taking down, right now, power that can be
used and needed because of the attacks from the environmental
communities, but we need to be very clear-eyed about what is
happening globally and what China is trying to do in terms of
trying to become the military, the economic, and the
technological superpower of the world. So thank you, Mr.
Chairman.
The Chairman. Thank you, Senator.
Senator Hickenlooper.
Senator Hickenlooper. First, let me thank each of you for
your service and being involved in this. Mr. Chair, I
appreciate you for calling this meeting.
Let me start. Dr. Kaushik, AI obviously can improve
efficiency and reliability of our electric grid in predicting
and managing supply and demand. While we work toward that goal,
however, the National Electric Reliability Cooperation, as well
as the Bravo Group, projected, as we have just been discussing,
significant increases in demand on the grid in part from data
centers that are training and running these AI systems. So Dr.
Kaushik, how much do we know, and what more information do we
need about what our benefit is here? In other words, how much
more electricity will AI and these emerging technologies
require, and how can we begin to estimate the benefits and the
savings in energy that they will provide us?
Dr. Kaushik. Thank you, Senator. I think predicting this
future demand is going to be challenging, one, because we don't
have the insights into proprietary plans of companies for
training new models, or sometimes they have speculative,
unfulfilled requests for our new capabilities from data center
vendors. But what we do know is that the International Energy
Agency has predicted that the electricity demand for data
centers will double by 2026--the global demand. And I think
data, even if it does not double, even if it is just, you know,
increasing by ten percent, we cannot take that risk. We need to
be prepared for that. I think, as Dr. Gleason mentioned, that
there are some reports that predict that 10 percent of global
energy usage will be by AI data centers.
I think we have to reduce that uncertainty, and I believe
DOE is already working on that as part of a mandate they have
from the Energy Act of 2020, but I think overall, the facts of
the matter on the ground are that in your home state, for
instance, in Colorado, a hyperscaler recently invested in
geothermal energy west of the Rockies. I think the private
sector is seeing this, and they are like, we need to invest in
the capabilities. Just this week, on Oracle's earnings call,
they shared that they are willing to build a one-gigawatt data
center and they have started the process on that. And these are
things that companies are actually investing in, and I think
that is a clear signal for why we need to be prepared.
Now, what needs to happen to reduce that uncertainty? I
think AI can help there. We have all these earnings reports
coming out. We have all these announcements from companies. We
have the GIS data. We can use that to better reduce our
uncertainty on how much power demand will be in particular
regions. It may not be an issue nationally, but it's certainly
creating strain on particular regions like Northern Virginia,
for instance.
Senator Hickenlooper. A lot of regions, I agree completely.
Ms. Fu, the National Institute for Standards and Technology
(NIST) has established the U.S. AI Safety Institute to work
with government agencies and the private sector to develop
metrics and benchmarks and tests that ensure AI systems are
both safe and reliable. And as we have been hearing today, DOE
is home to incredible technical expertise, a world-class
computing infrastructure, valuable scientific data that
positions the Department to be, really, a leader in all this AI
research and development. Ms. Fu, could you describe how DOE
collaborates with NIST and NSF and other agencies, but that
collaboration to advance AI safety research which, I think, you
know, we are getting the cart before the horse here a little
bit.
Ms. Fu. Thanks so much for that question. Certainly, AI
safety and security is of key importance for DOE and for the
country. This is something that the AI executive order speaks
to. We have been coordinating with commerce very closely. We
are working on an MOU with the AI Safety Institute so that we
can bring the expertise at DOE to bear on these questions of AI
safety and security. One thing I will note is that our National
Nuclear Security Administration has been focused on this issue,
even before the issuance of the AI executive order because of
how important it is to our mission, and they have been doing
evaluations of open models and proprietary models for some time
now. And we think that there is expertise that we have
developed through that process that we hope to bring to bear in
our cooperation with the AI Safety Institute.
With NSF, we have been working very closely to try and see
where we can align some of our workforce efforts. We talked a
little bit about that before, about how we can draw the line a
little bit more closely between the workforce efforts they are
doing at universities all across the country and the resources
and training and capabilities we have at our national labs. We
also are co-funding a research coordination network on privacy-
enhancing technologies because we think that this is an
incredibly important field of work that is important to
coordinate on.
Senator Hickenlooper. Of course, absolutely.
And I figure there is an opportunity here for the White
House Office of Science and Technology Policy to begin to help
coordinate all these different efforts, because there does need
to be some sort of, you know, the hive mentality has to be
directed properly and coordinated.
Dr. Gleason, Colorado's Elevate Quantum recently received
an award from the Department of Commerce to expand our--what we
feel is the nation's leading quantum ecosystem, and Colorado
companies, universities, and the workforce are determined and
ready to build on this federal investment to create quantum-
enabled technology that solves real-world problems in
navigation, communications, computing, et cetera. Quantum
computers and classical computers are most effective at solving
different and distinct problems. So how will DOE bring together
the different advanced computing technologies, including
quantum and classical computers, to solve these, the most
challenging science and technology questions?
Dr. Gleason. So, very good question. It's a very important
question. I think what I will start with is that it comes down
to heterogeneity of computer systems. So what I mean by that
is, bringing different computing technologies together to bear
that can attack different parts of a problem where they are the
most efficient technology to attack that piece of the overall
challenge. And I will just use history as an example. So high-
performance computing used to be CPU-focused only. Then, they
became more heterogenous by adding GPUs.
Senator Hickenlooper. Right.
Dr. Gleason. And GPUs are very good at AI training, et
cetera. CPU is good at modeling and simulations. So there is an
example of where heterogeneous computing can be used to attack
different types of problems, different parts of a problem.
I think the future--one of the things that DOE and the
national labs are very focused on are what are the next
versions of heterogenous computing that are on the horizon. And
you have already mentioned one. We talked a lot about quantum
recently in this hearing. Quantum computers are very good at
solving optimization problems, for example. Logistics problems.
They are very good at simulating molecular dynamics or
chemistry because of the nature of the quantum mechanics that
are used to drive the quantum computers. So integrating
classical HPC with future quantum computers, I think, is a
really important thing that the Department and the national
labs need to look at, and that is a big challenge. How do you
integrate these very different types of architectures that run
with very different software stacks and software platforms? How
do they connect together? How do they communicate? And how do
you make a quantum co-processor, for example, that lives in a
data center with a frontier supercomputer and takes on some of
these challenging parts of the problem, probably in a more
energy efficient way and probably faster? Other technologies--
neuromorphic computing is another good example of another
heterogeneous option. Neuromorphic co-processors can also
tackle unique challenges.
And so, and then, I think there are computational
architectures that we haven't even thought of yet, and the
Department of Energy is exploring and the labs are exploring
what those might be, but I think increasing the heterogeneity
is one really strong path that DOE and the labs should follow.
Senator Hickenlooper. I think we are going to need several
more hearings to really explore this thoroughly.
Senator King [presiding]. On behalf of the Chairman,
Senator Hawley.
Senator Hawley. Thank you very much, Senator King. Thanks
to all the witnesses for being here.
Dr. Gleason, if I could just start with you. I just want to
start with a question about Oak Ridge National Lab, since you
are here from that institution. You have worked there for more
than 30 years, is that right? Thirty-four I think you said in
your opening statement. Do you mind if I ask, do you happen to
live in the area as well?
Dr. Gleason. I do.
Senator Hawley. Here is why I am asking. You are probably
familiar with the role of Oak Ridge in the Manhattan Project, I
bet. Is that fair to say?
Dr. Gleason. I am, yes.
Senator Hawley. Could you just give us, for the benefit of
those who are watching the hearing, could you just give us a
thumbnail sketch of what Oak Ridge did for the country, to the
extent that you know it? It doesn't need to be detailed, but
just give us a synopsis of what Oak Ridge did for the country
during the Manhattan Project?
Dr. Gleason. Yes, I mean, Oak Ridge, you know, was stood up
in the early 40s to try to help build the nuclear material that
would power the bomb. So we built the world's first
continuously operating nuclear reactor, called the Graphite
Reactor, which is part of that process. The Graphite Reactor is
no longer operational but you can go look at it. It is a
wonderful tour stop when you come to Oak Ridge, but we are very
proud of our roots in the Manhattan Project and it continues to
inform our national security mission, part of what we do. We do
Office of Science work. We do national security work. But our
roots in the Manhattan Project--one of the big areas that came
out of that was our materials science strength. And we have a
huge pillar and strength in materials science that's roots go
back to that Manhattan Project and Oak Ridge's contributions.
Senator Hawley. Very good, thank you for that.
I asked because Oak Ridge was such a critical part of the
Manhattan Project, as you say, and the Oppenheimer movie of
this last year, I think, raised awareness about what that
project looked like and how it got started and the importance
of Los Alamos, but really at Oak Ridge, you know, the
government produced plutonium and then there were other uranium
processing sites around the country, including in my home State
of Missouri, in St. Louis. And as you probably know, in both
Oak Ridge and St. Louis, despite the fantastic vital
contributions of those processing plants, facilities, and labs
to our national security, the nuclear waste was not properly
disposed of in either place, and in Oak Ridge, the effects on
the community have been severe. And I just want to quote from
another one of your fellow residents there in the Oak Ridge
area--Tanvi Kardile is her name. She is the coordinator for the
Oak Ridge Environmental Peace Alliance, who has said that, ``It
is time that people in Oak Ridge receive compensation for being
exposed to radiation from nuclear waste.'' And now, Oak Ridge,
as we speak, is a Superfund site, and that is good. It is well
long overdue that it be cleaned up, but not only should it be
cleaned up, those folks, your neighbors in the area who have
been exposed to nuclear radiation over the years, ought to be
compensated for it, just like in the city and region of St.
Louis.
And I bring all of this up because this body, to its
credit, passed legislation that would compensate the good folks
of Oak Ridge and the St. Louis area and other similar nuclear
sites and other folks who were exposed to downwind testing
during the Manhattan era and the Cold War era. We passed that
legislation by 69 votes. And I look across the aisle, Senator
King voted for that. I appreciate that, Senator. And many--most
of my colleagues, it was a huge vote. It's now in the House of
Representatives. And I just, having you here from Oak Ridge,
considering everything Oak Ridge and the entire region there
that you work at has done for this country, I couldn't let the
opportunity pass to thank the good people of Oak Ridge, to
thank the scientists that you work with, to thank that
community, and also to call on my colleagues in the House to
pass this legislation. Compensate these good Americans, who
bore the brunt of our effort in the Second World War and the
Cold War, which we won because of the effort of people at Oak
Ridge and in St. Louis, and the residents who weren't
themselves scientists but lived in the area and have suffered
the effects of the nuclear radiation--they are proud to be
Americans and to have served in this effort. They deserve
compensation, however, just as they deserve to have their
communities cleaned up.
So thank you, Dr. Gleason, for testifying about the
significance of that.
Dr. Kaushik, if I could just come to you in my remaining
moments here, just about the dangers posed by China and AI. You
highlighted in your opening statement--your written statement--
the 2017 National Intelligence Law in the PRC requiring Chinese
citizens and organizations to share information with state
intelligence. Can you just speak to the dangers of American AI
companies doing business with China, investing in Chinese AI,
partnering with the Chinese businesses, whether overtly state-
controlled or not? Can you speak to that a little bit?
Dr. Kaushik. Senator, thank you for that question. I think
it hits right at the heart of the argument on economic security
and national security. I think American businesses doing
business in the PRC--at one point, our government promoted them
to do it, and that is okay, like, that was our understanding of
the PRC back then. Things have changed. And today, if a company
goes and invests in China, what they are doing is inherently
supporting their economy, inherently supporting an
authoritarian regime. They don't know whether what they are
investing in is also fueling slave labor in Xinjiang. I think
that is an important consideration for any American company who
wants to invest in other countries--they have to take that into
account.
We have to be realistic about what is happening. We cannot
live in a dreamland of the China we want to operate with. We
don't control that. We control what is the China that we are
operating with.
Senator Hawley. I am so glad to hear that from you, and I
hope that every corporate CEO will listen and take to heart the
words you just said. I asked partly because I just had the
opportunity to talk with an Intel executive who was testifying
at a different committee. Intel, of course, is investing
billions in China, and in Chinese AI, in particular. And this
executive argued to me--this was just 48 hours ago--argued to
me that this is of great benefit to the United States. There
are no security concerns with it at all. It's really good for
America. But the telling thing was, you mentioned the Uyghurs--
Xinjiang Province. He would not condemn the persecution of the
Uyghurs. He said, ``you know, well, I just--I can't speak to
that. It's unclear if there really is any forced labor in
China.'' This, I think, is, frankly, the moral hazard that our
companies face if they do business in China. If they find
themselves compromised by the PRC, they are effectively--they
run the risk of supporting slave labor, of helping the Chinese
AI program in a way that is materially to our detriment, but is
also, frankly, just morally wrong.
So thank you for your clarity on that. And I just think
these companies, these so-called American companies, it's time
that they actually did something for the values that we cherish
as Americans, and ending slave labor ought to be at the very
top of that list. And I, frankly, am sick of these companies
taking billions--Intel is getting billions--$8.5 billion--just
approved to go to Intel under recent laws that this body has
passed. And yet, they are turning around and spending billions
in China, and won't condemn slave labor--in fact, they may be
benefiting from it. And I just think that's wrong.
I see my time is expired. I have a couple more questions
for you, Dr. Kaushik, just about safeguarding our national
labs. So I will give those to you as written questions.
Thank you, Mr. Chair.
The Chairman [presiding]. We might have a second round too,
if you can stay, okay?
Senator Cortez Masto.
Senator Cortez Masto. Thank you. I want to thank the
Chairman and Ranking Member for this conversation. I love this
chart, by the way, thank you.
I understand one of my colleagues talked about it, but I am
just going to put a finer point on this because I understand
that this conversation just highlights the fact that a rapid
increase in projected electricity demand is happening because
of AI and because of data centers. And I know this in Nevada
because we actually have both happening here, and we are a
major hub for data centers. You just had conversations with
Switch in my state on this very topic. So I am very interested
in how we address that demand. And I think it is everybody's
responsibility to figure this out if we want to lead in this
country in emerging technology, right? So if you have any
additional information, I would love to hear it, on how we
address and be a part of the solution here to address that
emerging demand as well.
Ms. Fu.
Ms. Fu. Yes, this is something, as I mentioned earlier,
something that the Department is laser-focused on. And I think
it is due to our role in leadership in advanced computing,
understanding what the energy needs of AI will be, but also in
our role with the energy part of our name. We see an enormous
opportunity here. It is true that energy demand is growing. It
is doubling. Energy demand from data centers is doubling from a
very low base. However, it is a local issue that we need to
really get our heads around. We think that DOE has a really
unique role here in helping to convene stakeholders. This is
something that the Secretary has been thinking through. She has
charged her Secretary of Energy Advisory Board to look at this
issue and they recently came out with recommendations. We would
be happy to share them. We are starting to implement and look
through those recommendations to see what we can do. We will be
convening stakeholders around the country in areas of high
load-growth. This is going to be an effort over the next
several months.
We do think that there are things that we can do now. And I
think part of the challenge with AI load-growth is that it's
both a very large load and the expectations are very fast, you
know, this large load is going to come online in the next--now,
through the next few years. And so, we have many different
kinds of technologies that we are looking at that are going out
to 2030. We are doing everything that we can to look at what we
can unlock now.
Permitting, of course, is one piece of that. We actually
have an AI and permitting pilot that we have already launched
to see how we can use AI to expedite and streamline the
permitting process. And one of the things that we have been
able to do is to take the entire corpus of NEPA documents,
which normally go in a huge binder and they get put on the
shelf, and make them AI-ready, digitize them, and make them
available to the entire scientific community and to industry to
develop new tools to help with this process.
But we know that there are near-term needs, mid-term needs,
and longer-term needs, and DOE is focused on all of those.
Senator Cortez Masto. And I know you are, because of your
footprint in Nevada, by the way. And so, thank you. My only ask
is, and you talked a little bit about it, you invite industry
in to have them as part of the solution. I talk with the CEO of
Switch regularly, and I just talked to him about this issue.
They have ideas.
Ms. Fu. Yes.
Senator Cortez Masto. Right. And so, they should be part of
this discussion.
Ms. Fu. Absolutely.
Senator Cortez Masto. Nevada is a perfect place to have
this discussion because of DOE's footprint, but also there is
this nexus between energy and water and the challenges we have
there. There is also, in Nevada, a large footprint just of the
Federal Government that should be partnering with the state and
private sector in figuring this out. And for that reason, I am
hoping DOE is also partnering with the Department of Interior.
You talked a little bit about permitting, but it's Department
of Interior that has/owns a lot of the land, right? So I am
hoping that--and can you talk about that? Is there that
partnership that is happening with other federal agencies,
including the Department of Interior?
Ms. Fu. Yes, there is a lot of focus on this. We are
working through the CEQ that is convening the interagency body.
I will say the AI and permitting pilot, we are working with the
13 agencies in the permitting council on this issue. And we
recently, through Savannah River National Lab, issued an RFI
that looks at potentially siting data centers, even on federal
land.
Senator Cortez Masto. Okay.
Ms. Fu. So we are looking at all of these issues, looking
at all of these options. We understand the urgency of the
issue.
Senator Cortez Masto. Okay.
And the only other stakeholder that must--must be at the
table is our state and local folks, right, because we manage
our economic development, our population growth, our needs with
the Federal Government because the Federal Government owns most
of the land in the state. So I am hoping that you pull our
folks in as well in this conversation. It is crucial. Thank
you.
You talked a little bit about the workforce that we need
for the future. What else should we be aware of here in
Congress that we have to focus on, particularly for that
workforce of the future that is going to kind of lead and be a
part of these emerging technologies? Anything that we didn't
talk about that I need to know, or we need to know?
Dr. Kaushik. Happy to take that on. I think, Senator, that
is an incredibly important question that is at the heart of
this conversation. The population of the PRC is four times
ours.
Senator Cortez Masto. Right.
Dr. Kaushik. They are producing four times as many
bachelor's degree holders in STEM, twice as many master's and
twice as many Ph.D.s. This is not a competition that we faced
in the Cold War. The population of the Soviet Union was nearly
as much as ours. The economy was far below, and here the
population is much higher, much more talent to tap into, and
they have these sophisticated programs and they are trying to
bring more talent into their country. I will give you an
example. They have this national innovation-driven development
strategy, which aims for the PRC to become a key hub for global
high-end talent by 2050. Now, they have changed their visa and
permanent residency programs. They have established a
``thousand foreign talents'' program, alongside the Thousand
Talents Program, to recruit foreign nationals to China. And I
think we need to have a talent strategy here.
What does our talent strategy look like? DOE has some
fellowships for U.S. nationals. I think NSF has their own
fellowships. Agencies have their work cut out here, but at the
same time, there is a big chokehold that only this body can
solve, and that is--I will quote the Singaporean leader, Lee
Kuan Yew, who said when he was asked whether he thought China
would overtake the United States in the 21st century, he said,
``No, because the United States has long attracted the world's
best and brightest.'' He said that the United States fosters a
diverse culture of creativity, and China will struggle to do
so. Despite having 1.3 billion people to tap into, the United
States has seven billion to tap into because we can assemble a
rest-of-the-world team. But I think you see that in AI today.
Sixty-five percent of the top AI startups have at least one
person as a founder or co-founder who came here through legal
immigration means. I think that has to be an important part of
the conversation on workforce.
So we need an all-of-the-above approach here to be able to
match the numbers that they are putting out in STEM Ph.D.s and
STEM master's and bachelor's.
Senator Cortez Masto. And then finally, and I agree, but
that can be done, Dr. Kaushik, with what your caution was
before of how do we secure it? How do we ensure that we are
securing the technology, right, for our use in our labs? There
is a way to balance it and you believe that can be done?
Dr. Kaushik. Absolutely. There are, you know, the
guardrails I talk about, most of them are objectively laid out
in what we call the National Security Decision Directive 189,
which was issued during the Reagan Administration during the
Cold War, about protecting American technological advantage.
The National Academies also did a report in 2022 at the
direction of this body on protecting U.S. technological
advantage, and they said that it is possible. Now, what do we
need to do to actually get there? I think we have to recognize
the competition is a bit different. You know, we cannot
continue funding hypersonics research as basic research and
saying that we will make that openly accessible, right? We have
to classify the research when it needs to be classified. We
have to have that conversation.
A risk matrix cannot be a silver bullet. We have to have a
prior conversation on what is okay to be made openly accessible
and what needs to be behind closed doors, who should have
access to certain research and who should not. And I think we
need to be very careful about whether it is even people from
our partner and allied countries to be able to have access to
those, you know, those technologies that we are researching
that are of a sensitive nature. It's not just about preventing
PRC access to those technologies. Sometimes we just don't want
the fact that we are developing those technologies disclosed. I
think there is a way we can balance that, and I think NSPM-33
does a really good job at it. It was produced by the Trump
Administration as an all-of-government approach. The Biden
Administration, to its credit, has continued working on it.
Despite OSTP's massive delays in releasing implementation
guidance, it is the right approach. The Trump Administration
also issued the Presidential Proclamation 10043, which bars the
entry of certain Chinese graduate and postgraduate students who
have ties to military-civil fusion institutions. And I think
that is taking a scalpel rather than a sledgehammer to this and
that is the right way to go about it.
Senator Cortez Masto. Okay, thank you.
The Chairman. Senator King.
Senator King. Thank you, Mr. Chairman.
As we have been sitting here, a few minutes ago, I asked a
question of ChatGPT-4, and the prompt was ``give me a poem in
the style of Robert W. Service about the U.S. Senate Committee
on Energy and Natural Resources.'' Okay, Robert W. Service is
my favorite poet. He wrote about the Gold Rush, and wrote ``The
Cremation of Sam McGee'' and ``The Shooting of Dan McGrew.''
Anyway, that was what I asked. In three seconds, this is what I
got back:
``In the hallowed halls of Congress where the mighty stand
so tall, there is a group that tends the nation's needs, its
energy and all. The Senate committee gathers round with purpose
in their eyes, for the fires of coal and sun and wind they seek
to harmonize. Old Service might have marveled, for their work
is just as bold as the gold rush tales he spun so well of lands
both harsh and cold. They wrestle with the power lines, the
drilling, and the frack, the forests and the oceans deep and
what they give us back. Oh, the Senator from Texas . . .'' and
by the way, this proves AI isn't always right, there is no
senator from Texas on this committee, ``. . . with a mind on
oil's flow, speaks of wealth beneath the earth and profits yet
to grow . . .'' listen to this: ``. . . but across the room a
voice rings clear from Maine's wild rocky shore, there is wind
upon the sea my friend, we must demand much more. The coal
fields of Wyoming, the pipelines out of sight, they spar with
solar panels shining clean and bright, and as they hash their
quarrels out with power in debate, the future of the nation's
grid they strive to navigate.''
Isn't this amazing? This was three seconds. It continues:
``. . . yet all the while the rivers run, the mountain stands
so high, the forests whisper ancient tales beneath the open
sky, and those who sit and legislate with duty on their chest
know nature's gift is fleeting and they must do their best. So
let them weigh the costs and gain. Let compromise be found, for
in the hands of those who lead the nation's fate is bound. And
like the miner seeking gold and Service's rugged lore, the
Senate digs for answers deep on nature's precious shore.''
That is what AI can do, which I think is absolutely
astounding, but the fact that it made a mistake, I think, is
also a cautionary tale.
Okay, Senator Barrasso was talking about the electricity
issue and the coming problems, and that is one of the reasons
we are supporting--I don't know if you know, but two weeks ago
we reported out a comprehensive permitting bill in order to
unleash the power, particularly of renewables, because without
transmission we are not going to have those resources.
Okay, one question on quantum computing and advanced
computing, and I don't want to pursue this too deeply, but I
want to be sure, Ms. Fu, that we are not duplicating. There are
other people in the U.S. Government working on this problem and
I just want to be sure that there is coordination, and you
don't have to go into any detail, but assure me that you know
what is going on in other agencies and that we are doing this,
because this is an expensive proposition, that we are not
duplicating and overlapping because of competing jurisdictions.
Ms. Fu. We are not duplicating. We are working closely with
the White House National Quantum Coordination Office, and
working closely with other agencies on this issue. And I think
what we talked about earlier around AI, you know, and where DOE
plays a role is around developing the capability. And so, the
foundries that we are developing, the testbeds, the different
kinds of user access programs, these are things that only DOE
can do. We work with other agencies to build up the broader
ecosystem, but the large-scale science and the facility side of
this issue is something that is a DOE strength.
Senator King. Well, please keep an eye on that issue.
Ms. Fu. Absolutely.
Senator King. And in the context of this Committee, are
there ways that AI can assist us in the more efficient
allocation of power, the more efficient running of the grid?
There is a lot of inefficiency on the grid today, and we are
developing what are called GETs--grid-enhancing technologies. I
think that a complex system like the grid, and by the way, we
need to distinguish between power and the grid, between
capacity and the wires. But can AI be helpful to us in more
efficiently managing the grid?
Ms. Fu. It can, and I think just taking it back to the poem
that you wrote with ChatGPT earlier, I think that is a really
good example. ChatGPT and all of the commercial large language
models were trained on human-generated information off the
entirety of the internet. And the kinds of things that we are
talking about in our FASST proposal are not training on the
internet. It's not training human-generated data. It is
training on scientific data where we can trust the provenance
of that data and it can follow the laws of math or biology or
physics. And so, when we look at the grid and we look at the
things that we are going to need, and that is a very, very
high-consequence use-case, this is lights on or lights off. We
need to 100 percent have trust in where the data is coming from
to train this and what----
Senator King. I am assuming it's trustworthy data, my point
is, analyzing massive data about how the grid is operating will
enable us to operate it more efficiently. Is that correct?
Ms. Fu. Yes, we think that there is a huge use-case on
using AI to look at how we control massive systems like the
electric grid.
Senator King. Thank you.
Final point, and we have touched on this--we just did. Dr.
Kaushik, how much talent are we losing because of our not-very-
functional immigration system and, you know, the quote from
Lee, Kuan Yew, who is a genius, is very appropriate, but my
sense is it used to be a lot of foreign students would come to
the U.S., they would get a graduate degree from Stanford or
Notre Dame or NYU and they would stay. Now they are leaving.
Now we are making it harder for them to stay. How do we open up
this system safely? I understand there are security concerns,
but my sense is we are losing a lot of talent.
Dr. Kaushik. Certainly, Senator. I think there was a recent
study that was done which found that over 50 percent of AI
Ph.D. graduates who leave the country cite that as the reason
why they are leaving the country.
Senator King. Because of the immigration laws?
Dr. Kaushik. Yes.
Senator King. So there is a pile of talent who we are
chasing away?
Dr. Kaushik. Correct. And many of them are from nations
which we consider as partners and allies. I think that is
something that we have to recognize, that there is----
Senator King. So we are doing this not only to Chinese
people, or Russian people, but we are doing it to Australians
or people from France or Germany.
Dr. Kaushik. Correct, yes, Senator.
Senator King. That is just crazy.
Dr. Kaushik. That is totally the case. In fact, the CCP
recognizes this, and I am going to give you some quotes
directly from them where commenting on the U.S. retention of
Chinese STEM students, I will just say, the head of CCP's
Central Talent Work Coordination Group complained that the
number of talents lost in China ranks first in the world. But
now, if you look at their data--granted all the issues in their
data--the fraction of the Chinese students who are returning
home is increasing despite long-term stay rates. If you look at
students from the PRC who graduated from U.S. universities in
2015, over 90 percent of them are still in the United States,
but that number is declining overall, especially at the
undergraduate level, it's declining a lot.
A state-run consulting firm wrote in an AI policy white
paper in China that U.S. restrictions on immigration have
provided China opportunities to bolster its ranks of high-end
talent.
Senator King. Brilliant.
Dr. Kaushik. These are direct quotes from the Chinese
Communist Party.
Senator King. Shouldn't we--assuming sufficient security
analysis--shouldn't we just staple a green card to diplomas?
Dr. Kaushik. I think, Senator, that is a worthy
consideration. I would say that we do need security provisions
attached, like some of the provisions that were under
consideration here in the CHIPS and Science Act, for instance,
of applying sanctions to people who are found guilty of
economic or industrial espionage or academic espionage. I think
that is something that we should be----
Senator King. We can deal with that and that is a small
percentage.
Dr. Kaushik. Correct.
Senator King. In terms of gaining a huge amount of talent
that will drive this country into the future.
Dr. Kaushik. I think what you are saying is absolutely
right, and this is what Bill Evanina, who was a former
counterintelligence chief in the previous administration, also
said that we bring about 350,000 Chinese students here every
year. We have very liberal student visa policies for them.
About one in a thousand of them are bad-faith actors, and so we
need that scalpel-based approach to remove that one out of a
thousand rather than shutting the door on all 350,000. And I
think that is the right approach that the Trump Administration
took with its Presidential Proclamation 10043 and NSPM-33 and
the likes. I think there are obviously still a lot of issues
that we need to deal with and that can be done. At the same
time, we--I think it was Senator Cantwell who said during the
CHIPS and Science Act that either we recognize this problem
today or we will wake up in ten years and China will be ahead
in everything and we will have no time to work on it.
Senator King. Thank you.
Thank you, Mr. Chairman.
The Chairman. Thank you.
Senator Murkowski.
Senator Murkowski. Thank you, Mr. Chairman and thank you to
those of you who have been part of the conversation here this
morning.
Mr. Chairman, I want to thank you for your leadership and
inviting me to join you with the Department of Energy AI Act
that we are working to advance to make sure that DOE has the
tools that are necessary to really leverage some of the
benefits here of AI. I look forward to being able to advance
that and also to be supportive of the measure that Senator
Daines is introducing, the DOE Quantum Leadership Act of 2024,
I think all good measures that just add to the toolbox here.
Senator King, I sit in just amazement of what you shared.
Although, I will say that the ``The Shooting of Dan McGrew''
was still one of Robert Service's finer pieces of work, and I
think he would have been offended by the----
Senator King. ``A bunch of the boys were . . .''
Senator Murkowski. ``. . . whooping it up in the Malamute
Saloon.'' Yeah, we could go on, and I will regale you with my
Robert Service. It's one of my favorites, so you made me smile
here this morning.
I want to ask a little bit more about application of these
technologies. There has been a lot of focus in the Committee
here on national defense, competitiveness issues, power grid
resilience, and as Senator King mentioned, efficiencies. Talk
to me a little bit about how we can utilize these technologies
in other fields, whether it's resource exploration, disaster
prediction, climate resilience, adaptation planning. Let me
share with you--I had a meeting just yesterday with some
scientists from Woodwell, focused on what they call the
Permafrost Pathways. And it's essentially utilizing available
mapping to understand and determine where we have greater
vulnerabilities, whether it is with releases of carbon, as we
are seeing the permafrost melt or just the application in
whether or not you would even think to put a community or a
structure there, and a lot of discussion about where the data
is. As scientists, are they able to access the public
repositories? You know, BLM does certain analysis, or a service
in Alaska does analysis. What is available to them and how
would their work be facilitated in other ways? And I think
exactly about some of what we are talking about here.
The other thing that I will ask you to think about and help
me with, we had a deadly landslide in my hometown of Ketchikan
just a few weeks ago. This is an area in a neighborhood where
my family grew up, had the same house for 60 years, and not in
an area that is prone to landslide, and yet, as we are seeing
more intense weather, we are seeing these impacts in our
communities that are really quite frightening. And so, we are
asking what mapping have we done? What monitoring is going on
in terms of precipitation?
And we are told that, well, at the place that we were at
the day of the landslide, the rainfall was somewhere between
six and eight inches that day, whereas the weather monitoring
station, which is right across--maybe a half a mile away, is
less than two inches. Still a rainy day. And so, we are talking
about--we have these tools that are out here, but who is able
to communicate with who? And I am giving you a rambling open-
ended question, but I have a lot of problems that we are
looking to solve on the disaster side, or just preparedness
awareness, but also with being able to utilize some of the
public data that is out there as we are trying to assess is
there greater potential for a critical mineral in this area,
and do we need to do the actual drilling or can we extrapolate
from these tools and these technologies?
Have at it.
Ms. Fu. Thank you so much. And these are all super-
important questions and real application spaces for AI. And I
will just give a few examples of where this might be useful.
Climate modeling and environmental research--you mentioned
earlier the ability to look at vast datasets from the Arctic
region, including sensor and satellite data to understand
permafrost thaw and changes in the ecosystem. This is work that
PNNL and Oak Ridge are already engaged in.
AI for energy resilience--this is an area where we look at
how we can optimize energy systems for remote communities. And
we know that NREL is doing quite a lot of work in that space
where we can look at energy demand, plan out where storage
should be and reduce dependence on costly diesel fuel. These
are things that I think can be very impactful for everyday
people.
There is also, obviously, AI for infrastructure monitoring,
and this is work that Sandia has underway, looking at AI to
monitor critical infrastructure, like pipelines and
transportation, especially in remote areas where it will be
hard for people to go to regularly.
So to your point around data, this is an incredibly
important issue. We talked a little bit about what we are doing
on AI and permitting data, especially NEPA data, making it AI-
ready. I think it was millions and millions of pages that we
were able to make AI-ready so that it could be used more
broadly for tools and application development. I think there is
an opportunity to do this across this entire spectrum, from
climate data, energy data, the whole grid--the grid questions
here--and knit them together. And that is part of what we are
thinking about through FASST. How do we get all of the
different stores of data--data we collect, as well as public
data--and make them actually useful, because I think poems are
great, and these tools that we have commercially available
today are really great for efficiencies and they are really
great for language, but the kinds of problems that we have are
going to be hard scientific problems that will then need to be
applied to real-world situations.
Senator Murkowski. Well, I appreciate that, and I am
thinking about the application to the Department of Defense.
Right now, on Alaska's coastline, we have some older military
assets that are greatly threatened by what we are seeing with
coastal erosion. And you have to make decisions, expensive
decisions about do we stay there, do we relocate, and you want
to make sure that the decisions that you are making are going
to be in place for more than just the next decade. It has to be
more long-term, so, the tools that we have to put in place.
Dr. Kaushik, you wanted to----
Dr. Kaushik. Sure, Senator, I think that is exactly the
kind of work we should be doing at the Federal Government level
where it's not necessarily like the private sector has
incentives to build those kinds of tools. They don't have the
data. They don't have the sensitive data. I think, for
instance, NOAA collects about a terabyte of data every day that
nobody ever gets to see. I think that is something that we can
be doing, but at the same time, there are some private-sector
actors who are working on this, like NVIDIA developed a digital
twin of the Earth, right? And what that allows them to do is
conduct millions of simulations of extreme weather events,
compared to what we can today do, which is like just about
thousands of simulations to see, to better predict the weather.
Another area where I think we need to think about what the
national labs could be doing is, like the Pacific Northwest
National Laboratory is working on this idea called Cloud Labs,
right? We have changed dramatically how we work in every sector
since the industrial revolution, but science has still stayed
the same. You go into a lab, you prep your chemicals, you do it
how you do it. A cloud lab allows you to automate how to do
scientific discovery. You can run experiments at scale. They
are more applicable. They can be run faster, with more
precision. And I think the same goes with how we explore
critical minerals. It's a very CapEx-heavy stage at the
exploration level. And I think USGS has a program on that where
you are using AI along with radar data to understand better
what kinds of materials may be underneath the Earth.
And I think all of those things, not necessarily--the
private sector has near-term incentives to invest in or
develop. And I think those are the kinds of things that
generally people look to the Federal Government for.
Senator Murkowski. Dr. Gleason.
Dr. Gleason. I just wanted to touch on something Helena
mentioned, which is the data problem, and you touched on it as
well. Making sure that data is AI-ready is a huge undertaking.
A lot of people underestimate the challenge. You know, you
could have a lot of data, but it's not valuable until it is
made AI-ready. It has to be organized. It has to be labeled in
some cases, sometimes not, but it has to be ready to train an
AI system. So there is a huge part of FASST which is to make
data AI-ready, and public data, scientific data from the
national labs, a very important challenge that's not the most
exciting piece of the work, but it's really the most important
because data is the fuel for the AI engine.
The other thing I just wanted to mention, just as an
example, is something we are doing in disaster response and
disaster recovery is, Oak Ridge is stitching together satellite
images from across the entire world and then mapping the actual
building infrastructure from those images, in fact, to the
point where we are predicting some of the materials that those
buildings may be made of from the spectral reflections and the
information in the satellite image. The cool thing about that
is, now you can, if you see a disaster, predicted disaster,
from a climate model or even a military situation, you can
understand what the affects might be, and after it happens you
can do a before and after comparison and decide where do I need
to direct my emergency resources most effectively, most
quickly, to save the most people, restore function, et cetera.
So AI is a huge tool to help those kinds of things that are
connected to some of the things you mentioned earlier.
Senator Murkowski. Fascinating.
Senator King [presiding]. On behalf of the Chairman, the
distinguished former Governor and current Senator from the
State of North Dakota.
Senator Hoeven. Thanks, Governor, appreciate it.
Thanks to all the witnesses for being here. And my question
for all of you is, you know, there are probably a lot of
metrics we use to measure AI. What are the most relevant
metrics? For example, if we are saying, okay, how do we compare
it to China or someone else in terms of where we are with AI,
what are the metrics that we use to measure who is ahead and
why it matters?
And then my second question is going to be how do we really
manage the security, not only for ourselves, but you know, how
do we address the security issues for others that are
developing AI, be it China or anyone else? And you know, you
have this great workforce and you get them, you know, all
trained up on this stuff and they develop it, and I am guessing
there are a lot of folks in a lot of other countries that
decide they want to hire them and are offering some pretty big
wages and so forth to have them come over and it may be for
malign purposes.
So first, the metrics, how you measure them, why it's
important, and then the security aspects, both for us here and
for our adversaries and particularly regarding the workforce
that you are developing and training.
Ms. Fu. Thank you so much for that question. These are all
things that we are thinking very deeply about at the Department
of Energy. On metrics, the AI executive order refers to how
much power is used to train the model, or FLOPS. So in the AI
executive order it refers to 10 to the 26th as the amount of
power that is used to train some of the most powerful models
that are here today. Now, is power or the amount of money that
you spend to train a model a good metric for capability?
Senator Hoeven. It's just an input.
Ms. Fu. It's imperfect. It's imperfect, and we recognize
that. And I think what we are doing at the Department and
across the interagency as well is thinking about what other
ways of identifying model capability there are. I mean, one way
to look at it is the data that is used to train the models.
It's not always the size that counts, it's how useful that
model is. And I will point to work that is underway across many
of the national labs, but work that we are doing on red-teaming
of open-source as well as proprietary models through our
National Nuclear Security Administration. This is an area of
intense work where we are pairing our data scientists with our
experts in radiological and nuclear expertise to really
understand how capable some of these models are relative to
each other. And that is work that we hope to bring to the work
of the AI Safety Institute to help inform that effort.
More broadly, around the workforce issues, I think this is
incredibly important. We have a managed research environment
within the DOE national labs to look at risk across a
continuum. Of course, for classified work, we have extreme
restrictions on that. We have a science and technology risk
matrix that looks across different areas of critical and
emerging technology and we use that to help guide who works on
what kinds of projects. We also screen foreign nationals who
come into the labs.
Now, to your point of, well, once they are in the labs and
what if they leave, what if they take that knowledge, the
training and go somewhere else, to industry or elsewhere? And I
think that is part of the U.S. open ecosystem. We can't control
where people go. We can take measures while they are in our
system. And I think it calls for why the capabilities that we
have at DOE attract talent from all over the world, because
they are unique facilities. We have people at the labs who come
to the United States because of that facility, because of that
supercomputer, or because of that neutron source. And the
investments that we are making at DOE and our national labs
continue to keep that center of gravity for talent here in the
United States.
Senator Hoeven. Yes, I mean, I would argue that, you know,
China or somebody else could actually want to develop people
that come here and train under you, learn everything you have
got, all your advanced technologies, and go back home and make
sure that they have access to all that information and
everything else and that capability.
Ms. Fu. So I would say we are extremely clear-eyed about
those threats, and that is why we have a managed research
environment. That is why we work very closely with our Office
of Intelligence and Counter Intelligence. We have our eyes wide
open to those things and that is why we focus on where people
have access to, even when they come to the lab, the kind of
research that they are focused on, the kind of access that they
have at the national lab. It's not a free-for-all once they
come through. We have ongoing efforts underway. We have
training. People understand what those risks are and there is
an ecosystem around our labs to make sure that we are managing
those risks properly.
Senator Hoeven. Are you seeing people leave and go to other
countries?
Ms. Fu. I mean, we are seeing people leave and go to other
companies. There is really a competition for talent in these
areas of critical and emerging technology. If you speak to
companies, they will say they can't hire enough people to do
that. And I think if you talk to our national labs, they will
say the same. I think the things that attract people to DOE are
mission and the access to the kinds of resources that we have.
But that leadership is not assured. That leadership is not
assured, and people--other companies--are paying quite a lot
more than what DOE national labs are paying, but people come to
us because of the mission, the work that we do.
Senator Hoeven. Right.
Dr. Gleason. Just as a leader of a few organizations where
I have tried to hire and retain talent across these emerging
technologies, it is a big challenge. You know, recruiting--we
need to increase our domestic workforce, the supply. In terms
of retention, my personal experience is we have lost less to
other countries, I can't even think of specific examples, but
we lose a lot to industry. And I would rather lose to industry
in the U.S. that is, you know, creating new AI technology that
will advance the cause of the United States, but it is a
challenge. And as Helena said, the ones that stay at the
national lab for the long haul are the ones that are motivated
by the mission. They love the lab mission and they love to work
at a place where they can explore science and technology
objectively without a profit motivation, which is a good
motivation, but that's not why you come to the national lab.
So I just wanted to echo what she said our challenge is,
which is recruiting domestic talent with, you know, U.S.
citizenship in the STEM fields. So one thing Oak Ridge is doing
a lot, and other labs are doing this as well, is engaging in
middle school, in high school. By the time they get to college,
you are almost too late, right, because they have already
decided their path. But having our scientists go mentor at
local schools to try to encourage young men and women to enter
STEM fields because we have such a huge shortage of that. So I
think that is a big thing we need to focus on.
Dr. Kaushik. I will just add to that, Senator. On metrics,
I think no metric will give you a perfect picture. Every metric
is targeting a different conversation, like, when you look at
papers, you are asking about who is ahead in more basic
research. When you are looking at commercialization, how many
users are using iFLYTEK AI in the PRC actually outpaces how
many people are using AI in the United States. But at the same
time, nobody can dispute that the capabilities of our frontier
AI model providers, whether it's Open AI, Anthropic, Meta and
all those, are undisputedly in the lead. That said, their
development of DeepSeek, which used to be, by the way, a
financial firm in China, which the Chinese government said you
have to work on AI now. And now they are churning out such
amazing models--they actually publish all their results, and we
can see that their results are actually very competitive to our
models. You look at Huawei's new AI chip, which outperforms
NVIDIA's A100 on several metrics. So there is an aggregate of
statistics that we probably need to be looking at here, an
aggregate of metrics.
I think on the talent side, like, I think the labs are
really great at recruiting mission-driven, focused people. At
the same time, like, you can put, you know, you can have all
the fancy knives in a restaurant, but if you don't have the
Michelin-star chefs, you are not going to make a good meal. And
I think that is the challenge the labs have to deal with.
Senator Hoeven. Thank you, all three of you.
The Chairman [presiding]. Thank you, Senator.
Senator Daines.
Senator Daines. Chairman Manchin, thank you, and thanks for
holding this hearing on my bipartisan Department of Energy
Quantum Leadership Act, which will reauthorize and strengthen
DOE's programs under the National Quantum Initiative.
In 2018, this Committee and Congress passed the National
Quantum Initiative Act in order to focus research, development,
and encourage commercialization of the next generation of high-
powered computing. That bill has been very effective in
spurring quantum research in the United States and in Montana.
Since passage, Montana has seen a surge of economic development
in jobs surrounding the quantum supply chain. The smartest
people want to find the best places to also live and work, and
Montana fits that bill. In fact, Montana now boasts over 50
companies and 1,200 employees focused solely on quantum and
photonics. Montana State University is home to the MonArk
Quantum Foundry, which focuses on quantum materials research.
Through the CHIPS and Science Act, Montana hosts the Headwaters
Technology Hub, focused on smart photonics, which is a key
component in the future of quantum technology. Simply put, and
you may not be aware of this, in Montana right now, we are
helping lead the world in quantum and photonics, which is why I
am proud to say we have strong support from Montana and other
national leaders to reauthorize the DOE Quantum Initiative,
including the Montana Photonics and Quantum Alliance, Montana
State University, the Montana Chamber of Commerce, the Energy
Sciences Coalition, the Quantum Industry Coalition, the Quantum
Economic Development Consortium, and in fact, many more.
Chairman Manchin, I ask unanimous consent to enter into the
record the many letters of support I have received for the DOE
Quantum Leadership Act.
The Chairman. You want an answer on that?
Senator Daines. I need a yes.
The Chairman. Yes.
Senator Daines. Thank you.
[Laughter.]
[Letters of support for the DOE Quantum Leadership Act
follow:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Senator Daines. It's who you know in this business, isn't
it, Joe?
Thank you, Joe, I appreciate it--Senator Manchin--Chairman
Manchin.
Anyway, thank you.
The Chairman. Still Joe.
Senator Daines. And it's still Joe and I am still Steve,
too.
The Chairman. I know.
Senator Daines. We must continue to focus our energy on
developing the best technology and leading the world in quantum
research, and this starts with passing my bipartisan bill. So I
want to thank Chairman Manchin for entering those letters into
the record and his partnership and support of this as well, and
I yield back my time to the Chairman.
The Chairman. Thank you, Senator.
First of all, before I adjourn, and thank you all for a
great job and it was really wonderful, you know, just so much
to do here. I want to make a clarification on--I know I put you
all on the spot. I wasn't asking you to basically disparage on
our sister NSF. What I was asking is, for the record, we wanted
to avoid duplication. When we did the CHIPS and Science Act at
the highest of DOE and the labs and intended to do the same
with AI and other emerging technology, we were trying to avoid
that duplication, and DOE, the labs, we were working on their
behalf. I want to make sure that we have that balanced approach
and we are not robbing Peter to pay Paul and trying to
reproduce the same thing again, dual application. We just, I
don't think any Democrats or Republicans want that to happen.
So with your assurances that's not happening, and that you
have the resources and you are working with the NSF to try and
have a balanced approach with what they do and can do best for
you to support what you are doing and vice versa is what we are
trying to make sure happens. We just don't want you to think
you have to be muffled on this. If you want to speak out, this
is the place to do it.
But with that being said, I believe that this hearing
demonstrated consensus that we need to continue investing in
and leading these emerging technologies in a way that leverages
DOE's assets to avoid reinventing the wheel, as we have said,
or duplicating between agencies. We worked to avoid duplication
in the CHIPS and Science Act, and at the behest of DOE and
labs, intend to do the same here in AI. We also agree on the
need to secure the research from foreign espionage, and I
believe that Senator Barrasso and I prepared a solid bipartisan
compromise with the Intelligence Committee to do just that in
the NDAA, which will complement our efforts on this Committee.
It's a scalpel approach that protects our critical research
while continuing our labs to leverage the best scientific minds
in the world. We are concerned about not only the espionage
that goes on, but also people that are working within, and we
are training them to use it against us, but we understand the
delicate thing that you have to work with in the freedom-loving
country that we are and the democracy that we rule ourselves
under.
So members will have until close of business tomorrow to
submit additional questions for the record.
And the Committee stands adjourned.
[Whereupon, at 11:57 a.m., the hearing was adjourned.]
APPENDIX MATERIAL SUBMITTED
----------
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
[all]