[House Hearing, 116 Congress]
[From the U.S. Government Publishing Office]
[H.A.S.C. No. 116-90]
INTERIM REVIEW OF THE NATIONAL
SECURITY COMMISSION ON ARTIFICIAL
INTELLIGENCE EFFORT AND RECOMMENDATIONS
__________
HEARING
BEFORE THE
SUBCOMMITTEE ON INTELLIGENCE AND EMERGING THREATS AND CAPABILITIES
OF THE
COMMITTEE ON ARMED SERVICES
HOUSE OF REPRESENTATIVES
ONE HUNDRED SIXTEENTH CONGRESS
SECOND SESSION
__________
HEARING HELD
SEPTEMBER 17, 2020
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
__________
U.S. GOVERNMENT PUBLISHING OFFICE
42-135 PDF WASHINGTON : 2021
--------------------------------------------------------------------------------------
SUBCOMMITTEE ON INTELLIGENCE AND EMERGING THREATS AND CAPABILITIES
JAMES R. LANGEVIN, Rhode Island, Chairman
RICK LARSEN, Washington ELISE M. STEFANIK, New York
JIM COOPER, Tennessee SAM GRAVES, Missouri
TULSI GABBARD, Hawaii RALPH LEE ABRAHAM, Louisiana
ANTHONY G. BROWN, Maryland K. MICHAEL CONAWAY, Texas
RO KHANNA, California AUSTIN SCOTT, Georgia
WILLIAM R. KEATING, Massachusetts SCOTT DesJARLAIS, Tennessee
ANDY KIM, New Jersey MIKE GALLAGHER, Wisconsin
CHRISSY HOULAHAN, Pennsylvania MICHAEL WALTZ, Florida
JASON CROW, Colorado, Vice Chair DON BACON, Nebraska
ELISSA SLOTKIN, Michigan JIM BANKS, Indiana
LORI TRAHAN, Massachusetts
Bess Dopkeen, Professional Staff Member
Eric Snelgrove, Professional Staff Member
Caroline Kehrli, Clerk
C O N T E N T S
----------
Page
STATEMENTS PRESENTED BY MEMBERS OF CONGRESS
Langevin, Hon. James R., a Representative from Rhode Island,
Chairman, Subcommittee on Intelligence and Emerging Threats and
Capabilities................................................... 1
Stefanik, Hon. Elise M., a Representative from New York, Ranking
Member, Subcommittee on Intelligence and Emerging Threats and
Capabilities................................................... 3
WITNESSES
Clyburn, Hon. Mignon, Commissioner, National Security Commission
on Artificial Intelligence..................................... 12
Griffiths, Dr. Jose-Marie, Commissioner, National Security
Commission on Artificial Intelligence.......................... 11
Schmidt, Dr. Eric, Chairman, National Security Commission on
Artificial Intelligence........................................ 5
Work, Hon. Robert O., Vice Chairman, National Security Commission
on Artificial Intelligence..................................... 9
APPENDIX
Prepared Statements:
Langevin, Hon. James R....................................... 35
Schmidt, Dr. Eric, joint with Hon. Robert O. Work, Hon.
Mignon Clyburn, and Dr. Jose-Marie Giffiths................ 38
Documents Submitted for the Record:
[There were no Documents submitted.]
Witness Responses to Questions Asked During the Hearing:
Mr. Langevin................................................. 59
Questions Submitted by Members Post Hearing:
Mr. Waltz.................................................... 63
INTERIM REVIEW OF THE NATIONAL SECURITY
COMMISSION ON ARTIFICIAL INTELLIGENCE
EFFORT AND RECOMMENDATIONS
----------
House of Representatives,
Committee on Armed Services,
Subcommittee on Intelligence and Emerging Threats and
Capabilities,
Washington, DC, Thursday, September 17, 2020.
The subcommittee met, pursuant to call, at 1:00 p.m., in
room 2118, Rayburn House Office Building, Hon. James Langevin
(chairman of the subcommittee) presiding.
OPENING STATEMENT OF HON. JAMES R. LANGEVIN, A REPRESENTATIVE
FROM RHODE ISLAND, CHAIRMAN, SUBCOMMITTEE ON INTELLIGENCE AND
EMERGING THREATS AND CAPABILITIES
Mr. Langevin. The subcommittee will come to order.
I want to welcome members participating in today's hearing
and, for those remotely, members who are joining remotely must
be visible on screen for the purposes of identity verification,
establishing and maintaining a quorum, participating in the
proceedings, and voting. Those members must continue to use the
software platform's video function while in attendance, unless
they experience connectivity issues or other technical problems
that render them unable to participate on camera. If a member
experiences technical difficulties, they should contact the
committee staff for assistance.
A video of members' participation will be broadcast in the
room and via the television's internet feeds. Members
participating remotely must seek recognition verbally and they
are asked to mute their microphones when they are not speaking.
Members who are participating remotely are reminded to keep the
software platform's video functions on the entire time they
attend the proceeding.
Members may leave and rejoin the proceeding. If members
depart for a short while, for reasons other than joining a
different proceeding, they should leave the video function on.
If members will be absent for a significant period or depart to
join a different proceeding, they should exit the software
platform entirely and then rejoin if they return.
Members may use the software platform's chat feature to
communicate with staff regarding technical or logistical
support issues only.
Finally, I designated a committee staff member to, if
necessary, mute unrecognized members' microphones to cancel any
inadvertent background noise that may disrupt the proceeding.
Before I go to my opening statement, I understand that
there will be votes during the course of the hearing, very
likely. So if that occurs, we are going to keep the hearing
going is what has been worked out and, unless it becomes an
issue, we will see members who will go and vote and they will
return as soon as possible but the hearing will continue.
So with that, I am going to now give my opening statement.
Let me say that I am pleased to welcome four commissioners from
the National Security Commission on Artificial Intelligence, a
commission created by this committee in the John S. McCain
National Defense Authorization Act for Fiscal Year 2019, to
consider the methods and means necessary to advance the
development of artificial intelligence, machine learning, and
associated technologies to comprehensively address the national
security and defense needs of the United States.
Our intent with this Commission was to ensure a bipartisan
whole-of-government effort focused on solving national security
issues and we appreciate the leadership and the hard of work of
our witnesses in supporting the Commission's efforts in that
spirit.
Today, we welcome Dr. Eric Schmidt, chairman of the
Commission, the Honorable Robert Work, vice chairman, the
Honorable Mignon Clyburn, commissioner of the workforce and
ethics lines of effort, and Dr. Jose-Marie Griffiths,
commissioner on the ethics line of effort and the chair of the
workforce team.
I want to thank you all, first of all, for your service, as
well as your other commissioners and look forward to hearing
your testimony today.
Our understanding of artificial intelligence started in the
1950s and 1960s through research funded through the Department
of Defense's vital science and technology investments by the
Defense Advanced Research Projects Agency, or DARPA, and the
Office of Naval Research and was aided by the convening power
of universities.
Now, more than half a century later, this Commission is
working through the difficult issues requiring national
investments in research and software development and new
approaches on how to, among other things, apply AI [artificial
intelligence] appropriately for national security missions,
attract and hold onto the best talent, protect and build upon
technical advances, best partner with our allies on AI and stay
ahead of the threat posed by this technology in the hands of
our adversaries, and implement ethical requirements for
responsible American-built AI.
Indeed, last year, the Defense Innovation Board, which was
also chaired, until recently, by Dr. Schmidt, helped the
Department begin the necessary discussion on ethics in AI. Dr.
Schmidt, I want to thank you for the 4 years that you led the
Defense Innovation Board and I look forward to working with you
to make sure that you continue to be able to serve, in some
capacity, with the Defense Innovation Board. You are an
invaluable resource and we can't lose you.
So I applaud the Commission for being forward-leaning by
not only releasing an initial and annual report, as required by
law, but also releasing quarterly recommendations. Ranking
Member Stefanik and I, along with Chairman Smith and Ranking
Member Thornberry, were pleased to support a package of
provisions in this year's House version of the fiscal year 2021
NDAA [National Defense Authorization Act] based on the
Commission's first quarter's recommendations. The House version
carried 11 provisions, with the majority deriving from the
Commission's call to strengthen the AI workforce. We are
pleased that both Commissioner Griffiths and Commissioner
Clyburn are with us today to testify on the need for action on
AI talent.
On that note, we must implement policies that promote a
sound economic, political, and strategic environment on U.S.
soil, where global collaboration, discovery, and innovation can
all thrive. The open dialogue, in-depth resident in academia
and the research community can be an anathema to the
requirement for secrecy in the Department of Defense but we
must recognize and embrace how our free society provides the
comprehensive advantage that lets us innovate faster than our
great power competitors. Our free society enables a dynamic
innovation ecosystem and federally funded open basic research
focused on discovery has allowed American universities to
develop an innovation base that has effectively functioned as a
talent acquisition program for the U.S. economy that is second
to none. And that talent is required today, as much as ever, to
solve our most pressing national security challenges.
Indeed, great power competition is also a race for talent.
With that, we are looking forward to hearing about your
efforts, the observations and recommendations you have already
developed, and your plan to continue, until you submit the
Commission's final report in the spring.
With that, before turning to our witnesses, I will now turn
to Ranking Member Stefanik for her remarks. She has been an
outstanding leader on the issue of AI and I am proud to partner
with her on this whole effort.
I would like to now recognize Ranking Member Stefanik for
her comments.
[The prepared statement of Mr. Langevin can be found in the
Appendix on page 35.]
STATEMENT OF HON. ELISE M. STEFANIK, A REPRESENTATIVE FROM NEW
YORK, RANKING MEMBER, SUBCOMMITTEE ON INTELLIGENCE AND EMERGING
THREATS AND CAPABILITIES
Ms. Stefanik. Thank you, Chairman Langevin.
Welcome to our witnesses, Chairman Schmidt, Vice Chairman
Work, and Commissioners Clyburn and Griffiths. It is great to
have you before the subcommittee today. Thank you for all of
your continued service on this Commission.
I would be remiss if I didn't also thank Yll [Yll
Bajraktari] for his incredible work as staff director. I know
we are working so closely with you with our subcommittee staff
and your team.
On March 20th of 2018, I introduced legislation in the
House of Representatives to establish a national commission to
review the advances in artificial intelligence, the
competitiveness of our efforts, and the implications to our
national security. Just a year later, I had the honor of
speaking at the AI Commission's first plenary session, meeting
many of you and providing my thoughts on the importance and
direction of the Commission's work. And just a few short months
ago, I had the privilege of sponsoring, alongside my friend and
colleague, Chairman Langevin, 11 amendments to the NDAA that
originated from the Commission's first quarter recommendations.
This is truly a remarkable achievement and demonstrates the
value of your findings and recommendations to policymakers, and
in particular, to this committee. This impressive commitment
reflects upon your hard work, the dedication of the staff, and
also a recognition of how important and timely this
conversation on artificial intelligence is to our national
discourse and national defense.
In my comments at the Commission's first session, I spoke
about the need for artificial intelligence to be
transformative. I had stressed that, if AI doesn't
fundamentally change the way we operate, how we view our
collective defense, adapt our workforce composition, shift our
priorities, and invest our resources, then we are failing to
embrace this new technology to its fullest. I am pleased that
many of your initial recommendations addressed these issues and
I look forward to hearing your comments on how we are doing in
these regards.
Over the last several weeks, we have seen glimpses into the
power of artificial intelligence. DARPA's AlphaDogfight
demonstration, which pitted an experienced Air Force pilot in a
virtual dogfight against an algorithm developed by a small
woman- and minority-owned business in Maryland. It was a
decisive victory for artificial intelligence and one that
Secretary of Defense Esper accurately observed as a, quote,
tectonic impact of machine learning on the future of
warfighting.
In another noteworthy demonstration, we observed a
hypervelocity weapon shoot down a cruise missile with the help
of an advanced battle management system powered by powerful
data analytics and AI capabilities. The head of Northern
Command noted afterwards, quote, I am not a skeptic after
watching today.
Equally important as these AI technical demonstrations is
the formulation of policy governing how we use these
capabilities. The development of standards, ethical principles,
accountability, and appropriate level of human oversight will
be critical to ensuring the American people trust its use. Your
work, both on the Commission and in your personal and
professional endeavors, is key to ensuring a strong and
enduring partnership between the military, academia, and
private sector; a partnership built on trust, democratic
ideals, and mutual value.
Again, I look forward to discussing the Commission's
recommendations and your priorities for the remainder of the
Commission's work. Thank you so much for your service and the
hundreds of hours you have dedicated to this effort.
I yield back.
Mr. Langevin. Thank you, Ranking Member Stefanik.
Let me now introduce our witnesses. We are pleased to have
with us today Dr. Eric Schmidt, chairman of the National
Security Commission on AI. Dr. Schmidt is the technical advisor
to the board of Alphabet, where he was formerly the executive
chairman. His previous roles included the chairman of Google
Inc. and CEO [chief executive officer] of Google. He has a
distinguished record of contributions to the national security
technology community, including recently chairing the Defense
Innovation Board.
Dr. Schmidt, as a commissioner on the Cyberspace Solarium
Commission, I would like to begin by thanking you for your
commitment to ensure the two Commissions work closely together
and all that you have done to make the AI Commission so robust.
Next, we will hear from the Honorable Robert Work, vice
chairman of the Commission. Secretary Work is familiar to many
of us on the committee, as the former Deputy Secretary of
Defense and the Under Secretary of the Navy before that.
Secretary Work's commitment to innovative strategic thinking is
well known with his related work on the Third Offset Strategy.
Thank you for being here, Commissioner Work.
Next, we will hear from the Honorable Mignon Clyburn.
Commissioner Clyburn has spent 9 years on the Federal
Communications Commission, where her commitment to closing the
digital divide was well known. She has had a distinguished
career, fighting for diversity in the communications sector.
Thank you for being here as well, Commissioner Clyburn.
And finally, we have Dr. Jose-Marie Griffiths. Dr.
Griffiths is the president of South Dakota University.
Dr. Griffiths, first of all, I want to thank you, again,
for hosting me and my fellow Solarium commissioners 2 weeks ago
to release our white paper on the Federal Cybersecurity
Workforce. As you and I both know, our institutions of higher
education are vital resources in educating the digital natives
that we need to help us meet the AI and cybersecurity
challenges that we will face in the coming decades.
So with that, I again want to thank our witnesses for being
here today and I will turn now to Chairman Schmidt to summarize
your comments for 5 minutes.
Chairman Schmidt, the floor is now yours.
STATEMENT OF DR. ERIC SCHMIDT, CHAIRMAN, NATIONAL SECURITY
COMMISSION ON ARTIFICIAL INTELLIGENCE
Dr. Schmidt. Thank you so much. I cannot express how
grateful I am for the leadership of Chairman Langevin, Ranking
Member Stefanik, this Commission, and the things that I and our
Commission care so deeply about. It has been a remarkable year
working with you all to try to get these things going forward.
The progress we have made, in terms of improving the
situation of AI, is a good indicator of what is possible if we
continue to work very hard on this. I cannot say enough how
important this is.
I think addressing AI, in the way that we are describing,
is a unifying topic. It is a bipartisan priority. What is more
important than our national security? And when I hear that, I
say: What is more important than leadership in AI? I could go
on, and on, and on, to the point of boredom, I suspect, of how
AI is so exciting. I imagine, if I were a graduate student
today, the kind of amazing technologies and solutions I would
be able to provide using these new AI techniques that did not
exist when I was a computer scientist, as a younger scientist,
in particular, the application to biology, and to medicine, and
to health, and to the things that we all care and deal with so
much in our society.
There is a term in history called the Cambrian explosion
and it is the point in history where everything came together
to form modern life, and everything aligned at that point, and
we are in a similar position now with AI. These AI applications
will be the basis for the solution to the COVID pandemic. I
believe that, for example, the vaccines, essentially all the
ones I have looked at, have had AI as a core part of their
research enterprise.
I could just go on and on. Maybe it will help us plan how
to allocate the horrible fires and the resources. There are so
many areas where we struggle, where these new techniques can
make us more effective and efficient.
We have to understand, however, that there are darker sides
of this technology and, in particular, I will give you an
example of something I am hearing a lot about. AI systems are
trained from human behavior; humans have biases. And we don't--
we are Americans. We don't believe in prejudice and bias and so
we have to work on that. And, indeed, this is a large area of
research. Face recognition, for example, is full of biases that
are incompatible, today anyway, with the sort of rules of
America.
But I am also concerned, and I want to hit this very hard,
that the AI systems can be used in ways that really are counter
to how you want our country to evolve. It can supercharge
adversaries' disinformation campaigns. Most of the
disinformation campaigns that I have looked at have been done
by large groups of presumably poorly paid and badly managed
Russians. Imagine when the same technology is used in scalable
machine learning at a scale that is much more pervasive.
It is very clear, and Bob Work is an expert in this, that
AI could lead to forms of autonomous warfare. He will say, if
you talk to him, that it is fine to make the weapons more
effective but that you fundamentally don't want automatic
weapon systems that fire without human intervention. And
indeed, our military has a rule of human in the loop for that
reason.
We already know that authoritarian regimes, very
incompatible with our democracy, are using AI technologies to
try to consolidate power and homogenize thought; and
homogenizing thought gives you an army of sycophants and the
rest is history. And certainly, the technologies that are being
broadly distributed now could distribute this to terrorists,
and the future Osama bin Ladens, and sort of groups that we
just don't want access to this.
The other thing that is happening, and in my work with the
military I learned from them that they now view a very strong
strategic competition with China is on our plate. And I would
argue that China is no longer a near peer; they are a peer in
this area. They are close enough. And the Commission has spent
a lot of time discussing this--how close is it--but from my
perspective, within a year or two is close enough to be a
serious issue. And there is no question that if the Chinese
become leaders in AI, which, in most cases, they are not
today--perhaps in TikTok's algorithm but not otherwise--they
are going to use it in ways that are inimical to our country's
interest. So, we have got to take this really seriously.
So we are going to basically make and continue to make
strong recommendations to make AI for good but I want to say
right now that my approach, and I think the Commission's
approach, is very straightforward. We want America to win.
Right? It is really easy to articulate that way. We need to do
whatever it takes with respect to AI to be leadership.
One of you mentioned: Why don't we just set a goal of
leading and winning sooner? I completely agree. And part of the
reason that we talked about ethics was because we want to win
in a way that is compatible with American values, which you all
know and you all embrace.
So we have got a series of principles, which I will
highlight briefly. We have got to be global leaders in AI. It
is not okay if another country, specifically China but there
could be others, where they are the innovators ahead of us. Why
is this so important? And we can explore this, if you are
interested in it, because AI is a new knowledge and reasoning
system, it is at the beginning of every new area of inquiry. So
every new aspect of science, every new aspect of thought, every
new aspect of--every new thing now will start with AI as a
contributing accelerator with new data, new insights, and so
forth. That is why it is a pervasive technology. It is not like
a missile that just gets smarter. Everything gets smarter. It
has enormous systems implications to what we are doing and
probably, eventually, to society as a whole.
By the way, the government is important here. When I was a
graduate student, I was funded by DARPA and the National
Science Foundation. I wouldn't have been able to do it without
that funding. I didn't have the money. The remarkable
relationship that collectively you all established between
universities, the private sector, and the Federal Government,
primarily, some State governments, is at the root of American
exceptionalism in this area and I want to keep it. So, I cannot
express the importance of Federal funding in research and these
sorts of things. And we have talked about this before in this
committee and subcommittee, and I think everyone understands,
that the Federal Government funds the research that nobody else
can because it is not in their business interest. So, there is
a key role for Federal funding research.
Adopting AI for national security, as Congresswoman
Stefanik mentioned, is central. She used some examples that are
recent but there are example, after example, after example in
national security. The most obvious ones involve the
intelligence committees--community because they spend a great
deal of time with data and AI is very good at sorting through
data. I would much rather get heads-up from a computer system
that is constantly looking for threats and then have a human
say: Oh, that is interesting; I hadn't thought about that.
Right, that is what AI can do. That will keep us safe.
We have got to find ways, and we have some proposals, where
private sector individuals are flowing into and out of the
government and vice versa. The fact that that talent and
knowledge is in the private sector; we need it in the Federal
Government and we need the Federal Government people in the
private sector. We need to make that as easy as possible and we
have some recommendations there.
We are going to talk a lot about talent today. The majority
of the subsections of our recommendations, so far, have been
talent. After a while, when you work on this, you discover that
you can write as many papers as you want, but the fact of the
matter is that without the people who understand, this stuff is
hard. To be very honest, a lot of it is really hard. I have
Ph.D. in this area and it is hard for me. I can imagine what
somebody who is trying to struggle through all the
complexities. We need a next generation of talent and they need
to be in the government working for the Secretary of this, and
the Secretary of that, and the DOD [Department of Defense], and
the intelligence community, and working for you all on your
staffs, and so forth. You need that.
We are going to talk a lot about this and, indeed, Mignon
and Jose-Marie will go into it in some detail.
We really want to emphasize that we want to do this in an
American way--free inquiry, free enterprise, and the free flow
of ideas, right? The Chinese model is different; it is not
compatible with the way we work. There are other models. Let's
do this the American way.
And you guys, by the way, did a really good job in terms of
counterintelligence threat in research, taking action to
protect fields like microelectronics, which we are also very
worried about.
So, again, the government is beginning to understand this
and beginning to act correctly. What we need to do is we need--
we need to get the ethics stuff in agreement. I was part of a
team that did a DOD ethics group. I was also part of a team at
Google that did some ethics work. There is an emerging
consensus of what AI ethics looks like and we include that as
part of our report.
And then finally, I think, we need to win all of the tech
competitions, not just the AI ones. We have never had a
challenger at the level of depth and sophistication that China
represents in terms of their innovative capability. We need to
take it seriously, in terms of scale.
And I think, frankly, we should publish such a list. If you
were to ask me today, I would tell you the list of things that
are important are AI, obviously; biotechnology, the basis for a
gazillion dollars' worth of industry; quantum computing,
something which is hard to understand but incredibly important
for national security; semiconductors, huge fight over that;
5G, very important; and advanced manufacturing, a huge basis
for industry in our country.
But maybe there are others on that list and I think one of
the things that we all should collectively discuss is what that
list should be. And again, let me just emphasize, this has to
be all around, built around American values.
And I will finish up by saying that we have been working
hard with you and your staffs to translate these into specific
recommendations. What I have learned in this process is there
are all sorts of rules that govern how pieces of the government
work. You all knew this. And if we can adjust those rules to be
a little bit more focused on getting excellent AI techniques,
technologies, getting leadership, getting everybody talking to
each other, and all of that, the American model will not just
succeed but really thrive.
So, I want to thank you so much for letting me speak.
[The joint prepared statement of Dr. Schmidt, Secretary
Work, Ms. Clyburn, and Dr. Griffiths can be found in the
Appendix on page 38.]
Mr. Langevin. Thank you very much, Chairman Schmidt, for
your leadership and all the work you have done to lead this
Commission and give us a lot to think about.
With that, the chair now recognizes the vice chair of the
Commission, the Honorable Robert Work. Secretary, you are
invited to summarize your remarks for 5 minutes and, without
objection, your written testimony will be submitted for the
record.
Secretary Work, you may have to unmute your line.
STATEMENT OF HON. ROBERT O. WORK, VICE CHAIRMAN, NATIONAL
SECURITY COMMISSION ON ARTIFICIAL INTELLIGENCE
Secretary Work. Sorry about that, sir.
Chairman Langevin, Ranking Member Stefanik, and members of
the committee, thanks for the opportunity to testify today.
I would like to discuss the importance of capitalizing on
AI for our Nation's defense and intelligence capabilities, and
then discuss the Commission's view on this year's National
Defense Authorization Act in the current House and Senate
versions.
The Commission has found, in important ways and as Eric has
laid out, AI is going to change how we defend the American
homeland, how our intelligence agencies make sense of the
world, and how our military deters adversaries and fights on
future battlefields.
In the context of homeland security, we see promise in
applying AI to border protection, cyber defense, critical
infrastructure protection, counterterrorism, and
counterintelligence investigations. It will also be central to
countering malign information operations designed to create or
deepen fissures in our society and undermine confidence in the
electoral process.
In the intelligence realm, AI algorithms can sift through
vast amounts of data, define patterns, and identify
correlations, while automating imagery analysis, and other
labor-intensive analytical tasks.
For our military, AI-enabled autonomous systems open up a
vast new realm of possibilities for operational concepts and
command decision making that will give us advantages in any
fight. AI is going to enable new forms of what we call human-
machine collaboration, using machines to help humans make
better decisions, and human-machine combat teaming, both of
which will improve combat effectiveness. If employed
responsibly, we believe AI-enabled military systems can also
help reduce risk to U.S. service members in the field and
protect innocent lives abroad.
In addition, AI will make business functions of the
Department of Defense, the entire Federal workforce, for that
matter, and sector far more efficient and cost-effective.
The Commission is preparing recommendations in all of these
areas. We have moved with an urgency that is commensurate with
the opportunity and the national security threat presented by
AI. As you know, sir, we have released our interim report last
November that articulated a series of initial judgments of all
of the Commissioners and we have published over 80
recommendations since. We will publish more next month.
We will deliver our final report to Congress and the
President in March 2021. These are going to cover five key
areas: research and development, national security
applications, talent and workforce, promotion and protection of
critical technologies, international partnerships and ethics.
We are encouraged to see several of the Commission's early
recommendations reflected in both the House and Senate versions
of this year's NDAA and I want to comment on the importance of
the legislative action in five key areas.
The current bill encouraging actions to bolster government
investment in AI research and development, improve public-
private coordination, and establish technical standards. The
Commission shares these priorities and endorses them and
applauds them.
We want to emphasize the importance of creating a national
AI research resource. Right now, we have AI research haves and
have-nots. The haves are generally in the private sector and
the have-nots are in academia. We are very encouraged in the
new recent White House-led investments to establish seven
national AI institutes but we believe that the AI research
resource would complement and support these efforts.
In terms of defense, Department of Defense organization
reform, we have made several recommendations to make sure that
DOD puts the proper emphasis on AI and shepherds and monitors
the way it is transforming the force.
Microelectronics, we need to preserve our leadership in
this and we have put forth several recommendations to lay the
groundwork for long-term access to resilient, trusted, and
assured microelectronics.
The fourth area is ethical and responsible use. We have
spent a lot of time on this and Eric has talked about that, so
I won't dwell on it anymore.
And the fifth area is the Federal Government's AI
workforce. I am going to leave that up to Jose-Marie and Mignon
to discuss in detail.
So let me just stress we must grasp the inevitability of AI
and out-innovate, out-invest, out-strategize, and outwit our
competitors.
I thank this committee so much for devoting so much
attention to including AI in developing this year's NDAA. We
are extremely encouraged to see this process and we look
forward to working with the committee in the future.
Thank you.
Mr. Langevin. Very good. Secretary Work, thank you very
much for your testimony.
I understand the commissioners have worked out among
themselves that Dr. Jose-Marie Griffiths will go next. So, Dr.
Griffiths, the floor is now yours to summarize your testimony
for 5 minutes and, without objection, your written testimony
will be submitted for the record.
STATEMENT OF DR. JOSE09MARIE GRIFFITHS, COMMISSIONER, NATIONAL
SECURITY COMMISSION ON ARTIFICIAL INTELLIGENCE
Dr. Griffiths. Thank you very much. Chairman Langevin, and
Ranking Member Stefanik, and members of the committee, thank
you for the opportunity to testify before you today.
Within our Commission, I chaired our efforts to develop an
AI-ready Federal workforce and to improve the AI talent pool in
the United States more broadly. Over the last year and a half,
our workforce line of effort has held 11 working groups and
interviewed more than 150 AI and human capital experts from the
government, private sector, and academia. Through this process,
three broad themes have emerged about the government's
workforce.
First, building an AI-capable workforce doesn't lend itself
to neat and tidy solutions. We need to tackle the problem from
multiple angles. Second, it is difficult for agencies to
implement their own major workforce reforms. Anything other
than incremental change requires congressional leadership. And
third, every opportunity my colleagues have mentioned and every
challenge we describe in our report is, at its core, a
workforce issue.
When organizations fail to adopt AI, and it is almost
always because of their lack of qualified engineers and lack of
senior leaders with the right education and experience to
establish priorities and cut through red tape; when
organizations can't purchase the software and hardware they
need, it is often due to a problem with the limited knowledge
and understanding on the part of their acquisition and
contracting personnel. When organizations struggle to collect
and manage data, it often suggests a lack of training and
education geared towards these complex tasks.
To better understand the composition of the workforce the
government needs, we partnered with the Defense Innovation
Board and the Joint AI Center to create an AI workforce model,
which you can find in our interim report.
In broad strokes, we believe the government should focus on
five things: (1) build a technical workforce with tiered levels
of skill and educational requirements; (2) educate senior
leaders, who can better define strategic and enterprise
objectives; (3) train junior leaders, who will manage the
deployment and use of AI-enabled technologies and capabilities;
(4) train the end users of AI-enabled technologies, who will be
responsible for collecting and managing data; and (5) train and
educate people in critical support roles, including human
resource, acquisition, contracting, and legal professionals.
Our early recommendations are meant to help set a
foundation for Federal workforce improvements and we are
encouraged to see so many of them reflected in this year's
NDAA. I would like to highlight several provisions in
particular that we strongly support. They include AI training
courses for HR [human resources] professionals, the creation of
unclassified workspaces, a pilot program for using electronic
portfolios to evaluate applicants for technical positions, a
program to track and reward the completion of AI training, a
mechanism to hire university faculty on a part-time basis in
government laboratories, expanding talent exchange programs
between DOD and technology companies, and an adjustment to the
aptitude test that the armed services use so that it tests for
computational thinking skills.
In combination, these reforms would mark a significant step
forward and I urge Congress to ensure they are included as part
of this year's defense authorization.
Thank you, again, for the opportunity to appear here today
and I look forward to your questions.
Mr. Langevin. Thank you very much, Dr. Griffiths.
The chair now recognizes Commissioner Clyburn for your
testimony for 5 minutes and, without objection, your written
testimony will be submitted for the record.
STATEMENT OF HON. MIGNON CLYBURN, COMMISSIONER, NATIONAL
SECURITY COMMISSION ON ARTIFICIAL INTELLIGENCE
Ms. Clyburn. No objection. Thank you very much.
Chairman Langevin, Ranking Member Stefanik, and members of
the committee, thank you for the opportunity to testify today.
I would like to use my time to continue the theme that my
colleague, Dr. Griffiths, has just discussed, the state of the
Federal Government's AI workforce. In my time on the
Commission, it has become clear to me that talent is the
centerpiece of any winning AI strategy. We have examined the
government's current shortcomings and have found that, in
addition to the series of reforms Dr. Griffiths mentioned, we
need to take bolder action. Existing programs will not bring
enough digital talent into the public service workforce to meet
serious shortages. The current scholarship and service programs
are limited in scale and will not create a common set of ideas,
shared experiences, professional culture, or a common mission
to improve the government's digital talent. So, we must
fundamentally reimagine the way the U.S. Government recruits
and builds its digital workforce.
The Commission has put forward two significant proposals
and I will take these few minutes that I have left to briefly
describe them. First, we propose building a United States
Digital Service Academy. This academy will produce technically
educated graduates who would have a service obligation as civil
servants into Federal Government. The academy will be an
independent entity within the government. It would be advised
by an interagency board, which would be assisted by a Federal
advisory committee composed of commercial and academic leaders
in emerging technology. The academy would be a partnership
between public and private sectors, working together toward a
common goal of developing a modern digitally proficient
workforce.
We should consider now, before legislative action takes
place, how the private sector and academia can support an
academy. We are eager to discuss what barriers, limitations, or
other factors would prevent such cooperation, or work with
legislators to ensure that language is written with this
partnering in mind.
Second, we propose establishing a National Digital Reserve
Corps. Many of the most talented technologists in the United
States are eager to serve their country but are unlikely to
become full-time government employees or military reservists.
The government needs a mechanism to tap this talent reservoir.
The government should establish a National Reserve Digital
Corps modeled after the military Reserves that allows
individuals to work for government 38 days a year as advisors,
instructors, and developers. We could incentivize participation
with a training and education fund, and a scholarship program
modeled after ROTC [Reserve Officers' Training Corps]. While
short-term volunteers are not a substitute for full-time
employees, they can help improve AI education for both
technologists and non-technical leaders, perform data triage
and acquisition, help guide projects and frame technical
solutions, build bridges between the public and private sector,
and other important tasks.
I urge Members of Congress to take both of these proposals
into consideration and to develop the legislation that would be
needed to turn bold ideas into real institutions and programs.
Thank you again for the opportunity to share our
recommendations with you and I look forward to any questions
you may have.
Mr. Langevin. Very good. Thank you, Commissioner Clyburn.
I want to thank all of our witnesses here today for your
extraordinary contributions to the National Security Commission
on AI. We appreciate the testimony today and the written
testimony that you have submitted.
And we will now turn to recognizing members for questions
for 5 minutes. I will begin with myself.
Let me start with you, if I could, Chairman Schmidt. Given
that China views talent as central to its technological
advancement, U.S. policies that restrict foreign talent from
studying and working in the U.S. seems to play right into
China's hands. How do policies that restrict China's talent
from studying and working in the U.S. impact our national
security?
Dr. Schmidt. Well, thank you, Mr. Chairman.
I was quite surprised, when we looked at the quality of the
top papers, how many Chinese graduate students were part of the
top papers being produced in the United States. So in other
words, if you were to get rid of them, if you were to say none
of them are allowed in the U.S., U.S. research would suffer.
I don't know if it is appropriate or not, but I need to say
that this Pathway Act that you guys are proposing is exactly a
good answer to this because we need to identify the very tip-
top people we need for national security from all countries,
and we need to get them into America, and we need to keep them
here, and we need to keep them here producing research wins,
producing defense companies, producing high-tech companies, and
so forth, and so on. Put another way, if those people are--
using China as an example--if they are in China, they are going
to start up a whole bunch of companies that are going to become
a real pain in the ass, if it is okay to say that, for the
Congress in a decade.
So in other words, I would much rather have them creating
huge successes in America, for security and also for commercial
reasons, than doing the same thing in China or another country,
such as Russia, and then us having to deal with the
consequences of that.
And just to make it clear, if you look at TikTok, the core
achievement of TikTok, although it is a social phenomenon, is a
different kind of AI recommendation algorithm where they are
clearly ahead. The moment we started arguing with TikTok over
their U.S. operations, the Chinese Government banned the export
of that algorithm. How important is that? I don't know but it
is a good example of something that would have been available
to U.S. researchers that is not available today. That is not a
good thing.
Mr. Langevin. Thank you. I share your concern. And the
Pathways Act is a good remedy, again, arm the Secretary of
Defense to designate critical study areas where we would want
to keep that talent here. We would start out with 10
individuals but we would have liked to--very much like to
rapidly expand it to a larger number from there.
The next question: What are any specific AI research areas
in which you believe the United States is under- or over-
invested and how would you propose rebalancing the U.S. science
and technology investments?
Dr. Schmidt. So, my personal opinion--people can disagree
over this--is that we are not over-invested in anything. It is
clear that we are under-invested in the underlying
infrastructure that is needed.
There is a proposal that is, I think, being discussed in
the Congress called the National AI Research Resource. And the
idea here, it is a good idea, it is basically to try to create
an infrastructure that allows all of the creative people in the
United States access to the systems where they can do the
research. I can tell you that, if you work in a very large
company, you have that. But what about all of the 10-person and
20-person companies, where they don't have the money, and the
scale, and the time to get the kind of data, and data analysis,
and computing platforms? What about the researchers, the three
people in, you know, a little boat somewhere in their
university, who don't have access but they have a brilliant
idea?
One of the hallmarks of American creativity has been that
the greatest things come from the weirdest corners. We want to
make sure that those weird corners in the United States of sort
of clever people, who are staying up all night drinking Diet
Coke and eating hamburgers, whatever the stereotype you have
is, they have the tools that they need to do global solutions
very quickly.
Mr. Langevin. Thank you. And understanding that many of
your recommendations are focused on government-wide, what are
the other committees that you are meeting with to help
implement these provisions?
Dr. Schmidt. I am sorry. The committees of the Congress?
Mr. Langevin. That is right. Yes, what other committees of
the Congress are you meeting with to help implement these
provisions?
Dr. Schmidt. In general, this area on our remit is
controlled by the HASC [House Armed Services Committee] and the
SASC [Senate Armed Services Committee].
Mr. Langevin. Okay, all right.
Secretary Work. Mr. Chairman?
Mr. Langevin. Thank you very much.
Secretary Work. Mr. Chairman, if I could add.
Mr. Langevin. Of course.
Secretary Work. We have worked very closely with the
intelligence committees on several of the recommendations.
[audio malfunction] and any of the areas that have to do with
workforce, we try to reach out to as many committees that
oversee the government workforce, for example.
So I know Yll Bajraktari, we can take that as a question
for the record, sir, and bring it back to you but we are
talking with many of the committees.
[The information referred to can be found in the Appendix
on page 59.]
Mr. Langevin. Very good. Thank you very much.
I have additional questions but I am going to stop here.
And I will now yield to Ranking Member Stefanik for her
questions.
Ms. Stefanik. Thank you, Jim.
I wanted to follow up on Eric's comments regarding the
infrastructure side. So the Commission referenced many times
the importance of accessible robust data sets for the
development of machine learning and AI. However, we often hear
and we have worked through many of these impediments within
DOD, whether it be classified or controlled data sets from the
government side or concerns that exist over intellectual
property and data rights from the side of private industry
that, of course, partners with DOD.
How, specifically, would the Commission suggest AI
stakeholders alleviate these concerns and reduce those
impediments? Because I view that as a form of the
infrastructure. The data sets are the fuel for AI and this is a
really tough challenge for us to work through.
Dr. Schmidt. Maybe Commissioner Work can also add to this.
When I look at this, what I would like to see is a broad
research exemption that would allow the kind of data that is
being collected to be used for research with appropriate safety
safeguards and privacy concerns and so forth.
One of the key things to understand about AI is it needs
data. It eats data. It is how it trains. It is how it learns.
And the more data the better. There are a number of problems in
AI which seem to only get better with more data. There is no
limit to the amount of data you can feed them--language
translation or language understanding is such an example.
And so the combination of the computing resource that I
have highlighted, plus broader access to data under appropriate
safeguards is key. Each of the groups that controls this has
got to confront the fact that they have a lot of data that is
in databases that are not connected together, that nobody knows
how to get it out and so forth.
I believe, for a long time, that using intelligence as an
example, in the intelligence community, if they could simply
unify their databases, you would find an enormous number of new
things because the data is over here and not there and the AI
can see the pattern between the two that humans cannot.
Bob, could you add a little bit here?
Secretary Work. I think I can start by adding a real-life
anecdote, Congresswoman Stefanik. When we stood up Project
Maven, which was designed to sort through all of the full-
motion video and take analysts away from the screens, staring
at the screens for hours upon hours, and having the computer
work that data, Jack Shanahan, Lieutenant General Shanahan, who
was, at that time, the head of the ISR, the Intelligence,
Surveillance, and Reconnaissance Task Force, came to me and
said: We can't do what you want to do, Mr. Secretary, because
the data that we need to train the algorithms is all classified
Secret. And I said, so what do we have to do about it? And he
said, all you have to do is declassify it. So I just asked:
Okay, what would be the implications? And it turns out that it
was very easy to do and it had absolutely no impact on security
in the sense of us giving up any type of secrets.
So to your point, I believe that the JAIC [Joint Artificial
Intelligence Center] and the new chief data officer need to
have the authority to declassify data, when asked, to use for
an AI algorithm. They are the ones that are in the position to
determine whether or not declassification of the data would
pose any risk to the Department and it would make it faster and
easier to go after these AI algorithms.
Ms. Stefanik. Thank you. I yield back.
Mr. Langevin. Thank you, Elise.
Next, Mr. Larsen is recognized for 5 minutes.
Mr. Larsen. Thank you, Mr. Chair. I appreciate that.
Chair Schmidt, earlier in your testimony and we were
talking earlier, I had asked you whether or not we should make
a similar declaration on AI policy that the Chinese Government
has made, where they are going to be the global leader in AI by
2030. And I always thought, we will just say that we are going
to be the global leader by 2029. We are just going to beat them
to that. Why don't we and what would prevent us from doing
that?
Dr. Schmidt. I think your question was very prescient, in
my view, because the good news is I think we already are the
global leader and we would need to maintain it.
So, I would recommend that the Congress come up with some
mechanism to say where we must lead. There is a candidate list.
Indeed, in one of the bills you have proposed, you have made a
list, which is similar to--it is a biotechnology list, 5G, so
forth, and so on. I would view these as a matter of national
security, national priority. They are also at the basis for the
economy of America. So the most valuable companies in the
United States are all based on these technologies. We don't
want to give up that either. So that even if you don't care
about national security, you must care about our companies, and
our economic growth, and the GDP [gross domestic product], and
the wealth of our citizens.
So regardless of your point of view, I know you care about
both, you are going to want a plan. And what I would recommend
is that you ask us to produce the list and we will work with
our colleagues and the other commissions and come back to you
for your consideration.
Mr. Larsen. And if we made that ask, could you still do
that within the timeline of the Commission?
Dr. Schmidt. We have a team looking at this question. I am
sure we could do a good first start. I would also say that
these sorts of list are--they are one, they are controversial
within the community because people are fighting for their own
fiefdoms but there is also----
Mr. Larsen. We don't have that here, so fine.
Mr. Schmidt [continuing]. But in the technical world, these
battles occur.
But also, there is evidence that America's greatness is
because of our ability to integrate these things quickly. So it
is not only the areas but our ability and the flexible way in
which we work as a society to create the companies, create the
initiatives, create the resurgence, and create the health
solutions, and so forth combining them.
So I would recommend that not only do we give you something
that is interesting for your review but then you also ask, in
some other forum, for a continual review of this. I think it is
part of national security.
Mr. Larsen. Okay. On Monday, the DOD released their AI
education strategy, which was directed in section 256 of the
fiscal year 2020 NDAA. It is really focused more on educating
the women and men who wear the uniform in the DOD about the
basics of AI and how it might apply in everything that they do.
Has the Commission evaluated that and would you evaluate
it?
Dr. Schmidt. Let me ask Mignon and Jose-Marie.
Mr. Larsen. Sure, I will ask Commissioner Clyburn.
Ms. Clyburn. The short answer is yes, sir. We recognize
that, without talent, without pathways inside and outside of
the communities, that all will be troubled. I will just put it
that way.
So, we have outlined, especially in our first quarter
recommendations, a series of pathways, of opportunities, of
synergies that should be realized and adopted across multiple
platforms, including recognizing and affirming AI as a
priority, no matter what your rank, no matter what your job
description. So that holistic inclusive approach to learning
and embracing AI as a way of life, as a way of your job, as a
way of this mission as a national strategic priority, is
definitely--you will see that all through this report, sir.
Mr. Larsen. Thank you. Thank you. And I want to move to a
final question.
Chair Schmidt, you know Kai-Fu Lee, I imagine.
Dr. Schmidt. I do, very well.
Mr. Larsen. Yes. In his 2018 book on AI and superpowers, he
made a distinction between the United States being better at
innovation and with the Chinese system is better at application
of AI.
Do you agree with that assessment? And, if you agree with
that assessment, maybe, for the record, you could get back to
us how we could be better in both.
Dr. Schmidt. So Kai-Fu and I have been colleagues and
friends for a decade. And my view of his book was that it was
the case for China. I think we don't really know if his claims
are correct. But the argument that he made that is important to
state right here and right on the record is that they have a
massive investment in this area coming. And we know that there
are areas where the application of this technology is a scale
problem. In other words, we invent it and they apply it. And I
am always worried that we are going to do something that will
prevent us from having a global market.
Part of the genius of America is our companies are global
companies. So, they have a huge market. And I want to make sure
that that is a huge market and not taken over by China. So I am
worried about what he says.
We can try to give you more clarity on some of the things
that you can do. Many of them are more or broader than our
mission. So, many of them involve essentially promoting
entrepreneurship, trying to get more dynamism in the economy,
trying to get more founders to found the great companies, more
high-skills immigration, trade policy which promotes American
exports. I suspect these are things that you would agree with
but that is sort of the list.
Mr. Larsen. Yes, all right. Thank you.
I apologize, Mr. Chairman. I yield back.
Mr. Langevin. Thank you, Mr. Larsen.
Mr. Conaway is now recognized for 5 minutes.
Mr. Larsen. I will just note, Mr. Chairman, Mr. Conaway is
not present at this point.
Mr. Langevin. Okay. Is Mr. Bacon there?
Mr. Larsen. There are no Republicans present, at this
point. So it would be the next in line, the Democrat next in
line. Sorry, Mr. Chairman.
Mr. Langevin. Okay, then we will take the Republicans when
they come back.
Next will be Ms. Slotkin is recognized for 5 minutes.
Ms. Slotkin. Great. Thank you, Mr. Chairman. Great to see
you both. Thank you for coming.
You know I was reading through the recommendations of the
various quarters and I think there is nothing to contradict
anything. You guys are the experts in this. But I think for
former Deputy Secretary Work and for Mr. Schmidt, head of the
Defense Innovation Board, we always have these commissions but
the structure at the Department of Defense does not support an
easy, efficient incorporation of new technology, not for any
one person's fault but because of the incentive structure
there.
And so, while I see a lot of suggestions, I am having
trouble understanding how to incorporate this in a way that is
practical.
So from each of your perches, one in the DIB [Defense
Innovation Board] and one as DEPSECDEF [Deputy Secretary of
Defense], can you talk about the structural change that, no
kidding, no joke, would allow for this innovation to be
incorporated and not just repeating the past of the services
being able to do their own kind of decision making on this and
only a notional kind of signoff at the top level that misses
the real opportunities.
And I will turn to Mr. Schmidt first.
Dr. Schmidt. Actually, why don't we--with your permission--
--
Ms. Slotkin. Of course. Of course.
Mr. Schmidt [continuing]. Could we have Chairman Work speak
first?
Ms. Slotkin. Of course.
Dr. Schmidt. I will follow him.
Secretary Work. It is good to see you again, Representative
Slotkin.
Ms. Slotkin. It is good to see you.
Secretary Work. I could imagine you asking me this question
in my office but you are exactly right.
On a transformation of the scale that the Commission
believes is necessary, just how important artificial
intelligence will be, not only to the business applications in
the Department but, more importantly, the operations and combat
capability and effectiveness of the Department, you are going
to have to have a strong top-down push. You are going to want
to have the thousand flowers bloom within the services. That is
going to be a very good thing to see and I actually think that
is happening now. But without that strong top-down push, you
are not going to get the broadest transformation that you are
looking to.
The Commission talked about this a lot and came down with
two things. One, I think you remember the old Advanced
Capabilities and Deterrence Panel.
Ms. Slotkin. Yes.
Secretary Work. And using that as an exemplar, we said we
should have a steering group, a technological steering group,
consisting of the Deputy Secretary, the Vice Chairman of the
Joint Chiefs of Staff, the Principal Deputy Director of
National Intelligence, the Under Secretary of Defense for
Research and Engineering, and they would try to look at the
forest, instead of the trees, and approach it the way that Eric
talked about: How do you integrate all of these technologies
for military advantage? And without someone doing that on a
consistent basis, you are not going to have the transformation
that you would otherwise have.
The second thing is we feel strongly now that the Secretary
of Defense has come out and said that AI is, if not the top
priority, one of the top three priorities and try [audio
malfunction] underneath the Secretary of Defense, who could
then delegate it to the Deputy, if he or she so desires.
The CIO [Chief Information Officer], we believe, should be
responsible for the broader digital transformation of the
Department. It makes sense to have things like the cloud, and
the data strategy, and the infrastructure that Eric talked
about. The CIO is natural for that job but we need to have a
single organization that is focused on applications like a
laser beam and, in our view, the CIO shouldn't lead that
effort. It should come directly from the top, be top-down
driven, and the executive agent for that should be the JAIC.
The combination of the technical steering group, which is
looking at the forest rather than the trees, and then the JAIC,
which is helping to plant and tend to the trees, is the way we
think that you can have this be a sustained transformation.
Ms. Slotkin. Great. And Chairman Schmidt, in the last 30
seconds of my time?
Dr. Schmidt. I guess I should have realized that the
military would be top-down and it is so top-down, everything
that Bob just said is a requirement.
I would add one more thing. If I were advising the
Secretary of Defense, I would just say I want everything
faster. I want these prototypes faster. I want a design
engineering mechanism where I get stuff faster.
These product cycles for weapon systems, which are 15
years, are crazy because the technology has already moved past
what the spec was. It doesn't serve our Nation well. Let's get
to a different model, where the stuff is happening very
quickly, we are canceling and starting things, we are giving
you all choices of things to approve, and so forth.
The proposal that Bob made is very consistent with that but
I would tell you the metric I would apply, if I were Congress,
is I want things faster.
Ms. Slotkin. Okay, thank you very much. I appreciate it.
Mr. Langevin. Thank you, Ms. Slotkin.
Mr. Brown is now recognized for 5 minutes.
Mr. Brown. Thank you, Mr. Chairman. I want to thank
Chairman Schmidt and all of the commissioners for your work on
this Commission in the area of artificial intelligence. Thank
you for being here today.
I am enthusiastic about the opportunities, the potential
for artificial intelligence, particularly how it is going to
enhance not only the lethality but the survivability of our
warfighters. I think about autonomous vehicles, air, land, and
sea. I think about enhanced human decision making. I think
about improved targeting, which is so important, particularly
with the ever-increasing number of sensors in the various
physical demands on the battlefield.
But I do have some concerns. I have some concerns that you
have mentioned, Mr. Chairman, and that has been a topic of
conversation, and that is biases in both the development and
the deployment of AI, biases by culture, race, ethnicity, and
even gender. And I am concerned particularly when we talk about
AI for targeting, whether that targeting is done by the
military or by law enforcement.
NIST [National Institute of Standards and Technology], in a
study I think last year, found that African-American and Asian
faces--we are talking about facial recognition in targeting,
identifying--10 to 100 times more likely to be falsely
identified than Caucasian.
Now, the second quarter recommendations did include the
following: that R&D [research and development] is needed to
advance capabilities of AI technologies to perceive and
understand the meaning of human communication, including spoken
speech, written text, and gestures. This research should
account for varying languages and cultures, with special
attention to diversity, given that AI typically performs worse
in cases in gender and racial minorities.
So, there is a recognition of this. You probably know that
of the $4.1 billion that the DOD invests in research and
development at universities and colleges, less than 0.5 percent
goes to historically black colleges and universities [HBCUs],
and minority-serving institutions.
So I have a concern about the bias in the development and
deployment and I have a concern about the lack of diversity and
inclusion to address the bias. And nowhere in the first quarter
interim report that you issued or the Commission issued did I
hear much of anything about diversity inclusion.
So what are we doing to ensure that, as we are considering,
and studying, and developing, and deploying AI in all of its
many ways, that we are addressing biases? And in my opinion,
you do that by making sure you have a diverse and inclusive
team that is actually researching, and developing, and
delivering this technology. What are we doing now? What should
Congress do to move this along faster?
Dr. Schmidt. Speaking for the Commission, we completely
agree with the framing that you just said on the issues and
they are starkly correct.
Let me ask my fellow commissioners to comment on the
solutions. I can tell you that this is a huge issue in the
American research institutes, universities. Many, many people
have ethics groups and concerns over bias. There may be
algorithmic ways to change the algorithms, to eliminate this
bias in such a way that we don't have to worry about it as much
but right now, it is a very real issue.
Mignon and Jose-Marie?
Dr. Griffiths. I am happy to start. There are really two
answers to the question. One relates to how we build the AI
algorithms to eliminate bias and how we make sure and test
against all different scenarios that the bias doesn't exit. The
other goes back to workforce again and to have a broadly
diverse and inclusive workforce that represents the population
of the Nation. That becomes very, very important as well.
I believe it was GM [General Motors] that first brought
women in to design minivans for the soccer moms and the design
of those vehicles suddenly became very different.
So I think two ends, one is the R&D and the actual
algorithm development testing and evaluation and, on the other
side, trying to ensure that we have the broadest possible
representation coming into the workforce, which relates to
actually some of the work that we are doing for our next
report, which relates to looking not just at the universities
and their production of graduates in AI and increasing those
numbers, but actually reaching down into K-12 and also looking
at alternative pathways into the workforce that go beyond the
college degree.
So we are looking at those things at both levels.
And I think, Mignon, you may have something you wish to add
here, too.
Ms. Clyburn. Right. One of the reasons, Congressman, I
mentioned the two proposals that would encourage and bolster a
diverse--diverse sets of talent, including civilian talent, is
when it comes to the development, when it comes to those teams,
they have to be inclusive. They have to be diverse. When models
are designed, there has to be an inclusive and expanded table.
That is the problem.
One of the things--and I know I am running up on your
minute--firsthand, a few years ago, I went in with this AI-
enabled product and it didn't see my face at all. I was
invisible in a room. So if I am invisible, that was a kind of
passive invisibility but invisibility in terms of presence, in
terms of you know being able to not only see but predict in a
productive way, if that is not at the design phase, then we are
going to have a perpetual problem at the implementation phase.
So, it has to be inclusive. It has to be diverse and we have to
be conscious and intentional about production and application.
Mr. Brown. Mr. Chairman, in yielding back, if I could just
comment that last year's NDAA we directed the Secretary of
Defense to commission a study, the national study of defense
research at HBCUs, in an effort to increase the research
dollars, and the capacity, and capability. I would commend and
request that this Commission take a look at that work--they are
at the very early stages--and perhaps provide them guidance and
input on how we can get more research dollars in those
universities and colleges, where you have a high concentration
of diverse candidates doing extraordinary things at the
graduate and undergraduate level, and AI may very well be a
part of that.
Thank you, Mr. Chairman. I yield back. Thank you for your
indulgence.
Dr. Schmidt. And that is a yes.
Mr. Langevin. Thank you, Mr. Brown. Thank you, Mr. Brown.
It is certainly a very important topic to raise.
With that, Ms. Trahan is recognized for 5 minutes.
Mrs. Trahan. Thank you, Mr. Chairman, and I really
appreciate the expertise represented in this Commission, and
the depth of your recommendations laid out in the Q1 and Q2
reports.
I think my question is for Commissioners Griffiths and
Clyburn. I am interested in just digging deeper into your line
of effort on AI talent. There seems to be a severe lack of AI
knowledge in DOD and other parts of government, where that AI
requirements and capabilities get automatically bundled with
cyber missions.
And so, one, it would be great if you could just explain
why it is important to decouple our AI workforce from the cyber
workforce. And then, also here, how you would recommend the
government create a system for measuring and tracking its AI
knowledge.
Dr. Griffiths. I will jump in, if I may.
You are absolutely right. I think that decoupling cyber
from AI is very important because cyber has a mission and it is
very, very clear what the mission of cyber is. And while they
may have common roots in fundamentals of computer science, they
branch off after that and focus on different missions and
confusing them confuses everyone.
From an academic perspective, academic programs are very
different for producing graduates in those areas. So, I think
decoupling is one and we have made recommendations not for
eliminating one or crowding out a program in cyber, but
actually adding to the vehicle or adding to the mechanism to
ensure that AI receives appropriate attention, as well as
cyber.
On the issue of the talent base within the government, you
have no ways, at the moment, of knowing who has the talent and
who doesn't have the talent. You don't even really know in the
military who has capabilities of coding. And so a number of our
recommendations are addressed to sort of testing for
computational thinking, which now is the prevailing thought,
the kind of capability and underlying fundamental skills, and
the mindset that people have, the talent that they have that
can then be developed into AI-related capabilities.
So, and the other area is, of course, a lot of educational
training at different levels. And the workforce model that we
developed jointly with the DIB and the JAIC I think really
addressed all those different layers. We have three very
technical work roles and four non-technical work roles, and all
of them need to have some and different levels of understanding
of AI, including ethical issues associated with the acquisition
and application of AI.
Ms. Clyburn. And I will just finish up by saying AI will
continue to transform. It is not passive and nor should we--
should it be passive, in terms of our intent, in terms of it
being a single focal line of effort, if I may borrow our
jargon, so to speak. And this general purpose technology has
multiple applications that are significant with our focus when
it comes to national security but it has a ripple effect
throughout society.
So not targeting, not streamlining, not separating, so to
speak, would be to our peril because of the expansive, the
significance, and how much it is interwoven into our everyday
lives.
Mrs. Trahan. I appreciate that. And, you know, if I have
time for just one--I think I do have time to slip this last
question in.
But, you know, we hear all the time that, you know, we are
not recruiting enough technical talent and, certainly, we are
not moving at the speed that we need to, and our onboarding
authorities are often not able to meet the demand. You
addressed some of these concerns in your quarter, I think it
was the Q1, recommendations around strengthening the AI
workforce and using a Cyber Excepted Service.
I am wondering if, you know, whenever we have folks in the
private sector in front of us, we always love to, you know,
borrow your best practices. In terms of specific
recommendations the government should consider to be more
competitive in recruiting this technical talent, knowing full
well that some of the most competitive companies with the
largest market share across the globe don't take the--you know,
they don't necessarily pay the most, are there anecdotes that
you can share with us on how we can think of novel ways of
recruiting technical talent more quickly?
Dr. Schmidt. This is Eric. May I answer your question?
Mrs. Trahan. Please.
Dr. Schmidt. In my service to the DOD, for the 4 years I
was the chairman of the DIB, I was struck by how many people
want to volunteer to serve their Nation, and they are willing
to do so at very low salaries, and under lots of difficult
situations. The things that drive them crazy are things like it
takes 3 months to get an offer out, or that they get classified
in the wrong way, or they can't--they have 5 minutes per day to
clean up their email because they only have a megabyte of email
and their email system doesn't work.
So, they are willing to serve at a lower income level for
the Nation but it has got to work operationally and there are
many such simple things that the DOD and others could do, it is
true of the Federal Government in general.
Mrs. Trahan. Thank you so much. Mr. Chairman, thanks for
indulging me. I, obviously, yield back.
Mr. Langevin. Thank you, Ms. Trahan.
The chair now recognizes Ms. Houlahan for 5 minutes.
Ms. Houlahan. Thank you, Mr. Chair, and thank you to all of
the witnesses. And I just have some questions for any of you.
The Commission specified a belief that the JAIC should be
elevated from its current position with the CIO to a direct
report to the Secretary and it notes that the Secretary can
delegate responsibility to the Deputy Secretary. The
recommendation, however, does speak to your position that AI is
an issue that has to be raised in order to, and this is a quote
from the report, provide the requisite level of senior
oversight and support needed to preserve the Department's
initial AI projects, enable their growth, and ensure that the
Department can develop the capabilities needed to successfully
adopt AI applications.
So the Secretary and the Deputy Secretary have the ability,
right now, to take such a step today. However, they haven't
acted on that recommendation to date, necessitating legislative
involvement on the House's side in our most recent NDAA.
So my question is: Can you all help us understand,
potentially, the reason that the Department has not accepted
your recommendation and can you potentially reiterate for us
the reasons why our colleagues on the Senate side should be
interested in accepting the Commission's position that the JAIC
should be placed under the Secretary's direct authority?
Dr. Schmidt. Bob, can you comment on that?
Secretary Work. Yes, ma'am. The JAIC, I think right now, is
starting to hit its stride. Dana Deasy, who is the current CIO,
is an extraordinarily capable public servant and the JAIC has
grown up under his supervision. And I think the Department of
Defense is quite happy with the way things are progressing, as
they should be.
What the Commission was thinking is that, over time, the
CIO--you should split the responsibility, where the CIO focuses
on the digital transformation of the Department, like getting
the cloud set up and settled, doing the data strategy for the
Department, and focused on all of the infrastructure, and
having the JAIC focused, really like a laser beam, on AI
applications.
At some point, someone is going to have to be designated
kind of the system architect for large DOD programs and the
system architect would say these are the AI applications that
we think would have the broadest and the most consequential
impact on the joint enterprise.
And so we think that having the JAIC doing that and having
everybody in the services understanding that they are working
under the direct supervision of either the Secretary or the
Deputy Secretary, that that is the fastest way to get
transformation. That I think is----
Ms. Houlahan. So, sir, what you said was you recommended
that over time and then you ended by saying that is the fastest
way. And so what is the timeline that your recommendation is or
is it best to be immediately you know rip the Band-Aid and move
forward? What is the recommendation, in terms of timeline and
immediacy?
Secretary Work. I think I can speak for the Commission,
where we said the sooner you do this, the better--split the
responsibilities, really start to run. As Eric said, we believe
that urgency and scale is absolutely important. So the sooner
you would do something like this, we think, the better.
Ms. Houlahan. Okay. So, urgency, the sooner the better. Do
you all have anything else that you would like to contribute to
that response?
Dr. Schmidt. So, we are celebrating the success of the JAIC
today. Two or three years ago, there were people who didn't
want to do it at all and it took, again, top-down leadership
from the Secretary of Defense at the time to really force it.
If you want to take a large bureaucracy and reform it in a
way that is consistent with national security and modern
principles, you have to force it from the top. This is a
maneuver of that nature. There are probably others.
Ms. Houlahan. No, I appreciate that. I had the opportunity
to visit the JAIC and I do understand how innovative it is and
how impossible the lift is from an organization as big as the
DOD to create something new and innovative.
I only have 24 seconds with my time, so I think I will
yield back, Mr. Chair.
Mr. Langevin. Thank you, Ms. Houlahan. And we are going to
go to second round anyway, so if you want to ask another
question, you will have another opportunity.
So I understand Elise has not returned yet but hopefully
she will be back soon.
So I just had two additional questions. I just want to
thank our witnesses again, as we go to a second round.
My first question is: What are the most significant gaps in
current legislation that relate to artificial intelligence and
national security and what do you believe are the Commission's
most consequential recommendations that Congress has not yet
acted upon?
Dr. Schmidt. Maybe each of the Commissioners could answer
your question, Mr. Chairman.
My reaction is that the profound change that is needed is
the workforce one, that without addressing the workforce one,
the gains will be lost in the bureaucracy, et cetera.
Other commissioners?
Ms. Clyburn. Investment, sir. This will not be a free
endeavor but it is one that is a priority, it is one that it is
critical. And so, honestly, right now I would say budgetary.
Dr. Griffiths. If I may, Mr. Chairman, I believe that--I
endorse what Eric said. I believe workforce is the core. As we
have discussed at our various Commission meetings, workforce
has an impact on every other line of effort and it needs to be
done and incremental changes are not going to make a
difference.
So, we do need to make sort of moves that generate and lift
the scale up, so that we can get the workforce ready quickly to
enable some of the innovations to move forward quickly.
Secretary Work. I agree with my colleagues that workforce
is number one. The National AI Research Resource I think would
be a close second, along with the increases in investments, as
Mignon has talked about.
The recommendations we have made on microelectronics and
how we maintain our lead in that area I think are
extraordinarily consequential. And then the recommendations we
make as far as pursuing with our allies this together, since we
see very clearly that this is a competition in values, as much
as it is a competition in technology, and we want to involve
all like-minded democratic nations so that AI reflects the
values of democratic nations with respect for personal privacy,
rule of law, et cetera.
So, I would put those four as very close together but with
workforce on top.
Mr. Langevin. Thank you, Secretary Work, and that leads--it
is a good segue to my last question.
Obviously, as you pointed out, we are not developing AI in
a vacuum. Other nations, who are competitors, or adversaries,
enemies, are also developing AI technology and they may not be
approaching it with the same type of safeguards that we are
putting in place to make sure that AI is used ethically and
responsibly.
What do we do about that? How do we protect ourselves
against it? What is the way forward to pressure other countries
to approach use of AI ethically and responsibly so we do have
some safeguards in place?
Dr. Schmidt. Well, AI can be misused by countries that have
different values than ours, the most obvious ones being the
misuse of surveillance. And that, to me, is a political and
nation-state issue. The technology is already in China. They
are going to do whatever they are going to do with it. I don't
know how to stop that technically.
I think it is very important that if a technology is
invented in America that is dual-use and which could be used
for very bad things, there would be a moment of reflection as
to whether that should be broadly released or kept more close
for that reason.
Mr. Langevin. Thank you. Any other commissioner have a
comment?
Ms. Clyburn. I think the values need to be baked in. I
believe, as was mentioned, those--the allies, the
relationships, the norms need to be socialized, agreed, and
expanded robustly. I believe to strengthen all of this, you
know, would lie in our principles and those ethical standards,
as we talk about efficiencies and all of the other benefits.
But the strength, to me, are the principles that the U.S.
and its allies would, I believe, promote and amplify.
Secretary Work. Sir, this is----
Mr. Langevin. Have you done anything with respect to
putting safeguards into play with the use of AI?
Ms. Clyburn. There was a little bit--I heard. I don't think
I----
Mr. Langevin. I know Secretary Work was going to respond
there.
Secretary Work. Well, sir, on Monday, I just wanted to
check on something because there was an extraordinary meeting,
where 100 officials from 13 democratic countries met online
Tuesday and Wednesday to discuss how their militaries could
ethically use AI. It was the first summit of its kind. It
follows on the heels of the AI ethical principles that Eric
spoke to, which were the first published by any organization of
the DOD, and it was hosted by the JAIC. And it has kicked off
what JAIC is calling the AI Partnership for Defense. This is
something that we haven't had an opportunity to talk about, as
a commission, but I can state with certainty that we would
endorse it and embrace it.
And central to this is talking about the limits, the moral,
legal, and ethical limits that we want to establish for AI and
national security applications. So, I really take this meeting
as a very positive first step.
Mr. Langevin. Agreed.
Do any of you know of any international efforts, including
at the U.N. [United Nations], where this discussion and a
framework for ethical use of AI is being undertaken with
seriousness?
Secretary Work. The primary place, right now, is looking at
the ethical--the use of autonomous weapons and that is in the
U.N., the United Nations' GGE, Group of Government Experts, as
part of the CCW, the Convention on Certain Conventional
Weapons. And that is very well attended by the Department of
Defense and the Department of State and that is the primary
place where the debate over how far and in what form AI-enabled
autonomous weapons can be or should be used on the battlefield.
Mr. Langevin. Very good. Thank you.
With that, let me go to Mr. Larsen for any questions he may
have.
Mr. Larsen. Thank you. I was wondering if any of the
panelists have a view on your recommendation or your guiding
principle 5, where you discuss principles of free inquiry, free
enterprise, and free flow of ideas.
The last line of that section says: At the same time, we
must not lose sight of enduring American principles and overly
securitize basic research or the private sector.
I was wondering if you have some guidance for us on how to
not overly securitize or to appropriately securitize basic
research?
Dr. Schmidt. The Commission has spent a fair amount of time
on this question of protection and it is an obvious one. An
American firm, an American researcher invents something very
important. What should we withhold, and what should we try to
withhold, or maybe we will be not successful? And the consensus
of the Commission seems to be that it is very difficult to
withhold algorithms or even software because the algorithms are
broadly known and they will be discovered by the competitor
anyway, and the software is relatively reproducible. However,
the insight from the Commission is that there are significant
parts of the hardware value chain which are very specialized
and that deserves very special attention.
Maybe the other commissioners may want to add something to
add color to my statement.
Mr. Larsen. Do any other commissioners have any views on
that to add? If none--okay, thank you very much.
Dr. Schmidt. Thank you.
Mr. Larsen. I yield back, Mr. Chairman.
Mr. Langevin. Thank you, Mr. Larsen.
Ms. Houlahan, do you have additional questions?
Ms. Houlahan. I do. I do. I appreciate that.
My first question has to do with my interest in developing
the U.S. Digital Service Academy which, of course, would create
a pipeline of security-cleared Federal employees, and that was
part of your proposal.
Can you highlight the components of that program and have
you done any outreach to determine if there is an appetite for
students for such a program?
Dr. Schmidt. There appears to be infinite appetite for this
idea, if we can find the money.
Mignon, are you the best expert on this to comment on it?
Ms. Clyburn. I will give it a try. Thank you very much.
One of the things, highlights that makes this particularly
appealing, we believe, it is modeled after, you know, existing
academies. We also know that existing academies, they are
overflooded with applications. So, we think the interest would
be there. We also think that the interest would be there for
those who may not be able or might not have the interest in
moving into military service or reservist service afterwards.
So the thing that is the most attractive, however, it is an
incredibly targeted pipeline for all of the issues, when we
talk about the lack of talent and having broad pathways for
educational opportunities that are STEM [science, technology,
engineering, and mathematics] or AI oriented. This would happen
more robustly under this framework.
Dr. Schmidt. Can I add, Congresswoman?
Ms. Houlahan. Yes, please.
Dr. Schmidt. In our specific proposal, this is an
independent entity under the Federal Government but it has a
public sector-private sector board. In our formulation, it
would be able to raise private money in addition to Federal
money. In theory, it would offer--it would be able to charge
tuition and have requirements of payback and things like that.
So, very similar to the way our military academies operate.
And this is, perhaps, our strongest recommendation and I
don't think it is that difficult for the government to do.
There is huge demand. There is plenty of universities that
would love to advise us and help us. We have had outpourings
from many of the States that we have spoken with. They said,
look, how can we help; what can we do?
Ms. Houlahan. Yes, it sounds like a really intriguing
concept and I think something that we probably should take a
very hard look at.
I also understand that you have been talking with industry
experts in academics that have indicated that they would like
to contribute to government missions because of a sense of
civic responsibility or interest in unique government missions
but they don't want to leave their current career field, even
temporarily.
So, as a result, you have made the recommendation to create
a National Reserve Digital Corps. Could you all also please
elaborate or highlight on the purpose and the fundamentals of
the Reserve Digital Corps, please?
Dr. Schmidt. Mignon or myself?
Ms. Clyburn. Either. What this would do, it was patterned
itself after ROTC. It will allow for a number of less than 40
days of service. It would enable the infusion of talent to
troubleshoot, to triage, and overall to benefit the current
ecosystem when it comes to our government, with giving those
the flexibility to meet whatever other needs, in terms of
economic needs, that they wish to.
The fact of the matter is, as it stands right now, one of
the biggest barriers for serving full-time in government is
money, is resources. Academia and the private sector attract,
oftentimes, some of the better talent and this is one of those
ways that we would recognize that, leverage that, without
somebody having to make a hard permanent choice.
Ms. Houlahan. So I am just making sure that I understand. I
was ROTC in the Air Force and went to school with a full
scholarship at Stanford and repaid that by serving. Are you
suggesting that this is the same kind of a thing with existing
colleges and a pathway to repay; whereas, the other concept is
the same as the Air Force Academy in this analogy?
Ms. Clyburn. If I am speaking with the right program, in
terms of the program, that you are saying the reservists,
correct?
Ms. Houlahan. Yes, the Reserve Digital Corps. Yes, ma'am.
Ms. Clyburn. Right, it would be patterned after that, all
of the benefits of that would be----
Ms. Houlahan. I am just trying to drill down and understand
just for myself what you mean by they don't have the time to
commit because it seems as though I had a lot of commitment on
the other side of the opportunity.
Ms. Clyburn. I guess what I am saying, in this particular
case, you know sorry about the miss, is that someone who does
not want to take a permanent military track.
Dr. Schmidt. This is a somewhat lighter version of what you
are describing. The ROTC, by the way, is a fantastic program
that America has in our universities. Many of the leaders I
have worked with have gone through ROTC and it is a great way
to build bridges between the civilian and the military sector.
And those are lifetime--lifetime commitments. And you are,
obviously, a success from that.
One way to think about this is we have got to get a way for
people to be able to work in the military--work for the
military but not be in the military.
Ms. Houlahan. Okay.
Dr. Schmidt. This is a mechanism to do this. It is a
structural mechanism, where they can say to their employer I am
required to do this; you have to let me do this.
Ms. Houlahan. So it is a reservist, rather than an ROTC. I
have got it.
Dr. Schmidt. A reservist. It is a reservist model.
Ms. Houlahan. Okay, I apologize. I appreciate the
clarification and I yield back.
Ms. Clyburn. No, and I apologize. I apologize, too, for the
confusion.
Mr. Langevin. Very good. Well, we will wrap up this hearing
now. I want to thank our witnesses for their extraordinary
testimony, for the extraordinary commitment that you have made
to the National Security Commission on AI. We look forward to
continuing to follow your work and the final report. And again,
your testimony today has just been extraordinary and valuable.
So, with that, I want to thank you again. And this hearing
now stands adjourned.
[Whereupon, at 2:39 p.m., the subcommittee was adjourned.]
=======================================================================
A P P E N D I X
September 17, 2020
=======================================================================
PREPARED STATEMENTS SUBMITTED FOR THE RECORD
September 17, 2020
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
=======================================================================
WITNESS RESPONSES TO QUESTIONS ASKED DURING
THE HEARING
September 17, 2020
=======================================================================
RESPONSE TO QUESTION SUBMITTED BY MR. LANGEVIN
Mr. Work. The National Defense Authorization Act for Fiscal Year
2019 names the House and Senate Committees on Armed Services, Select
Intelligence, and Commerce as the key committees of oversight for
NSCAI. However, to get AI for national security right, many of the
Commission's recommendations have included the jurisdiction of other
committees. Advancing the AI for national security and defense needs
requires taking action beyond traditional national security departments
and agencies. Recommendations such as expanding research and
development, growing an AI-ready workforce, and establishing ethical
standards are examples of actions that often begin within other
committees. To that end, NSCAI Commissioners and staff have engaged
with individual members of Congress and a number of committees beyond
our committees of oversight including, but not limited to, the House
and Senate Committees on Appropriations; House Committee on Ways &
Means; House Committee on Oversight and Reform; Senate Committee on
Homeland Security and Governmental Affairs; House and Senate Committees
on the Judiciary; House Committee on Science, Space, and Technology;
House Committee on Homeland Security; House Committee on Foreign
Affairs; Senate Committee on Foreign Relations; and Senate Committee on
Energy and Natural Resources.
As we approach NSCAI's final report in March 2021, we welcome
engaging further with you and your colleagues, including colleagues on
other committees, to inform our findings and recommendations and
educate Congress on the vital importance of U.S. leadership on this
transformational emerging technology. [See page 15.]
=======================================================================
QUESTIONS SUBMITTED BY MEMBERS POST HEARING
September 17, 2020
=======================================================================
QUESTIONS SUBMITTED BY MR. WALTZ
Mr. Waltz. Looking forward, there will continue to be a need for
the government to pave the way on very specific innovations that
satisfy specific government needs. But where there is invention in the
private sector that can be harnessed, why not focus DOD efforts on
enabling and co-opting what is already being done for our national
defense? Industry is already heavily investing in R&D for AI
technologies, so how can we make sure we are partnering with them?
The Commissioners. As the Commission noted in our Interim Report,
private sector leaders and government officials must build a shared
sense of responsibility for the welfare and security of the American
people. The government needs help from industry and academia to
maximize the promise of AI and minimize the national security risks
posed by AI. American companies are at the forefront of AI
developments. Their investments dwarf federal R&D; they generate many
of the major breakthroughs; and, they are on the frontlines of
defending against cyber threats and malicious uses of AI applications.
To harness the full potential of private sector innovation, the
government must adapt internally, modernizing its platforms, policies,
and procedures to improve access to commercial breakthroughs and create
the environmental conditions for success. At the same time, the
government must also grow the innovation base, investing in early-stage
AI research across the commercial sector and academia. In our first
quarter and second quarter memos, the Commission recommended a number
of Executive and Legislative Branch actions that the U.S. Government
could take to improve its ability to leverage commercial advances in
AI. While no single recommendation will ensure successful partnership
between the private sector, academia, and the government, the
Commission believes that these actions could set in motion a closer
collaboration between the two sectors, as they would forge a common
commitment to protecting our values, free market principles, national
defense, and security.
Mr. Waltz. Keeping up with the general theme of maintaining AI
leadership globally, how can the U.S. compete with countries like
Russia and China who have much laxer laws on competition, privacy and
the like?
The Commissioners. History has shown that the United States has a
unique ability to galvanize international coalitions around core shared
values. The Commission believes that the U.S. approach to AI enshrines
these shared values and represents an asymmetrical advantage that the
United States has over authoritarian regimes. U.S. leadership will
endure to the extent the United States can continue to be at the
forefront of innovation and technical expertise. Global leadership in
AI technology is a national security priority. Given the centrality of
AI to the future of our economy, society, and security, the U.S.
Government must pursue an investment strategy that extends America's
technological edge. Global leadership gives our defense and security
agencies access to the best technology, and puts the United States in
the best position to secure that technology against vulnerabilities and
develop international norms and standards for responsible use. While
American companies play a significant role in advancing AI research and
development, the government retains a core responsibility to steer
advancements in ways that protect the American people and foster a
robust basic research environment. People are still essential. Talent
remains the most important driver of progress in all facets of AI. We
must prioritize cultivating homegrown talent by making long-term
investments in STEM education. In the near term, high-skilled
immigration is important for rapidly growing America's talent pool. One
of America's advantages is the fact that its universities, companies,
and innovation culture are magnets for the world's best AI talent. We
need to encourage that talent to come, contribute, and stay. Within
government, recruiting, training, and retaining AI-talent will be
essential to maximize AI's potential. Diplomatically, to maintain
global leadership in AI requires the United States to attract partners
and allies around these values and take steps to make those
partnerships and alliances more resilient--focusing on the economic,
security, and technological benefits of alignment with free and open
states rather than repressive regimes around world-changing technology
like AI. In recent years, we have seen this in action. The rise of
several important multilateral coalitions, including the D10 initiative
and the Global Partnership on AI, bring together like-minded nations to
develop AI norms, principles, and applications consistent with their
values. Likewise, U.S. efforts to galvanize worldwide support against
Chinese 5G technology has proven significant as the number of nations
opposed to Chinese technology continues to grow. The United States must
continue to lead the world. The Commission is committed to driving
changes that will maintain this lead by maximizing the role of AI in
protecting U.S. security, extending American leadership in emerging
technologies, and strengthening our core values, and the Commission
will continue to push for these important changes as we work toward our
final report due to Congress in March 2021.
Mr. Waltz. I am concerned about the challenges you identified for
the government to find and maintain an adequate AI literate workforce.
And I am interested to explore how we can harness the power of our
universities to create and strengthen the AI workforce we need. I
noticed a new and remarkable partnership between the University of
Florida, a top educational institution in my home state that is
currently ranked #6 among public universities and #1 in technology
transfer and innovation, and NVIDIA, the major computer hardware
company in the world. Come December, UF will be the first institution
of higher learning in the U.S. to receive NVIDIA's DGX A100 systems,
which are designed to accelerate diverse workloads, including AI
training, inference, and data analytics. Together with their
HiPerGator3 supercomputer, UF will be home to one of the fastest, most
powerful computers in higher education. As a comprehensive educational
institution, UF will be among the nation's first to integrate AI across
all disciplines and make it a ubiquitous part of its curriculum. It
will offer certificates and degree programs in AI and data science,
with curriculum modules for specific technical and industry-focused
domains.
Can you tell us about the concept of creating a U.S. digital
academy or AI university that would be on par with the other service
academies?
In addition to the academy model, are you considering mirroring the
ROTC model of progressive scholarship with a service requirement upon
completion to create AI ready graduates to step into government
positions?
The Commissioners. The concept of the United States Digital Service
Academy (USDSA) began with our finding that there is a severe shortage
of AI knowledge across our United States Government. While addressing
issues with the hiring process and expanding scholarship for service
programs are both necessary and helpful, they are also insufficient to
address the government's overall technologist deficit. To solve this
problem, we argue the government must re-imagine the way it recruits
and builds its digital workforce.
USDSA would provide two major benefits. First, it would produce
large numbers of technically educated graduates that would serve in
government. USDSA would be an accredited, degree-granting university,
with a mission: ``to develop, educate, train, and inspire digital
technology leaders and innovators and imbue them with the highest
ideals of duty, honor, and service to the United States of America in
order to prepare them to lead in service to our nation.''
Students would receive a technical education during the school
year, and participate in government and private-sector internships
during the summer. They would graduate with a degree and the technical
competencies that are needed within the government, and owe no student
debts. Instead, and to your question, USDSA graduates would become
civil servants with a five-year obligation to serve in the U.S.
Government. We believe that USDSA should be an independent entity
within the Federal Government, advised by an interagency board and
assisted by a federal advisory committee composed of commercial and
academic leaders in emerging technologies. This interagency board would
allow agencies to bring forward their digital workforce requirements
through a formal process annually and help align graduates of USDSA to
the government's workforce needs by agency and by career field.
With regard to mirroring ROTC, that model is closely aligned to
existing scholarship for service programs, which offer a three-year
program in school for three years of government service upon
graduation. Our Second Quarter memorandum includes a recommendation to
significantly expand CyberCorps: Scholarship for Service and SMART:
Scholarship-for-Service, with an increased focus on AI-related topics.
Our Second Quarter Recommendations memo provided a three-pronged
approach to developing a ``Digital Corps'' concept. The first prong
being a civilian reserve force with the National Reserve Digital Corps
(NRDC) concept for part-time civilian government work. The second and
third prongs were focused on full-time civilian government work in the
expansion of scholarship for service programs and the creation of an
USDSA. These three recommendations together provide multiple paths to
address our current and future government civilian part-time and full-
time needs with diverse pipelines to bring talent into the government
civilian workforce.
[all]