[House Hearing, 116 Congress]
[From the U.S. Government Publishing Office]
MACHINES, ARTIFICIAL INTELLIGENCE,
AND THE WORKFORCE: RECOVERING AND
READYING OUR ECONOMY FOR THE FUTURE
=======================================================================
HEARING
BEFORE THE
COMMITTEE ON THE BUDGET
HOUSE OF REPRESENTATIVES
ONE HUNDRED SIXTEENTH CONGRESS
SECOND SESSION
__________
HEARING HELD IN WASHINGTON, D.C., SEPTEMBER 10, 2020
__________
Serial No. 116-30
__________
Printed for the use of the Committee on the Budget
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
Available on the Internet:
www.govinfo.gov
___________
U.S. GOVERNMENT PUBLISHING OFFICE
42-322 WASHINGTON : 2021
COMMITTEE ON THE BUDGET
JOHN A. YARMUTH, Kentucky, Chairman
SETH MOULTON, Massachusetts, STEVE WOMACK, Arkansas,
Vice Chairman Ranking Member
HAKEEM S. JEFFRIES, New York ROB WOODALL, Georgia
BRIAN HIGGINS, New York BILL JOHNSON, Ohio,
BRENDAN F. BOYLE, Pennsylvania Vice Ranking Member
ROSA L. DELAURO, Connecticut JASON SMITH, Missouri
LLOYD DOGGETT, Texas BILL FLORES, Texas
DAVID E. PRICE, North Carolina GEORGE HOLDING, North Carolina
JANICE D. SCHAKOWSKY, Illinois CHRIS STEWART, Utah
DANIEL T. KILDEE, Michigan RALPH NORMAN, South Carolina
JIMMY PANETTA, California KEVIN HERN, Oklahoma
JOSEPH D. MORELLE, New York CHIP ROY, Texas
STEVEN HORSFORD, Nevada DANIEL MEUSER, Pennsylvania
ROBERT C. ``BOBBY'' SCOTT, Virginia DAN CRENSHAW, Texas
SHEILA JACKSON LEE, Texas TIM BURCHETT, Tennessee
BARBARA LEE, California CHRIS JACOBS, New York
PRAMILA JAYAPAL, Washington
ILHAN OMAR, Minnesota
ALBIO SIRES, New Jersey
SCOTT H. PETERS, California
JIM COOPER, Tennessee
RO KHANNA, California
Professional Staff
Ellen Balis, Staff Director
Becky Relic, Minority Staff Director
CONTENTS
Page
Hearing held in Washington, D.C., September 10, 2020............. 1
Hon. John A. Yarmuth, Chairman, Committee on the Budget...... 1
Prepared statement of.................................... 5
Hon. Steve Womack, Ranking Member, Committee on the Budget... 7
Prepared statement of.................................... 9
Susan Athey, Ph.D., Economics of Technology Professor,
Stanford Graduate School of Business, and Associate
Director, Stanford Institute for Human-Centered Artificial
Intelligence (Hai)......................................... 13
Prepared statement of.................................... 16
Daron Acemoglu, Ph.D., Institute Professor of Economics,
Massachusetts Institute of Technology...................... 38
Prepared statement of.................................... 40
Darrell West, Ph.D., Vice President and Director of
Governance Studies, Brookings Institution.................. 49
Prepared statement of.................................... 51
Jason Matheny, Ph.D., Founding Director, Center for Security
and Emerging Technology at Georgetown University, and
Commissioner, National Security Commission on Artificial
Intelligence............................................... 58
Prepared statement of.................................... 60
Hon. Sheila Jackson Lee, Member, Committee on the Budget,
statement submitted for the record......................... 95
MACHINES, ARTIFICIAL INTELLIGENCE,
AND THE WORKFORCE: RECOVERING AND
READYING OUR ECONOMY FOR THE FUTURE
----------
THURSDAY, SEPTEMBER 10, 2020
House of Representatives,
Committee on the Budget,
Washington, D.C.
The Committee met, pursuant to notice, at 1:04 p.m., via
Webex, Hon. John A. Yarmuth [Chairman of the Committee]
presiding.
Present: Representatives Yarmuth, Boyle, Schakowsky,
Kildee, Panetta, Morelle, Scott, Jackson Lee, Sires, Khanna;
Womack, Woodall, Johnson, Flores, Hern, Burchett, and Jacobs.
Chairman Yarmuth. This hearing will come to order. Good
afternoon and welcome to the Budget Committee's hearing on
Machines, Artificial Intelligence, and the Workforce:
Recovering and Readying Our Economy for the Future.
Before we begin, I want to welcome the newest Member of the
Budget Committee, Chris Jacobs. Welcome, Chris. Before coming
to Congress, Chris was a New York State Senator, and the
Committee is happy to have you here.
Mr. Jacobs. Thank you.
Chairman Yarmuth. Now before I welcome our witnesses, I
will go over a few housekeeping matters.
At the outset, I ask unanimous consent that the Chair be
authorized to declare a recess at any time to address technical
difficulties that may arise with such remote proceedings.
Without objection, so ordered.
As a reminder, we are holding this hearing virtually in
compliance with the regulations for committee proceedings
pursuant to House Resolution 965. First consistent with
regulations, the Chair, or staff designated by the Chair, may
mute participants' microphones when they are not under
recognition for the purpose of eliminating inadvertent
background noise.
Members are responsible for unmuting themselves when they
seek recognition or when they are recognized for their five
minutes. We are not permitted to unmute Members unless they
explicitly request assistance. If I notice that you have not
unmuted yourself, I will ask you if you would like the staff to
unmute you. If you indicate approval by nodding, staff will
unmute your microphone. They will not unmute you under any
other circumstances.
Second, Members must have their cameras on throughout this
proceeding and must be visible on screen in order to be
recognized. As a reminder, Members may not participate in more
than one committee proceeding simultaneously. For those Members
not wanting to wear a mask, the House rules provide a way to
participate remotely from your office without being physically
present in the hearing room.
Now, I will introduce our witnesses.
This afternoon we will be hearing from Dr. Susan Athey,
Economics of Technology Professor at Stanford Graduate School
of Business, and Associate Director at the Stanford Institute
for Human Centered Artificial Intelligence.
Dr. Daron Acemoglu, Institute Professor of Economics at the
Massachusetts Institute of Technology.
Dr. Darrell West, Vice President and Director of Governance
Studies at the Brookings Institution.
And Dr. Jason Matheny, Director for the Center For Security
and Emerging Technology at Georgetown University and
Commissioner for the National Security Commission on Artificial
Intelligence, who I might add has just informed me he is from
Louisville, Kentucky, so we are especially glad to have him
here with us.
Thank you all for being with us today.
I will now yield myself five minutes for an opening
statement.
This year Labor Day felt different than previous years.
While most of us still honored our workers and celebrated their
vital contributions to our nation, especially our frontline
workers, we also recognize the hardships faced by millions of
laid off Americans and their families struggling to get by amid
global pandemic and the worst economic downturn since the Great
Depression.
Yet these twin crises have amplified problems that existed
long before the coronavirus: devastating healthcare inequities,
the loss of stable well-paying jobs, and stagnating wages.
While our economy has slowed, exacerbating these underlying
issues, technological change has marched on creating even more
challenges.
As we look to the future artificial intelligence, or AI,
has significant potential to disrupt the world. It presents
opportunities to improve lives, livelihoods, productivity, and
equality. However, it also poses serious risks of large scale
economic changes.
Today's hearing will help us ground our thinking in facts
and better prepare for this impending economic transition.
Like the arrival of the steam engine, electricity, and
computers, AI will reshape a broad swath of industries and
jobs. However, history shows us that while technological
advancements can create new jobs that increase productivity and
growth, these benefits have been paired with the elimination of
old jobs and increased inequality as some workers are left
behind.
Today we are losing jobs because the administration's
failed response to the pandemic and economic crisis, but as the
economy eventually recovers, workers may find it difficult to
get their job back as companies replace jobs with new AI
enabled automation.
So while advancements in AI technologies could create more
opportunities for workers with advanced education or
specialized skills, workers without these skills could see
fewer opportunities in the near future, and it is low and
middle waged jobs that are most at risk.
Since the mid-1980's, but prior to the pandemic, 88 percent
of middle skilled job losses associated with the automation of
routine tasks took place within 12 months of a recession.
Absent concerted efforts to foster inclusive recovery, AI and
automation could exacerbate income inequality, widen racial and
gender income gaps, and push more people into poverty when we
eventually emerge from this recession.
There is already a large and persistent racial wealth gap
in America. And since Black and Latino Americans are over
represented in occupations at high risk for automation, they
are disproportionately at risk of job and wage losses.
Additionally, there are 40 percent more women than men who work
in occupations at high risk for automation.
The Organization for Economic Cooperation and Development
estimates that AI and automation could eliminate upwards of 14
percent of today's jobs and disrupt an additional 32 percent.
Current AI technologies have also raised concerns around
replicating human biases and discrimination in algorithms.
Given the range of AI applications emerging in employment,
housing, healthcare, financial services, and criminal justice,
improved transparency and oversight are needed to ensure AI
tools do not replicate or expand discriminatory practices.
Just like previous technological breakthroughs, AI will
broadly impact the federal budget. Along with IT modernization,
AI can directly improve the efficiency and effectiveness of
government operations leading to savings.
With the industry set to generate additional economic
activity of up to $13 trillion worldwide by 2030, Federal R&D
investments will remain essential to U.S. leadership and
competitiveness in AI technology. However, the benefits will
only be available to all Americans if paired with strategic
investments to support our workforce through this impending
evolution.
The pandemic and economic crisis have already shown that
income security and related programs are crucial for supporting
Americans during challenging times. The shifting job landscape
expected with widespread AI implementation could further
demonstrate this need. This will require strong federal
investments and social programs and affordable healthcare,
childcare, and housing, as well as new approaches for
retraining and upskilling our workforce.
IBM estimates that between 2019 and 2022, more than 120
million workers in the world's 12 largest economies may need to
be retrained and reskilled as a result of AI-enabled
automation. If we fail to plan ahead, the underlying problems
illuminated by the pandemic and recession will continue to
create barriers to success for American workers.
We have a responsibility to get Americans through the COVID
crisis, but we also must address the long-term economic
challenges we know are coming. These issues are complicated and
nuanced, but that is why we are here today. With the help of
our expert witnesses, we can begin to chart a path forward that
leads to inclusive economic growth, broad social benefits, and
a better prepared workforce.
I look forward to learning more about the magnitude of the
potential changes to our economy and job market and the federal
policies that will be needed in response.
I now yield five minutes to the Ranking Member, Mr. Womack,
of Arkansas.
[The prepared statement of Chairman Yarmuth follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Mr. Womack. And I thank the Chairman for holding this
hearing and my thanks to the witnesses who will be with us
today.
I would like to also add my bit of welcome to Chris Jacobs,
the newest Member of the Committee, native of Buffalo, New
York. Long history of public service, Erie County clerk, state
Senator now joining us as a Member of Congress. This isn't his
first tour of duty at the Capitol. He began his career working
for former Congressman and Buffalo Bill quarterback Jack Kemp.
Chris, welcome to the Budget Committee. To your wife and
daughter, Martina and Anna, thank you for allowing your husband
and father to continue his public service career by spending
time in Washington with all of us. Chris, we welcome you to the
Committee.
We are here to talk about AI capabilities, both current and
future, and the impacts on the economy and the federal budget.
It is a critical technology to be sure that will benefit the
lives of many Americans and touch nearly every sector of the
U.S. economy. While it will likely change the way many jobs are
performed as technological advances have for many decades, we
must harness the capabilities of AI to help drive our economy
and society forward.
Congress has to ensure that its actions do not stifle
innovation, rather government should work in partnership with
the private sector to move our country forward in AI research
and development.
By making strategic federal investments in AI R&D,
Washington can help unleash America's pioneering and
entrepreneurial spirit. It also means creating a regulatory
environment that supports, not hinders, private industry by
allowing technological advancements to flourish in a safe,
trustworthy, and effective way. Congress should also move to
encourage more American high-tech manufacturing in general.
The U.S. currently relies on countries located in
geopolitical hot spots for many critical components and as the
coronavirus pandemic has shown with medical supplies, we need
to ensure we have reliable, secure, and diverse supply chains
for vital materials.
Now while this is an interesting, important topic, it
should not be the reason why the Budget Committee is convening
this afternoon, in my strong opinion.
The dire fiscal outlook--notably the recent deficit and
debt projections--and the discussion on how to tackle these
challenges should be the focus of today's Committee meeting.
Last week the Congressional Budget Office released its budget
outlook update, and let me tell you the findings are incredibly
sobering, but not surprising.
We did not do our job when this pandemic--before the
pandemic hit. This Committee is charged with writing a budget
to put our country on a responsible fiscal path, but we failed
in that duty. Once COVID hit, we were obligated to respond to
the crisis. For those of you who don't know, let me summarize
where our nation stands fiscally. And let me just warn you, it
isn't good.
The deficit for fiscal 2020 is protected to be $3.3
trillion, more than triple the previous year's deficit and by
far the highest in American history. Every single year for the
next 10 years, CBO is projecting that annual deficits will
exceed a trillion dollars and total $13 trillion over this
period.
The public debt is projected to be larger than the size of
the entire economy by next year, that is 104 percent of GDP,
and will continue to increase to more than $33 trillion by
fiscal 2030. That is 109 percent of projected GDP. Once again,
CBO confirmed the driver of the fiscal problem, federal
spending, particularly mandatory spending.
Mandatory spending, including interest payments on the debt
is expected to account for 75 percent of total federal spending
by 2030. And I don't need to be the guy to tell you, you
already know. That is squeezing resources for many
discretionary priorities. The job of this Committee is to write
a budget resolution that sets a fiscal path for the government
to follow. We didn't write one. We don't have one. Instead of
considering a budget resolution, we are talking about
artificial intelligence, which is, as I mentioned before, while
an interesting topic and an important topic, it is not the
mandate of this Committee.
The Democrat majority has neglected to do a budget
resolution for the past two years. CBO's projections illustrate
the necessity for the Democrat majority to do its job--write
and pass a budget resolution that provides a responsible,
fiscal framework to correct this current, fiscal trajectory.
With that, I look forward to hearing from our witnesses,
and always look forward to the discussion. Thank you, Mr.
Chairman.
I will yield back the balance of my time.
[The prepared statement of Steve Womack follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Mr. Womack. Chairman, you'll need to unmute, I think.
Chairman Yarmuth. Thank you. Should be up to that by now.
In the interest of time, once again, if any additional
Member has an opening statement, you may submit those
statements electronically to the Clerk for the record.
Once again, I want to thank our witnesses for being here
this afternoon. The Committee has received your written
statements and they will be made a part of the formal hearing
record. Each of you will have five minutes to give your oral
remarks. As a reminder, please unmute your microphone before
speaking.
I now introduce and yield five minutes to Dr. Susan Athey.
Please unmute your mike and begin when you are ready.
STATEMENTS OF SUSAN ATHEY, PH.D., ECONOMICS OF TECHNOLOGY
PROFESSOR, STANFORD GRADUATE SCHOOL OF BUSINESS, AND ASSOCIATE
DIRECTOR, STANFORD INSTITUTE FOR HUMAN-CENTERED ARTIFICIAL
INTELLIGENCE (HAI); DARON ACEMOGLU, PH.D., INSTITUTE PROFESSOR
OF ECONOMICS, MASSACHUSETTS INSTITUTE OF TECHNOLOGY; DARRELL
WEST, PH.D., VICE PRESIDENT AND DIRECTOR OF GOVERNANCE STUDIES,
BROOKINGS INSTITUTION; JASON MATHENY, PH.D., FOUNDING DIRECTOR,
CENTER FOR SECURITY AND EMERGING TECHNOLOGY AT GEORGETOWN
UNIVERSITY, AND COMMISSIONER, NATIONAL SECURITY COMMISSION ON
ARTIFICIAL INTELLIGENCE
STATEMENT OF SUSAN ATHEY, PH.D.
Dr. Athey. Hello, Chairman Yarmuth, Ranking Member Womack,
and Members of the Committee. Thank you so much for inviting me
to speak today.
Artificial intelligence seems to inspire extreme views.
Some focus on a future where robots take all the jobs, while
others point out that so far its effects on the economy are
barely detectable. My own view is that AI has enormous positive
potential for society and for the efficiency and finances of
government, and that governments and universities have a
crucial role to play in ensuring that the potential is
realized.
AI, of course, also creates challenges, contributing to an
era where workers transition more frequently and require more
reskilling throughout their careers. So we need to ensure that
our institutions are prepared to meet this reality, especially
in light of the many fiscal and labor market challenges created
by an aging population and workforce. But when R&D is directed
at technology that augments human workers and support citizens
in their lives and health, we may be able to expand the
circumstances in which people engage in rewarding work while
experiencing a high-quality of life in areas with a more
moderate cost of living.
Some of the most promising areas where technology can be
part of the solution include education, training, remote work,
medicine, and government services. In each case, digital
technology powered by AI can be used to make services cheaper
to provide, higher quality, more tailored to the individual
need, and substantially more accessible and convenient.
The accessibility matters particularly to people with
limited time, like working people with caregiving
responsibilities and especially rural residents who face a dual
burden of high transportation costs and insufficient density to
support specialized services and job opportunities in their
local communities.
One reason the potential is so great for these problems is
that digitization and the adoption of AI can lead to low
marginal cost scalable and thus more efficient services.
Digitization and AI are inextricably linked to measurement
and optimization, which naturally improves the accountability
and effectiveness of the organizations who adopt them,
including the government. In addition, general trends that have
led to the rapid diffusion of AI relate to the lowered fixed
cost in time required to adopt it.
One trend is just a digitization of everything from
consumer interactions to supply chains. That creation of data
is what powers and makes possible AI to be an optimization. The
way IT is implemented has also changed. Cloud computing allows
companies to rent computing as they need it rather than buy
allowing infrastructure to be shared across firms and that
reduces cost.
Software as a service lets companies subscribe to services
and purchase the best products use case by use case and that
software as a service then can also make available AI and
machine learning innovation without firms having to do that R&D
themselves.
Finally, we have seen a big expansion of open source
software and, in general, data management analytics tools are
widely available. They are shared across firms and across
academia, and thus diffuse very quickly. The latest machine
learning algorithms are typically free. For example, for my
class we used algorithms that we downloaded that were trained
using Facebook's image data setting computing infrastructure
allowing the students to move on to the analytics on top of the
image recognition.
The reason that firms are willing to share those types of
algorithms is that it is customer relationships and data, as
well as know-how to optimize the algorithms at large scale that
give companies their competitive advantage.
And the general purpose technologies in algorithms have
actually been fairly widely available. That means the cost of
developing services is reduced as these general purpose
innovation from academia and for-profit organizations can be
repurposed by entrepreneurs, governments, and social impact
organizations.
Now an important precursor to a policy discussion is
demystifying the technology. In practice, rather than sort of
general intelligence, most of what we have seen in the past 15
years can be thought of as more automation on steroids. For
software, automation is like following prespecified rules
without real-time human direction, but the latest innovations
have concerned implementing automation using decision rules
that are learned from past data using machine learning.
And a common example of machine learning and algorithm
might take as input a digital photo and output a guess of what
animal is in the photo.
Traditionally, analysts had to do a lot of manual work to
customize the statistical models so the models were simplified,
but modern machine learning allows the analyst to just feed in
raw data and the algorithm does a lot of work to determine what
is important for the task. This makes things general purpose,
but the fact that they are general purpose also means they are
black boxed and sometimes even the engineers building them
don't understand them. Thus we need a lot more research and
best practices to make sure that this technology is implemented
safely and without unintended consequences.
Just to close, machine learning is diffusing across the
economy use case by use case, but in most cases, this has led
to an incremental innovation and incremental changes over time
rather than sudden shifts.
So I look forward to continuing the discussion in the
question and answer. Thank you.
[The prepared statement of Susan Athey follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Chairman Yarmuth. Thank you for your testimony.
I now recognize Dr. Acemoglu for five minutes. Please
unmute and the floor is yours.
STATEMENT OF DARON ACEMOGLU, PH.D.
Dr. Acemoglu. Chairman Yarmuth, Ranking Member Womack, and
Members of the Committee, thank you for inviting me to testify
on this important subject.
The U.S. economy today and U.S. workers are suffering from
what I view as excessive automation. The extent of automation
is excessive in that it is not leading to sufficient
productivity growth, creating new tasks for humans or
increasing wages.
Automation, the substitution of machines and algorithms for
tasks previously performed by labor, is nothing new. It has
often been an engine of economic growth, but in the past, for
example, during the year of the mechanization of agriculture,
it was part of a broad technology portfolio and its potentially
negative effects on labor were counterbalanced by other
technologies. Not today.
Recent advances in AI and machine learning are not
responsible for these trends. In fact, AI, a broad
technological platform with great promise, can be used for
helping human productivity in creating new human tasks. But it
could exacerbate the same trends if we use it just for
automation.
The COVID-19 pandemic will also contribute to this
predicament as there are now more reasons for employers to look
for ways of substituting machines for workers and recent
evidence suggests that they are already doing so.
Excessive automation has already been a major drag for the
U.S. economy. Private sector spending on workers, which
increased steadily and rapidly almost every year in the four
decades following World War II, has been essentially stagnant
over the last 20 years. The decline in the share of labor in
national income, the stagnation of middle class wages, and a
huge increase in inequality are all connected to our recent
unbalanced technology portfolio prioritizing automation and not
much else.
Excessive automation is not an inexorable development. It
is a result of choices and we can make different choices. While
there is no consensus on exactly what brought us to this state,
we know of a number of factors that have encouraged greater
automation. Chief among these has been the transformation in
the technology strategies of leading companies.
American and world technology is shaped by the decisions of
a handful of very large and very successful tech companies with
tiny workforces and business models centered on the
substitution of algorithms for humans.
There is, of course, nothing wrong with successful
companies pushing their vision, but when this becomes the only
game in town, we have to watch out. Past technological
successes have often been fueled by a diversity of perspectives
and approaches. The dominance of the paradigm of a handful of
companies has been exacerbated by the dwindling support of the
U.S. Government for fundamental research. The transformative
technologies over the 20th century, such as antibiotics,
sensors, modern engines, and the internet have the fingerprints
of the government all over them. The government funded and
purchased these technologies and often set the agenda, but no
longer.
Last but not least, government policies encouraging
automation excessively through its tax code. The U.S. tax
system has always treated capital more favorably than labor. My
own research estimates that over the last 40 years, via payroll
and federal income taxes, labor has paid an effective tax rate
of over 25 percent.
Even 20 years ago, capital was taxed more lightly, with
equipment and software facing tax rates around 15 percent. This
differential has significantly widened with tax cuts on high
incomes, the shifts of many businesses to S-Corporation status
that are exempt from corporate income taxes, and very generous
depreciation allowances.
Software and equipment are now taxed at about 5 percent,
and in some cases corporations can get a net subsidy when they
invest in capital. This generates a powerful motive for
excessive automation. One result of this has been the
disappearance of good jobs, especially for workers without
postgraduate degrees or very specialized skills.
The only way to alter this technology is to redirect
technological change. That will require changes in federal
policy. A first step would be to correct the asymmetric
taxation of capital and labor. This would go a long way, but is
not sufficient by itself.
A second step is to re-evaluate the role of big tech
companies in our lives, including in the direction of
technology. This, of course, goes beyond debates about
automation and AI as it relates to the issue of limiting the
size and dominance of big tech.
These measures can be strengthened with government R&D
policies specifically targeting technologies that help human
productivity and increase labor demand. Research policies that
target specific classes of technologies are rightly
controversial. They may be particularly challenging in the
context of choosing between automation and human-friendly
technologies since identifying these is nontrivial.
Nevertheless, I would like to end my comments by
emphasizing that such policies have been adopted and have had
successes in the past. Four decades ago, renewable energy was
prohibitively expensive and the basic know-how for green
technology was lacking. Today, renewables already make up 19
percent of energy consumption in Europe and 11 percent in the
United States, and have costs in the ballpark of fossil fuel
based energy. This has been achieved thanks to a redirection of
technological change away from a singular focus on fossil fuels
toward greater efforts for advances in renewables.
In the U.S., the primary driver of this redirection has
been the government subsidies to green technologies, as well as
the changing norms of consumers in society. The same can be
done for the balance between automation and human-friendly
technologies.
Thank you.
[The prepared statement of Daron Acemoglu follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Chairman Yarmuth. Thank you very much for your testimony.
I now recognize Dr. West for five minutes.
STATEMENT OF DARRELL WEST, PH.D.
Dr. West. Chairman Yarmuth, Ranking Member Womack, and
Members of the Committee, thanks for the opportunity to
testify. I am coauthor with Brookings' president, John Allen,
of a new AI book entitled, ``Turning Point: Policymaking in the
Era of Artificial Intelligence.'' And I also am the co-editor
of the Brookings technology policy blog Tech Tank and
coproducer of the Tech Tank Podcast.
In my testimony, I am going to argue that artificial
intelligence is one of the transformative technologies of our
time and likely to have major ramifications for the workforce.
AI is being deployed in a number of different sectors, and its
usage will accelerate in coming years. Its development is going
to necessitate rethinking our policies in the areas of
budgeting, infrastructure, healthcare, education, workforce
development, and economic development.
As AI and other emerging technologies become widely
deployed, there are several possible ramifications for the
workforce--job loss, job dislocation, job redefinition, job
mismatch, and job churn.
For example, there could be job losses in entry level
positions as firms automate routine tasks. There can be
geographical dislocations as positions migrate to urban
population centers and there can be job churn as people move
from company to company.
In an economy where benefits are tied to full-time
employment, any increase in job churn will create instability
in people's ability to maintain income and benefits.
Most of the issues noted above have grown worse with the
advent of COVID-19. The pandemic has revealed stark inequities
in access to online education, telemedicine, and opportunities
for remote work. As an illustration, African Americans are far
less likely than Whites to access online educational resources,
but far more likely to suffer from the coronavirus.
It is hard to estimate the precise impact of technology
innovation on the federal budget because there is so many
ramifications for government revenues and expenses. But one
thing that appears clear is we are going to need greater
investment by both the private and the public sectors.
One area is digital infrastructure. Right now there are
around 18 million Americans who lack sufficient access to the
internet. You need an online connection to apply for many jobs.
A number of people do not have the connectivity required for
online education, telemedicine, and remote work. So it is vital
that we close that gap so that all can benefit from the digital
economy.
The emerging economy presents challenges with respect to
ensuring health and retirement benefits. Any increases in
unemployment or people having part-time jobs will create some
hardships. In today's digital world, workers need benefit
portability to survive a turbulent working environment.
Organizations need to shorten their vesting periods for
people to become eligible for company retirement contributions.
Right now many organizations do not vest employees until they
have worked at the firm for one or two years, and if there is
increased joblessness, lengthy vesting periods will lead to
shortfalls in retirement income.
In the world of rapid change it is imperative that people
engage in lifelong learning. The traditional model in which
people focused their learning on the years before age 25 and
then get a job and devote little attention to education
thereafter is becoming obsolete. In the contemporary world,
people can expect to see whole sectors disrupted and they will
need to develop additional skills. The type of work that people
do at age 30 is going to be very different from what they will
be doing at ages 40, 50, and 60.
One possibility to encourage continuing education is
through the establishment of lifelong learning accounts. They
would be analogous to individual retirement accounts or state
government-run 529 college savings plans, but the owners of the
account could draw on that account to finance online learning,
certificate programs, or job retraining expenses.
As America deploys AI and moves to a digital economy, its
two coasts have fared much better economically than the
heartland. According to research by my Brookings colleague,
senior fellow Mark Moro, only about 15 percent of American
counties generate 64 percent of GDP. Far too many parts of the
United States are being left behind. One way to address this is
through regional innovation districts. These are public-private
partnerships that boost innovation in heartland cities. And the
districts include regulatory relief, tax benefits, workforce
development, and infrastructure.
To summarize, it is crucial to think proactively as tech
changes unfold. The longer we wait, the more painful the
transition will be. Now is the time to start having the
discussions required to make meaningful changes. And I applaud
the Committee for providing a platform for this important
conversation.
[The prepared statement of Darrell West follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Chairman Yarmuth. Thank you, Dr. West, for your testimony.
I now recognize Dr. Matheny from Louisville, Kentucky.
STATEMENT OF JASON MATHENY , PH.D.
Dr. Matheny. Thank you, Chairman Yarmuth, thank you Ranking
Member Womack, and Members of the Committee. And thanks also to
my colleagues at the Center for Security and Emerging
Technology at Georgetown whose research I will be drawing from
today.
AI is a general purpose technology with a broad range of
applications in healthcare, agriculture, energy,
transportation, national security, and scientific discovery.
Advances in AI are likely to be applied across many sectors of
the economy spurring growth and enabling new technologies.
Policies to strengthen U.S. leadership in AI have enjoyed
bipartisan support, at least during the decade that I have
worked on the topic.
I worked on AI strategies for both the current
administration and the last administration, and there are more
similarities than there are differences. Both administrations
emphasize the points that I will make here today and each had a
positive outlook on the potential for AI to improve American
health and prosperity.
As Michael Kratsios, the U.S. chief technology officer
recently said, our future rests on getting AI right. AI will
support the jobs of the future. Jason Furman, the previous
chair of the Council of Economic Advisers in the last
administration said that his biggest worry about AI is that we
do not have enough of AI.
So while AI will cause changes to the labor market, this
has been true of every technology since the industrial
revolution and this country has adapted. I believe we will
adapt to AI and will be helped by more economic research on the
likely effects of AI and automation on the labor force. And by
benchmarking to assess progress in various applications of AI.
The United States is in a strong position globally. By most
measures, we lead the world in AI and our lead is key--is due
to key structural advantages. We have an open society that
attracts the world's top scientists and engineers. The National
Science Foundation shows that over the half of the master's and
Ph.D.-level computer scientists who are employed in the United
States were born abroad. We have a competitive private sector
that spurs innovation, and we maintain strong international
partnerships.
While the U.S. alone funds only 28 percent of global R&D,
with our allies we fund more than half. We should double down
on these strengths. We should ensure that we remain an
attractive destination for global talent by broadening and
accelerating the pathways to permanent residency for scientists
and engineers. Most research suggests that increases in high
skilled immigration yield increases in jobs and wages for
Americans due to immigrants' contributions to economic growth
and the creation of new companies.
We should also ensure that small and mid-sized businesses
have access to the computing power needed for AI applications.
We can leverage the purchasing power of the federal government
to buy commercial cloud computing credits in the private market
and award them through federal grants and contracts
competitively as the National Science Foundation has done
through its cloud bank program. We should also strengthen our
alliances and foster the responsible use of AI through
organizations, such as the Global Partnership on AI, of which
the United States is a founding member.
China has made extraordinary technological progress in
recent decades and its future prospects should not be
underestimated, but U.S. policy should be based on an
appreciation of the strengths that have driven our leadership
in AI thus far and how they can be leveraged in the future.
While our private sector leads in AI, the federal
government plays a key supporting role. Federal research
funding laid the foundation for the current wave of AI
progress. Federal funding should continue to focus on areas
where the private sector is likely to underinvest. That
includes basic research, safety and security, testing and
evaluation, and verification and validation.
The National Institute of Standards and Technology should
be given the resources needed to lead interagency and public-
private collaborations on AI testing and evaluation, including
establishment of a national AI test bed: A digital platform
containing public and nonpublic data sets, code, and testing
environments on which AI systems from industry, academia, and
the government can be developed, stored, and tested.
Fourth and last, the United States should ensure that it
has access to leading edge microelectronics. This country is
the birthplace of microelectronics and we continue to design
most of the world's leading edge systems, but most devices are
now manufactured elsewhere.
Offshoring most of our semiconductor industry has increased
the risk of supply chain disruptions during crises. The United
States should strengthen U.S. based semiconductor manufacturing
to reduce supply chain risks and to increase the number of
high-quality jobs at home.
At the same time, we should work with our allies to ensure
that democracies remain at the leading edge of microelectronics
by investing in joint research programs and by enforcing
multilateral export controls on the manufacturing equipment
needed to produce advanced chips.
The United States and our allies produce more than 90
percent of this equipment, so we are in a particularly strong
position. Legislation, such as the bipartisan proposals for the
CHIPS for America Act and the American Foundries Act can help
maintain that position.
With these four points on the benefits of AI as a general
purpose technology, the sources of U.S. leadership in AI, the
federal government's role in supporting the private sector, and
the importance of microelectronics, I thank the Committee for
the opportunity to speak with you today, and I look forward to
your questions.
[The prepared statement of Jason Matheny follows:]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Chairman Yarmuth. Thank you for your testimony.
Once again, I thank all the witnesses for those statements.
We will now begin our Q&A session.
As a reminder, Members can submit questions to be answered
later in writing. Those questions and their responses will be
made part of the formal hearing record. Any Members who wish to
submit questions for the record may do so by sending them to
the Clerk electronically within seven days.
As we usually do, the Ranking Member and I will hold our
question periods till the end.
So I now yield five minutes to the gentleman from
Pennsylvania, Mr. Boyle.
Mr. Boyle. Thank you. I hope you can hear me OK, Mr.
Chairman.
All right.
Chairman Yarmuth. Yes.
Mr. Boyle. Thank you.
So whoever said this in the very beginning was right, that
no topic, perhaps more--while exciting some people, I would
say, inspires more fear, consternation/paranoia than AI. So all
of the presentations were very interesting. I want to go back,
though, to a point that was made very early on by one of the
witnesses because something that I learned, certainly today if
you look at our tax code, we treat capital and labor very
differently and certainly make it much more attractive for
capital versus labor.
I didn't realize that that is not just been a recent
tendency, which was always my impression, but one of the
witnesses stated that that goes back a ways.
So I was wondering what any of the witnesses would think
about constructive ways that we could bring equality to our tax
code, ideas like treating capital gains as ordinary income.
There is a discussion right now obviously, perhaps started
unintentionally by the President in terms of the use of the
payroll tax for funding Social Security and Medicare.
I am curious about these ideas because as my line of
questioning probably suggests, I am certainly one who thinks
that, at the very least, labor and capital should be treated
equally in our tax code and we shouldn't have our thumb on the
scale, which in my view we heavily do in treating capital more
favorably and thus making it actually more attractive for
companies to replace the McDonald's worker with the touch
screen that I now use--I am advertising my bad eating habits,
but--that now we have at so many of our fast food restaurants.
So I will open that up to any of the witnesses and
certainly any ideas or proposals you would have, and if you
agree with my view that capital and labor should be treated
more equally in our tax code.
Chairman Yarmuth. Who wants to take that? Dr. Acemoglu.
Unmute, please.
Dr. Acemoglu. OK. Yes. Somehow I was muting and unmuting
and muting myself at the same time. Thanks for that question.
Yes, I was the one who talked about the taxation of capital
and labor and it is a complex topic. Economists actually
differ, in all full disclosure. There are some economists who
passionately think that capital should be taxed at the very low
rate or not at all, but I think in the age of automation, the
asymmetric treatment of capital and labor in the tax code has a
lot of costs.
If you live in a world where every piece of capital needs
to be combined with some human operators, there are still
problems with asymmetric treatment of capital and labor,
certainly distributional consequences, but there is a better
case that increased demand for capital equipment is going to
trickle down to workers.
But during our current era where automation is so prevalent
from the McDonald's checkout kiosks, to customer service, to
machinery, numerically controlled machinery, robots and
algorithms everywhere, I think the asymmetric treatment of
capital and labor does create more severe problems.
Now if that wanted to be reversed, for example, going back
to the 1980's or the 1990's when capital and labor were still
treated asymmetrically, but the gap was smaller, you know, a
couple of items would help a lot. For example, reversing the
very generous depreciation allowances which were often
introduced during recessionary times as temporary measures and
then weren't completely reversed later. That would be a very
major step.
Then there is also the issue of, you know, why we have
corporations that choose their own tax status, S-corporations
versus C-Corporations, and that has changed a lot over recent
decades.
And often that is a way of reducing the tax base for
capital through some sort of tax arbitrage and I think that is
something that needs to be followed through especially since
Ranking Member Womack said this Committee is going to look for
ways of increasing tax revenues.
And exactly like you have expressed, Representative Boyle,
one has to also consider the taxation of capital gains and
other items.
If you wanted to go on the other side, there has been a
long line of argument in economics going back several decades
that payroll taxes are particularly problematic. And in the
United States that is actually a very important part of the
taxes that labor faces, but, of course, I recognize that right
now, with the budget deficit, makes more sense to think of, you
know, creating that symmetry by increasing the taxation of
capital especially broadening the tax base for capital rather
than eliminating taxes, but certainly payroll taxes are
something to think about in the future as well.
Thank you.
Chairman Yarmuth. All right. The gentleman's time is
expired.
I now recognize the gentleman from Georgia, Mr. Woodall,
for five minutes. Is Mr. Woodall on? Unmute if you are on. Not
answering. Well, then, I will recognize the gentleman from
Ohio, Mr. Johnson for five minutes.
Mr. Johnson. Well, thank you, Mr. Chairman. And I
appreciate you holding this hearing today and many thanks to
our witnesses too.
I am in my car, so I apologize if things are jumping
around, but I am an IT guy. I was in undergraduate and graduate
school in the late 1970's, early 1980's when AI first came on
the horizon. And today there is no doubt, we all know it,
technology spans every sector of our economy.
Investments and emerging technologies, such as AI,
blockchain, and the internet of things have the exciting
potential to drastically improve our economy, national
security, and our very way of life through greater efficiency,
increased global competitiveness, and creating countless other
applications.
In addition to domestic uses for AI, the Department of
Defense has been developing and utilizing AI applications for a
range of military capabilities, including intelligence
collection and analysis, cyber and information operations,
logistics, command and control and also for semi or fully
autonomous vehicles, but we all know, the United States is not
the only country developing AI capabilities.
China, in particular, is investing billions in AI. It is
imperative to our national security that the United States
continues to be the leader in developing AI and other emerging
technologies. However, China resorts to stealing innovation or
subsidizing state-owned enterprises. This is not the role of
America's federal government, nor should it be given the
innovative spirit of the American people and the exciting
advances in technology already occurring right here at home.
Rather it is imperative that our federal government enable
the private sector to flourish by removing barriers to
innovation, something that President Trump and his
Administration has taken important steps to do so.
And supporting private sector research and development
collaboratively through strategic federal investments in
agencies such as the National Institute of Standards and
Technology.
So I am very pleased to have introduced H.R. 6940, the
Advancing Tech Startups Act, which is part of a larger public
energy and tech agenda to create policies that foster American
innovation, secure our supply chains, and protect American
consumers.
Specifically, my legislation promotes a national strategy
for encouraging more tech focused startups and small businesses
in all parts of the United States. It is vital to our national
security to reduce our reliance on other countries, such as
China. And as I have stated, we don't need to rely on any other
country. We should, once again, tap into American ingenuity and
unleash the American innovation and entrepreneurship that we
are famous for.
So Dr. Matheny, some suggest that the United States may be
at risk of falling behind in AI development. For example, some
experts predict China could in the near future surpass the U.S.
and take the lead in AI development. In your opinion, where
does the U.S. currently stand globally on AI? Are we leading
the way or falling behind?
Dr. Matheny. Thank you, Congressman Johnson. I think that
the U.S. has a strong lead, but that we can maintain it by
drawing on our structural advantages compared to China.
First, most scientists and engineers in the world weren't
born in either the U.S. or China, and many more of those
scientists and engineers would prefer to work and live in the
United States than they would like to live or work in China.
That is a great advantage to the United States.
High-skilled immigration was key to our victory during
World War II and during the cold war. We were simply able to
get more scientists and engineers on our side to win.
Second, as we do have a more competitive private sector
and, sir, I think your efforts to empower small businesses
where there is so much innovation is really key to our success.
Thank you.
Mr. Johnson. OK. Are there some actions that we, in
Congress, can take or do we need to do more to maintain our
global competitive edge in AI, especially given where China's
going and the major investments that are in place?
Dr. Matheny. I think there are two things that I would
emphasize. The first is just how important our immigration
policy is to allowing us to lead globally given that this is
one of the asymmetric strengths that the United States has
compares to China.
And the second is our lead in microelectronics. It would be
very difficult for China to match us if we play our cards
right. We shouldn't rest on our laurels, but if we pursue
policies that strengthen our semiconductor industry while also
placing the appropriate controls on the manufacturing equipment
that China doesn't have and that China currently doesn't have
the ability to produce itself and is probably a decade away
from being able to produce itself, we will be in a very strong
position.
Thank you.
Mr. Johnson. OK, well, thanks. Mr. Chairman, I yield back.
Chairman Yarmuth. The gentleman's time is expired.
And I now recognize the gentlewoman from Illinois, Ms.
Schakowsky, for five minutes.
Ms. Schakowsky. Thank you, Mr. Chairman, and thanks to our
witnesses.
So technology has certainly addressed some of the isolation
problems that people have felt during the pandemic, and look
how we are communicating today, so there has been a lot of
important changes that technology and opportunities that
technology has provided for us, but even before the pandemic, I
think there were many, many consumers that reported feelings of
helplessness when it comes to with respect to the digital
economy.
You know, on my subcommittee in Energy and Commerce, which
is the Consumer Protection Subcommittee, we have talked a lot
about technology and its ups and downs, and we know that big
tech has actually allowed fraud and fake news and fake reviews
and counterfeit and stolen products that are thriving on their
platforms.
And we have talked about--they come in and talk about self-
regulation, and I think it is pretty clear that we need--that
that isn't working very well.
But here is the other question. They ask about consumers
are concerned about privacy. So Dr. West, I want to ask you. A
functioning AI needs data, but we also need to protect consumer
privacy.
So in your view, what are the main issues related to
consumer privacy and control and ownership of data that we need
to consider through as we think through the use of AI
technology?
Dr. West. Thank you, Congresswoman, for that question. It
is a great question. Privacy is very important to consumers.
You can look at any set of public opinion surveys and that
often ranks very high on the list next to security.
The problem with our current approach to policy at the
national level is it is mainly based on what is called notice
and consent. Like when you download software or even install an
ad, you get this 20-page document that nobody reads and at the
bottom, if you want to use the app, you basically have to agree
to it.
Nobody reads these. We did a national public opinion survey
and basically found that that to be the case. So my Brookings
colleague Cam Kerry has been doing a lot of work on thinking
about new privacy legislation and what he proposes is basically
get rid of the notice and consent because it is not effective
in protecting peoples' privacy and basically holding companies
responsible for their data sharing practices.
Improving transparency so consumers know more about what is
going on, improve everybody's sense of how--what kind of data
practices are being deployed.
I mean, there are all sorts of geolocation features that
are now built-in to apps. Like if you check the weather,
basically the weather app is localized to you so there is a
geolocation feature there, there is all sorts of privacy
problems that get developed there. So basically we suggest a
greater accountability for companies.
California, of course, at the state level has adopted a
much tougher law than what we have had nationally. We really
encourage Congress to embrace the issue of privacy at the
national level. We don't want to end up in a situation where
there are 50 different privacy laws. I mean, that creates havoc
for the tech companies, makes it difficult for them to
innovate. We need a national privacy law that can really
address the concerns that consumers have.
Ms. Schakowsky. Thank you.
In the remaining time that I have, I wanted to really ask
any of you who wanted to comment on this. You know that AI is
used in policing, in social work, in banking, in healthcare,
and we also know that this is a time of racial reckoning in
this country, disparities that we see, and we had a hearing in
my subcommittee on the issue of discrimination built into
algorithms, built into our technology, built into AI.
And I wanted to just ask whoever to talk about how we can
ensure that there is accountability to make sure that there is
not the kind of built-in bias that discriminates against many
in our population?
Someone grab this.
Chairman Yarmuth. Anyone want to handle that real quickly?
Dr. Athey. I can speak quickly. I think that we have to
consider what the algorithms are replacing, and in some cases
they are replacing human decisions, which have perhaps a
different set of biases sometimes driven by the fact that the
humans are using less information or don't have a full view of
someone's circumstance like in resume screening being too
superficial.
So in principal, when well-designed, when training data is
carefully selected and when best practices are used, actually
digitization can improve the situation and reduce bias, but it
has to be done well and it has to be done carefully.
So I believe that we need more research, we need more best
practices, and whenever they are used in government situations
or regulated situations, we do need to include accountability
and ongoing monitoring in order to make sure there are no
unintended consequences.
Often engineers themselves don't understand the source of
the problem and they won't look for it unless they are asked
to, but they also like to use best practices if those are
delineated and available. And so this is partly just a
maturation of the industry and a maturation of the best
practices.
So I am long-term optimistic, but we have to do the hard
work to make it improve things rather than make them worse.
Ms. Schakowsky. Thank you so much.
I yield back.
Chairman Yarmuth. The gentlewoman's time is expired.
And I yield five minutes to the gentleman from Georgia, Mr.
Woodall.
Mr. Woodall. Thank you, Mr. Chairman, for giving me a
second chance. You have always been gracious in that way.
Chairman Yarmuth. Of course.
Mr. Woodall. Mr. Matheny, I wanted to thank you for
mentioning the stability in national AI policy between the two
administrations that you have had an opportunity to serve. We
tend to focus on the chaos, which I think leads to less
confidence as opposed to the leaders in the room who are
working hand-in-hand administration to administration.
Could you talk a little bit about that? We are about to
come up on another major election. Do you anticipate that
stability in policy continuing whether it is into a second term
of the Trump Administration or the first term of a Biden
Administration?
Dr. Matheny. Thank you, Congressman Woodall. I would expect
there to be continuity in the U.S. strategy on AI. I think
there really has been a bipartisan consensus that I have seen
and a lot of continuity at the Office of Science and Technology
policy, in particular, which I think has done a great job both
in the last administration and in this administration
continuing much of the strategic work that was laid out.
Michael Kratsios and Lynn Parker, in particular, at the
office had been outstanding in coordinating the interagency.
They led a smart and ambitious AI strategy, which I hope to see
continued. It is, I think, one of the best cases of bipartisan
coordination around a key technology topic.
Thank you.
Mr. Woodall. I very much appreciate that.
Dr. Athey, you mentioned not just in your response to Ms.
Schakowsky, but also in your opening statement a need to be
aware of unintended consequences.
Are there particular unintended consequences that weigh on
you in your work or is that just a general admonition as we
plow new ground?
Dr. Athey. Well, I think maybe one--one answer relates back
also to the question about labor versus capital and excessive
automation. In general, firms are going to be thinking
primarily about their bottom line and they are often short-
term.
So it is going to be--you know, a firm might be indifferent
between a worker and a machine from a cost perspective and if
they are indifferent they will go with the machine. But, of
course, from society's perspective, we care about the jobs and
we care about the people and we care about their transitions.
So we do want to think about how are we investing in this
R&D generally. We can't always count on companies to take the
longer term perspective and our national innovation strategy
and R&D strategy can, in principal, develop this general
purpose technology in a way that focuses more on augmentation
of humans.
So one thing about this general purpose technology is that
somebody makes better AI and then lots of people adopt it. And
so if somebody makes AI that helps replace humans, lots of
other people can copy it. But if universities or a particular
company or government invests in AI that augments humans, it is
also the case that that can diffuse.
So I think that we can--we need to be intentional about our
strategies and recognize the places where we as a society care
about the direction of technological innovation so that it
pushes more and makes it cheaper and easier for the private
sector to then pick up augmenting technologies.
A second thing that I worry a lot about is just that the
most recent innovation has been in black box technology. It is
powerful in general purpose because it does the work for you.
An engineer who doesn't know anything about a domain can
plop down modern machine learning and the machine learning will
spit something out, but if it is just applied without a
context, without domain experts, without ethical experts or
legal experts or people who are thinking about your national
security consequences, we might end up in dangerous situations.
And so actually like the privacy and security issues, I
think, are especially concerning when we realize that all of us
are being observed and monitored sort of 24/7 by our phones and
by everything that we do as it all gets digitized. That can
create national security issues in that somebody always does
something wrong, so we are available for blackmail.
And we are also going to see in the future, because it is
so easy and cheap, a lot more worker monitoring, which can be
good for safety. We can make sure people are driving safely. We
can make sure truck drivers aren't asleep. We can make sure
that workers aren't going to be injured on an assembly line if
we use video to monitor them.
But, again, we are creating this massive corpus of
information about people. And we also need to make sure that
that information is applied in a fair way and for benefit
rather than being exploiting in various ways.
Mr. Woodall. Mr. Chairman, I know Dr. Acemoglu referenced
companies that were doing it wrong. I hope as this hearing
continues we will have an opportunity to talk about some of
those companies that are doing it right so that we can benefit
from that experience.
I yield back, Mr. Chairman.
Chairman Yarmuth. Absolutely. I will make sure we do that.
Thank you, Mr. Woodall.
I now recognize the gentleman from Michigan, Mr. Kildee,
for five minutes.
Mr. Kildee. Well, first of all, thank you, Mr. Chairman. I
assume you can hear me OK?
Chairman Yarmuth. We can.
Mr. Kildee. I very much appreciate you holding this
hearing. It is a very interesting and, I think, obviously very
important topic.
I represent an area that has seen a pretty drastic drop in
manufacturing jobs over the last 30 or so years. Often, and
almost exclusively, attributed to trade policy. And while trade
has clearly contributed to the loss of manufacturing jobs in my
region of East Central Michigan, Flint, Saginaw, Bay City area,
clearly technology has played a pretty significant role in that
job loss. We have gone from, in my hometown, of about 80,000
direct manufacturing employees in the auto sector to something
around 10,000 right now; but we still produce about half the
cars that we used to produce.
So that technology disruption obviously has had a pretty
dramatic impact on my community, and now we are trying to
imagine and you are thinking and researching about how AI may
have that same disruptive effect. So I am curious about any
thoughts that any of you have about the pace of development of
AI as it relates to manufacturing and specifically around the
production of automobiles. I know this might require some
speculation, but I think it is really important that we engage
in that speculation.
And I am particularly interested, Dr. West, your references
to the work of Mark Moro, I have a past relationship with
Brookings and did a lot of work around this space, particularly
around communities being left behind.
So I am curious if, maybe starting with you, Dr. West, but
others might comment on those two aspects: The pace of these
trends as they might relate to heavy manufacturing,
particularly the auto sector, and then any thoughts you have on
compensating interventions that we can deal with that might add
to the way we typically deal with trade disruption through
trade adjustment assistance, or something like that, how we
might think about support for those communities that are being
disproportionately impacted by these trends.
So maybe starting with Dr. West.
Dr. West. Thank you, Congressman.
I do worry about job losses, and manufacturing is one of
the areas where there is already a lot of automation and AI
that is being introduced, and we fully expect that to
accelerate.
When you look around the world, there are countries that
have almost fully automated factories right now where it is
basically a bunch of robots driven by AI technology and a
handful of humans just monitoring the computer control panels.
But it is not just manufacturing. Finance is going to be
disrupted. The retail sector, Amazon already has opened a
number of stores with basically no retail clerks. They
basically use computer vision to see what you have put in your
bag or, you know, what it is that you are purchasing, and they
automatically charge you as you are exiting it. So there
certainly is going to be, I believe, an acceleration of job
losses.
And in terms of the geographic thing, the thing I would
worry about here is if you look at venture capital investment,
three-quarters of it now is going into New York, California,
and Massachusetts. So, if anything, the geographical inequity
is going to accelerate. Already, you know, much of the high
tech industry is centered on the East Coast and West Coast and
a few metropolitan areas in between, but much of the country is
being left behind. My Brookings' colleagues in our Metro
Program have done a lot of work on this. This is very
worrisome. It creates political anger. People get upset. They
feel the system is rigged. They feel like they are being left
behind.
So we do need to think about public policies that will
address the geographic aspects. Now, one positive development
is the growing tendency to move toward remote work. It turns
out you no longer have to live in Seattle or San Francisco or
Boston or New York in order to work for these tech companies.
In fact, you know, the real estate is growing so expensive
in those areas that they are kind of pricing a lot of employees
out of that market. So they are starting to rely more on remote
work and telework, and so I think public policy can contribute
to that.
There is a rural digital divide where people living in the
country have less access to broadband and less access to high-
speed broadband. They are less able to take advantage of these
remote work things. So Congress should definitely invest in
infrastructure, development in the broadband area just to
reduce that digital divide so that, as companies start to move
to telework and remote work, everybody can take advantage of
that, including people living in the heartland.
Mr. Kildee. Thank you. It is a fascinating subject. I look
forward to pursuing it further.
My time has expired, so I yield back. Thank you, Mr.
Chairman.
Chairman Yarmuth. I thank the gentleman. The gentleman's
time has expired.
I now recognize the gentleman from Texas, Mr. Flores, for
five minutes.
Mr. Flores. Thank you, Chairman Yarmuth, and I appreciate
you holding this informative hearing today.
I want to note something that you said at the beginning,
that the government moves at 10 miles an hour when the rest of
the economy is moving at 100 miles an hour, and I will talk
about that in a minute.
I personally am excited about the opportunities that AI
brings moving forward. I know that several people are
apprehensive about it, but I think that we as policymakers need
to be excited about it.
A couple of things I want to comment on before I go on to
my questions. Number one is, I think that there has been a lot
of discussion about R&D, and I think one of the essential roles
of the federal government is robust investment in basic
research and development, and I say this from the perspective
that I represent two large tier 1 research and education
institutions and also have a great high tech footprint in
several parts of my district that rely on that, and those
discoveries that come out of the search for basic knowledge
from basic R&D.
The second thing is I think we as policymakers need to be
very careful about trying to get into adjusting the mix of
capital versus labor because, as you said early on, Chairman
Yarmuth, the government moves slowly, and I think we as
policymakers could wind up being well behind where the economy
is if we are not careful with that.
I would like to thank all of the witnesses for
participating. Dr. Matheny, I have a couple of questions for
you. As we are all aware, Taiwan through its TSMC Foundry is a
leading semiconductor manufacturer for many countries, and
particularly we in the U.S. rely on them for AI development.
The first question is this: Does the U.S. rely too heavily
on other countries for AI development, and what can the U.S. do
to put less of this reliance on other countries?
Dr. Matheny. Thank you. Thank you, Congressman Flores. And,
first, thanks also for your emphasis on research. I think one
of the most exciting opportunities is for AI to be applied to
research itself, to accelerate science and engineering. I think
some of the more exciting demonstrations that we have seen on
this include DeepMind's use of AI to solve protein folding
problems, which are really important for biomedical research.
So I hope we will see more of that in ways that can expand the
economy and produce jobs.
To your very good question about Taiwan Semiconductor
Manufacturing Corporation. I think the U.S. does rely too much
on imported semiconductors, which introduces at least three
risks.
The first is that our dependence on manufacturing in Taiwan
means that we have a supply chain that could be disrupted by a
conflict with China.
The second is that Taiwan is vulnerable to having its
workforce and its IP poached because of its proximity to China.
And third is we risk having our own know-how vanish in a
key industry the more we import.
I think Intel's recent announcement that they were
considering outsourcing their most advanced manufacturing,
which would be really the only U.S. based advanced
manufacturing of semiconductors, is extremely worrying. So I
think it is prudent to reshore some semiconductor manufacturing
to the United States, particularly the leading edge chips that
are used to power many of the AI applications that will be
valuable in the future.
And beyond the security benefits, this would also create
new manufacturing jobs for Americans.
Thank you.
Mr. Flores. Thank you.
You know what, you actually answered the second part of my
question that talked about the economic and national security
threats that exist if we rely on other regions and other
countries. Let me ask a second question.
As you are aware AI development requires talented workers
with particular skill sets. In order for the U.S. to continue
leading the way in AI development, it is critical that we
continue to develop domestic talent in addition to attracting
talent from abroad.
When you answered Mr. Johnson's question a minute ago, you
talked about attracting the best talent from abroad. What
policy recommendations do you have to ensure that the U.S.
successfully cultivates a domestic talent supply for AI?
And, for instance, talk in particular about what the
education system will look like for that group of persons.
Dr. Matheny. Thank you for asking the question.
My sister is a school teacher and a great one, so I have a
deep sympathy for school teachers who are trying to teach
computer science and mathematics. These are difficult topics to
teach, but we need to find ways of teaching more of our
students the strong math skills that they are going to need.
Mathematics is really the discipline that is most useful to
succeeding in AI. And we simply need to find better ways of
teaching math to our students and finding ways to teach more of
it.
We also need to address the AI labor needs that aren't in
research and development. I know discussions around tech talent
often center around the scarcest and most educated parts of the
workforce; but a critical talent gap also sits in skilled
labor, and for our skilled labor to compete globally, it will
need help from technology.
China enjoys a manufacturing advantage due to its vast
workforce, which is about 11 times the size of the U.S.
manufacturing workforce. But despite its larger size, the
Chinese manufacturing sector only produces about twice the
amount of value-adds. So the average U.S. manufacturing worker
is about six times as productive as the average Chinese
manufacturing worker.
So reshoring manufacturing will require that we both
increase the parts of our labor force while also increasing the
productivity per worker, which is going to have to be achieved
through both training and technology. One example is cobots,
robotic systems that complement human workers in order to
increase their productivity per person.
Thank you.
Mr. Flores. OK. Thank you, Dr. Matheny. And my time has
expired.
Chairman Yarmuth. The gentlemen's time has expired.
I now recognize the gentleman from California, Mr. Panetta,
for five minutes.
Mr. Panetta. Great. Thank you, Mr. Chairman. I appreciate
this hearing, appreciate this opportunity, Ranking Member
Womack. I apologize if my connection is spotty, but I have two
daughters learning remotely, sucking up a lot of the bandwidth.
I guess it would be in more ways than one, not just virtually
but mentally for their parents; but that is a whole other
story.
Let me just say I appreciate this opportunity to have this
type of hearing, especially when it comes to the risks of
automation for workers, especially for those jobs where
automation only provides a marginal cost in productivity
benefit over the human worker.
Now, I think we all sort of agree that we need to focus on
these workers and how such changes will affect them, but we
need to be very careful not to discourage automation or
technological progress because I think all of us agree that
automation and, yes, AI hold tremendous promise when it comes
to improving our lives and our economy.
Now, it can also eliminate, as we know, some very tedious
tasks so that workers can focus on being more productive, and
it does lower prices for consumers, improving our daily lives,
and raising the standard of living for low-income families. So
because automation has that ability to increase worker
productivity, it is our responsibility to ensure that workers
benefit from their increased value.
But taxing or otherwise disincentivizing automation I don't
think saves jobs, and I do think it will make our economy less
nimble and risk us falling further behind our competitors, like
China.
And that is why I believe we need to continue to invest in
automation and cutting edge technology like artificial
intelligence. That is why we should continue to keep the U.S.
competitive in these areas for the sake of our security.
And yet if the successes in these areas do lead to
displacement, it is our responsibility to help those workers,
and we should be prepared to support those workers, rethinking
our social safety net, how we retrain those workers, bolstering
their workers' rights, strengthening collective bargaining for
higher wages and job security so that the productivity
increases.
That also means we need to study how workers can best
complement automation and artificial intelligence, but we
should not, we should not shy away from these fundamental
challenges by stunting progress and protections for our
national, economic, and food security.
Now, here on the Central Coast, when it comes to food
security, we live up to our jobs, we live up to our
responsibilities. We have a lot of farms, farmers, and farm
workers. And as Dan Kildee will tell you, I live in the salad
bowl of the world because of it. We have a lot of specialty
crops that cannot be harvested like traditional row crops in
the Midwest, concerning corn, soy, and wheat. We have crops
where human discernment as to what is a ripe, safe, and
aesthetically pleasing product is really difficult to replace.
But our ag workforce is very necessary right now.
Unfortunately, though, it is an age thing and it is shrinking.
And the pandemic is highlighting not just how valuable that
workforce is but how vulnerable they are.
Now, obviously, it is a two-prong solution. Yes, one is
immigration reform, looking at the Farm Workforce Modernization
Act that passed out of the House this year. The other, though,
is investment in specialty crop mechanization, dealing with how
you can harvest those types of very difficult to harvest crops.
Now, obviously, the private industry is working more to
develop these technologies and to fulfill that labor gap, but I
believe the federal government has a critical role to play in
helping oversee and scale up these investments, if I may say
so.
Now my first question, Dr. Athey, is as we develop these
types of technologies to save labor, to save our food security,
what steps do you think are necessary to protect existing farm
workers and for them to transition and adapt to these new types
of existing circumstances?
Dr. Athey. Thank you for that question.
I believe that historically we have not really done the
greatest job in dealing with displaced workers in general. In
economics class we teach about, you know, all of the benefits
from trade and, you know, more efficient production of
products; but then as a society we forget about that second
step where you actually get the redistribution done and deal
with the consequences.
But where I am optimistic is that I think we have a lot
more tools at our disposal now to reach people, to use data to
figure out what is the best next step for a worker, what types
of up-skilling will actually work for a person in this
circumstance. And that in turn can help people feel comfortable
in the investment because, of course, for a worker to take
their scarce time and invest in trying to acquire a skill, they
need to have confidence that if they do make that effort and
take that scarce time and money, they will be able to use that
to get a new job.
And so I think we have just had services in the past that
haven't really responded to the individual worker, to the
individual worker's circumstance, and then provided them with
effective training and relocation services.
But I believe that we can do better. I am collaborating
with a project in Rhode Island working with the state
government to try to improve both the data to evaluate training
programs and as well try to help workers to have better
information for making choices. And I think that with
technology and data we can do better, and we can also reduce
the cost of delivery by bringing services to people remotely in
their homes at a time that is convenient for them, so they
don't have to get in their car, they don't have to hire a
babysitter and, you know, sacrifice income in order to receive
the training that they need.
So I am optimistic about the future, but we have to be
intentional about it, and we actually have to execute and
follow through on those types of promises.
Mr. Panetta. Agreed. Thank you, Doctor.
Thank you, Mr. Chairman. I yield back.
Chairman Yarmuth. The gentleman's time has expired.
I now recognize the gentleman from Oklahoma, Mr. Hern, for
five minutes.
Unmute. Mr. Hern, unmute. Oh, you need to be helped?
Mr. Hern. I did it twice. It is good. OK. We are good
again.
Chairman Yarmuth. There you go.
Mr. Hern. Thank you, Chairman, Ranking Member Womack, for
holding this important hearing today, and thanks to all of our
witnesses for being her today. This is a topic that I find
quite fascinating, being an engineer myself.
Unfortunately, due to the unforeseen spread of a particular
virus from China, economic growth has been stunted, and so this
really gets to be a real exacerbated issue right now that it is
up to us really to fix.
The U.S. economy has been forced--it is force built by hard
working, first starting Americans, and we only move forward as
a country if we continue to support innovation and encourage
workers to get back into the workforce.
AI can act as a great catalyst to both needs, and the U.S.
Government should create a regulatory environment which enables
growth and innovation, rather than creating hurdles to both, if
we want and would like to beat China and others in this space
as our own available workforce declines.
My question now is--there has been many answers to the
questions that I had; but one of the witnesses really is pretty
fascinating as we get into it. We talked a lot about the
technical aspects of this.
But, Dr. Athey, let's just talk about the workforce. There
has been a lot of talk about workforce replacement, but we
haven't talked at all about the lack of workforce. And for the
first time in at least a generation government figures show a
larger of open jobs than people out of work. Obviously, this
was pre COVID, but it was only six months ago. And certainly a
lot of us, probably all of us, hope we get back there very
quickly.
And part of that problem is demographics, labor and market
growth. The U.S. birth rate has been falling and is at a 30-
year low, and simultaneously baby boomers are hitting
retirement age, a big force behind the falling number of
unemployed. Some would argue it is the real problem of our
Medicare issues and our Social Security issues. We don't have
enough people working to fund those programs, along with the
aging population.
In fact, McKinsey Global Institute research on the
automation potential of the global economy focuses on 46
countries representing about 80 percent of the global workforce
and has examined more than 2,000 work activities and quantified
the technical feasibility of automating each of them. But the
proportion of occupations that can be fully automated using
current technology is actually pretty small, only about 5
percent.
And if you could speak to that AI as our workforce
continues and declines and our need for consumption grows, I
would like to get what your thoughts are on policies--and I am
being flippant in this; but, you know, if you go back to the
McKinsey group, it forces higher fertility and prevent us all
from getting older, which are two driving forces. And while
that is ridiculous, you can't, there is at least one--and I
would like to piggyback off on my colleague from California
when he talks about immigration. You know, there is a big push,
and the President has pushed for this, for merit-based
immigration, bringing people in that can add to where we need
to go from a technology standpoint to help us continue our
drive for AI.
So as you are aware, AI requires talented workers from
particular skill sets so that we can continue to lead the way
as our witnesses have testified. And so what policy
recommendations do you have to ensure the U.S. successfully
cultivates a domestic talent supply in this space? Will
students need a different education to pursue careers in AI
versus what they are doing right now?
Dr. Athey. Thank you, Congressman, for the question. And
you raised a number of really crucial issues.
Of course, everyone on the Budget Committee I am sure is
acutely aware that the amount of our budget that we are
spending on older Americans is increasing dramatically, and so
we need to really think about how we are going to deliver
services to our aging population more efficiently and also what
can we do to keep people in the workforce, preferably in the
workforce longer, which might be in a second career or a part-
time job that looks very different than how work was done in
the past.
So I think the first important consideration is to think
about what will all of these elderly people need and how can we
help them live independently, live fulfilling lives, and get of
the services they need. I think AI and automation can actually
help quite a bit because some of the things that make it
difficult to work as you age include, you know, physical
challenges, as well as memory challenges and, you know, certain
cognitive aspects of the job, all of which can be alleviated
through augmenting AI or physical robots, which might allow
humans to work longer and focus on the aspects of the job that
involve interpersonal relationships, comforting seniors,
helping them get their psychological needs met.
So you might have seniors helping seniors. It is also the
case that actually there is a lot of service work at that time
that in the end may not be fully replaced by automation.
So I see that this aging population is a challenge, but it
also points our way toward solutions for those people. And,
more broadly, the demographic crisis highlights for us that
immigration will be important because we see a shortage of
workers on the horizon and a shortage of taxpayers in the
working age when you look at the demographics.
It is much harder to predict what is exactly going to
happen to automation in 10 years, but we already know how many
20-year-olds we have in the country who will be 40 in 20 years.
Unless we bring in more 40-years-old, you know, we are kind of
stuck with what we have got.
So we can expand immigration, but we can also think about
how to most effectively use the people we have and allow our
aging population to contribute in meaningful ways as they age.
Thank you.
Mr. Hern. Thank you so much.
Mr. Chairman, I yield back.
Chairman Yarmuth. The gentleman's time has expired.
I now recognize the gentleman from New York, Mr. Morelle,
for five minutes.
Mr. Morelle. Thank you very much, Mr. Chairman, for holding
another really, really important issue facing the country.
Before I begin, I do want to also add my welcome to
Representative Jacobs, who I had the privilege of serving with
in the New York State Legislature, and I am delighted that he
has joined this Committee. I am looking forward to continuing
to work with my neighbor to the west in up-state New York.
I just want to say a couple of things. I think some of the
comments by the other members have been really, really
provocative, and there are a ton of questions here. To me this
isn't a question of whether or not AI, machine learning,
robotics, and innovative technology will reshape the landscape
economically and as it relates to the workforce. It is doing
it. It will continue to do it. It is happening, in many
respects, at breakneck speed. And I think then the question for
us, we have always marketed ourselves as a nation of
opportunity, a nation of innovation.
So the question is, as public policy challenges emerge
because of it, what do we do? How do we think through this? I
think that is why this hearing is so critical.
The way I see AI, I guess I think about it in a couple
different buckets. One is, to the extent that it could displace
human beings in some jobs and in some occupations, the more I
see it as ways to create tools that will allow people to do
their jobs faster, better, more efficiently. But there is no
question it is going to have an impact and we need to think
about it.
One of the things, as it relates to the budget--and perhaps
people can talk about this--you know, the President has talked
about elimination or deferral of payroll taxes. Obviously, that
has an impact on Social Security. It has an impact on Medicare.
But even beyond the call for reduction of payroll taxes, to the
extent that there is a displacement of workers or lessening of
wages because jobs become a focus of commodity-like activities,
what I am struck by is so much of what we have built on the
safety net, Medicare and Social Security being two of the most
obvious, built into a system where we get revenues based on
payroll.
So, you know, there have been suggestions by some folks
looking at this, to the extent that we look at displacement,
should there be alternative ways of looking at taxation so we
can continue to provide resources to Social Security, to
Medicare to make sure that particularly as the baby boom
generation starts to move into some of these programs, you are
going to see this significant percentage of the population in
Medicare, in Social Security, and given the reproduction rate
in the United States is at an all-time low, and mix that in
with AI and machine learning, robotics.
Could anyone--and perhaps, Dr. West, maybe you can help
answer this. Is there something we should be looking at in
terms of a replacement for payroll taxes that is based on--I
know people have talked about the difference between capital
and people when it comes to investment and tax payments.
Can you talk a little bit how we can make sure that our
revenue base doesn't decline if we see jobs displaced by either
AI, robotics, machines, et cetera.
Dr. West. That is a great question, Congressman.
I think we do need to think about the tax system both in
terms of tax rates, tax credits but possibly also new types of
taxes. And if you go back a hundred years to the start of the
industrial revolution, you know, we found our tax system to be
inadequate at that time, and so we developed new taxes, we
developed new social programs. And I think now as we are moving
to the digital economy, we need to be asking big questions like
that.
So I am not sure exactly what the kind of new taxes could
be, like people propose a financial transactions tax that would
kind of help with income inequality in general. Some countries
are implementing digital services taxes. So there is a lot of
new ideas that are being formulated there.
And on the first part of your question, you are right about
the importance of market competition, and the key in innovation
has always been small and medium size enterprises. We are
worried about a loss of market competition, and so I think
Congress should really think about ways to promote small and
medium size enterprises just so we can maintain the startup the
economy that has fueled American prosperity for several
decades.
Mr. Morelle. Yes, thank you. I think that is a really
important comment.
And I would just say in the few seconds that I have left,
what I do worry about is we don't want to create disincentive
for investment in innovative technologies. We also don't want
to put ourselves in a position where, as a result of that, we
have displaced workers and the payroll taxes that support much
of our social infrastructure.
So I want to thank the panelists. Thank you, Mr. Chair, for
a great hearing. I yield back.
Chairman Yarmuth. Thank you. The gentleman's time has
expired.
And now I recognize, in his debut Budget Committee
appearance, Mr. Jacobs from New York.
Mr. Jacobs. All right. Can you hear me, or no?
Chairman Yarmuth. You are live.
Mr. Jacobs. I am having some problems here.
Chairman Yarmuth. You may have muted yourself.
Now you are fine. You should be good.
Mr. Jacobs. Can you hear me?
Chairman Yarmuth. Yes.
Mr. Jacobs. OK. Sorry about that.
Thank you, Mr. Chairman. Thank you, everybody. Great to be
on. And I guess I need a little AI to help me with the unmuting
here.
I just wanted to first comment, Dr. Matheny, on some things
you talked about regarding semiconductors. I have an area,
Batavia, New York, in my district where they have been working
for a number of years in developing an advanced manufacturing
park. One of their hopes would be to lure a semiconductor
facility there because of the inherent assets we have in terms
of low-cost power due to the proximity of the Niagra Falls
Power Plant and also abundant water.
And in talking with them, they discussed this issue of the
loss of our semiconductor industry nationally, and one
statistic I just wanted to echo why this is so important what
you are talking about, in the year 2000, the United States had
24 percent of the semiconductor production in the world. Now we
are at 12 percent. In the year 2000, China had zero percent of
the production, and now they have 16 percent, and they are
investing another trillion dollars in this sector in the next
decade.
So, you know, this is a major issue and look forward to
pushing for policy nationally that will help level the playing
field so that we can make sure that we maintain and grow this
sector for the important reasons that are mentioned here.
I wanted to ask a question of Dr. West. My district, as we
talk about inequalities, would be geographic. My district is
rather rural and definitely have concern--we have major issues
with lack of high-speed internet access, and it is being more
pronounced right now with the needs for distance learning and
telehealth, but in an effort to be economically competitive in
the future.
And I was just curious--so, clearly, I am all in for any
additional money and programs to push for rural broadband
because it is important as any other piece of infrastructure
right now for our area. But in terms of you mentioned
innovation districts as the model of something to try to allow
areas that are not on the coast to be competitive in the new
day era, and I was wondering if there is examples of success
that you have of innovation districts? Thank you.
Dr. West. Well, it is funny you should ask that--and the
Chairman will love this because Louisville is actually an
example. Louisville is an example where they have developed a
pretty successful regional innovation district. Brookings
actually is helping advise some of the organizations there. It
is a public-private partnership. So you can talk to the Chair
about how they did that.
On the rural part of your question, I can really appreciate
this because I grew up on a dairy farm in rural Ohio many years
ago, and rural areas are really being left behind right now. So
we really need to address the infrastructure part and
especially the broadband part because, as I mentioned earlier,
like there are opportunities for remote work, like you don't
have to live in San Francisco, you could live in your district
and still work for any of these tech companies, but you need
high-speed broadband.
Just this week my Brookings' colleague, Tom Wheeler, had a
short report where he gave a couple of very specific ideas for
the Federal Communications Commission, which he used to head.
One is a reform of the E-Rate program, which was set up to
connect classrooms. It turns out there is a $2 billion surplus
in that fund, meaning there is unspent money that was designed
to connect classrooms. Now that so many people are engaged in
home schooling, you know, we could actually redirect some of
that $2 billion to improve rural broadband in order to
facilitate home schooling. It is very consistent with the
purpose of that program, so you should talk to the people at
the FCC about that.
And then, second, with the Lifeline program, which the FCC
also runs, including cable companies, not just phone companies,
in rolling out digital services and broadband, just because
today people are almost as likely to get their broadband via a
cable company as a phone company. So if we could broaden the
Lifeline program to basically address the ways people are
ordering broadband, that would help, and also making--including
companies that offer prepaid services.
So in the Tom Wheeler post, he talked about all of these
ideas. But I think they are particularly relevant for your
district and other rural areas across America.
Mr. Jacobs. Great. Thank you very much.
I yield back the rest of my time.
Chairman Yarmuth. The gentleman yields back.
I now recognize the gentleman from California, Mr. Khanna,
for five minutes.
Mr. Khanna. Thank you, Mr. Chairman. And thank you, Dr.
West. I highly recommend Mark Moro's paper, and it is not just
in Chairman's Louisville district, but even in Paintsville,
Kentucky, they have had quite a lot of success in bringing
technology to rural communities, and I appreciate your work and
Brookings' work on that.
I had a question for Professor Acemoglu, whose work I
admire very much. I was struck by this idea of excessive
automation, and I understand the tax incentives that may be
off, but bracketing that aside, what explains the move toward
excessive automation? Is it a sense that there is some kind of
market failure where companies are actually making irrational
decisions to automate in ways that aren't profitable, or is it
that it is marginally profitable but it is not having aggregate
productivity gains for society?
Dr. Acemoglu. Thank you very much, Congressman Khanna. I
think that is a great question. And it is a variety of factors.
First of all, it is indeed the tax incentives, so we cannot
ignore that. You know, there is no natural rate at which
capital and labor are going to be taxed, so it is a policy
choice, and that policy choice is going to have consequences.
A second important factor is that labor and capital use may
have social consequences and economic consequences that go
beyond what companies calculate.
So, for example, if people are better citizens or they
contribute more to their community or to their families when
they are employed, that is not going to be part of the
calculation of companies, and it is part of policymakers to
actually decide that.
So do we, for the same GDP, would we be happy when that is
produced by humans partly versus when a lot of it is produced
by capital? I think a lot of policymakers would say actually
for the same GDP, we would like it quite a bit if humans are
part of that equation, which means that we actually value as a
community, as a society, humans being part of that calculation.
And technology has gone in a way that makes it possible for
greater substitution of machines and algorithms, so some of
those external effects that were less relevant now become more
relevant.
And the third factor is that it is not necessarily
irrational, but different companies have different business
models. So if you look at the periods in which the American
economy has done very well while it was also automating, this
diversity of perspective, diversity of approaches was very
important.
Let me give you one example. Mechanization of agriculture.
That is an even more transformative automation event than the
ones that we are talking about right now. More than half of the
U.S. economy was agriculture, and there was a huge, tremendous
decline in labor share in agriculture as machines started
performing tasks that were previously done by humans.
But during that period, American growth didn't just come
from agriculture. It also came from other sectors that picked
up labor that was displaced or the children of the labor that
were displaced often because some greater human capital was
necessary. So the manufacturing sector introduced a lot of both
production and non-production jobs, a lot of the non-
manufacturing sector expanded.
So it is sort of diversity of approaches, diversity of
technologies was quite critical. So one of the things that may
be less active today is that we are not using the enormous
technological platform that AI presents us in ways that can
create jobs, tasks, opportunities for labor in other sectors of
the economy.
So, for example, when earlier on there was a discussion of
robots and what was going on in Flint, Michigan, you know, that
is absolutely central that there were a lot of production jobs
that were eliminated.
The same has happened everywhere. If you look at South
Korea, if you look at Germany, other countries that have
introduced a lot of robots, production jobs were eliminated in
more or less the same number as in the United States. But in
many of these cases, there were also non-production jobs that
were created more or less simultaneously, sometimes in the same
companies, sometimes in the same markets, and that is what we
haven't seen in the United States.
When you look at Flint, when you look at Saginaw or other
parts of the industrial heartland, you have these production
jobs disappearing, but we are not using the technology to
create other jobs to compensate for this.
Mr. Khanna. Very briefly, how would you create other jobs?
What would be one or two bullets points of what we could have
done in Flint to create those other jobs?
Dr. Acemoglu. Well, I think in Flint, you know, it is a
little bit hard for me to say from here what exactly the skills
that would be easily transferable. But when you look at
broadly, you know, there are many applications of AI in
education, in healthcare, in manufacturing that are completely
capable of creating jobs.
For instance, automation in manufacturing also enables job
creation because it reduces offshoring, so there is mounting
evidence that, you know, not the jobs that were destroyed to
trade with China or to the first wave of automation are not
going to come back. But there are certainly opportunities for
many jobs to come back, offshore jobs to come back as the
automation process continues because it is a cost-saving
possibility.
So many of those are not in the production line. They are
in the supporting capacities. But they are very, very important
and potentially high wage jobs. And, again, evidence from
Germany suggests that many of the jobs that were created, even
in the same companies that were automating at the same time,
were paying higher wages or comparable wages to the production
jobs that were destroyed.
Mr. Khanna. Thank you.
Chairman Yarmuth. The gentleman's time has expired.
I now recognize the gentleman from Virginia, Mr. Scott, for
five minutes.
Mr. Scott. Thank you, Mr. Chairman. I had to find it.
I am sorry, I came on a little late, so let me just ask a
couple of general questions.
First, to any of the panelists, how real is the threat to
laid off workers that their employers might decide to increase
artificial intelligence rather than rehire their workers?
Chairman Yarmuth. Any takers?
Dr. Acemoglu. I can give a quick answer to that.
We don't know. We don't know for sure, but in recent
surveys, about 75 percent of companies are saying that they are
either taking steps to increase automation or they are planning
to do so. So there is a real possibility that some of those
jobs will not come back even if the economy picks up.
The other issue that we need to think about is that the
sectoral composition of the economy is going to change in a
post COVID-19 world. The hospitality sector will probably be
much slower to come back, so there will be a natural
reallocation.
Some of that reallocation is, obviously, healthy and
efficient, but it will still have great costs on some of the
poorer communities and some of the poorer segments of U.S.
society.
So I think those are, as some of the earlier comments
indicated, questions related to the social safety net; but
broadly--and this has been one of the main themes that I have
tried to emphasize--it is not just a social safety net issue.
If we think that displacement is just a social safety net
issue, that would mean that we would be happy to have a lot of
workers being displaced and find ways of providing good social
services and a decent standard of living to them.
But, again, I don't think that would be a healthy economy
or a healthy society. That is why it is important for us to
find ways of using our existing and technology know-how and our
technological capabilities in order to find ways of deploying
our enormously productive, our very well skilled workforce in
other activities.
Thank you.
Mr. Scott. Let me ask you a followup question to that. If
there are going to be fewer workers, does that have budget
implications on your people having taxes and, therefore, lower
revenues?
Dr. Acemoglu. Oh, I am glad you asked that. That is a very,
very important question as well.
So if you--one of the themes that I emphasize is that our
tax system is asymmetric. It taxes capital less than labor, and
it has become more so. That has major budgetary implications
because if you look at the U.S. distribution of income, the
share of labor has gone down from around 67 percent of national
income to less than 58 percent.
So that means that income is shifting away from the more
heavily taxed factor to the more lightly taxed factor, and it
will have budgetary implications.
And another theme that I have tried to emphasize, but it
was very quick so this gives me an opportunity to underscore it
one more time, is that part of the reason is because our
capital tax base is very narrow.
It is not just a question of jacking tax rates on capital
and introducing huge wealth taxes or anything like that. There
is just a big chunk of capital income that we don't tax, and
that means it is costly, it is asymmetric, it may distort the
allocation of capital and labor in work places, but
increasingly has major budgetary implications.
Thank you.
Mr. Scott. Does that include--you know, we have tax credits
for investments in machines but not in education. Is that
something we ought to address?
Dr. Acemoglu. Absolutely. Absolutely, 100 percent. If you
look at decline in the tax rates basing capital that went from
over 15 percent to less than 5 percent in the last 20 years,
about half of that is because of the very generous investment
tax credits, which are so generous that if you have debt
financed capital investment in software or S corporations, you
may actually be getting a small net subsidy. We have nothing
similar to that for education or training.
Thank you.
Mr. Scott. Well, talking about education and training, to
get into an AI job, you don't sign up for an education for AI.
I heard math is important, but what should the Committee on
Education and Labor be doing for higher education?
Dr. Matheny. I can take a piece of this.
One thing that I think would be especially useful is a
tithe, a 10 percent allocation for public research grants to go
toward teaching because otherwise we are eating our seed corn.
We are spending all of our Federal R&D on the research rather
than on the teaching. And in most of the major universities
where AI is being taught, there is a natural tension for the
professors to focus on research as opposed to allocating time
to teaching. We need to make sure that we are training the next
generation, and a tithe, particularly on NSF grants, could help
with that, turning it over to others.
Mr. Scott. Well, thank you, Mr. Chairman. My time has
expired. Thank you so much.
Chairman Yarmuth. The gentleman's time has expired.
I now recognize the gentlewoman from Texas, Ms. Jackson
Lee, for five minutes.
Ms. Jackson Lee. Thank you very much, Mr. Chairman, and
thank you very much for the hearing both with you and the
Ranking Member, and thank you to the panelists.
Let me just add a description that should not be taken as
an offense, but we are all speaking now to the have's because
the have-not's are not in the room. And I think this is a very
important basis upon which we are responding because that is
the focus that I will have, along with maybe a more definitive
question about a tax scheme that would work to help AI.
I am going to start with Dr. West, who early in his
testimony mentioned the question of income inequality and
worker dislocation. Those people today are not in the room. We,
as Members of Congress, represent a wide landscape of
individuals.
Can you pointedly, Dr. West, talk about what should be our
response on the apparent and existing income inequality and the
potential worker dislocation?
Dr. West. That is a great question, Congresswoman, and you
are exactly right. There are income disparities. There are
racial disparities. This is a huge problem. We are almost in a
situation where technology is helping to fuel the inequality in
the sense that the have's are doing better and getting tax
breaks and have programs that support them, and people at the
lower end aren't even in the game. They don't have access to
the digital economy. There are 18 million Americans who do not
have broadband. A larger number doesn't have a high-speed
broadband.
So the way that we need to address these issues, certainly
infrastructure investment, the things we have talked about
earlier, a rural broadband, in underserved urban areas as well,
putting more money into education, and especially opportunities
for online education, because that would be a way to help
overcome the disparities; but you need the broadband in order
to be able to access that.
The same thing applies in terms of telemedicine. One of the
features of COVID is it has jump started what already was in
existence, a trend toward telemedicine, and has really
accelerated it, but not everybody is able to share in the
benefits of that. And given the racial disparities in the
incidents and fatality rates of COVID, like that is a scandal
that people who need it the most are not getting access.
So there are a lot of different things we need to do, and
we certainly need to address the inequities in the tax system.
Ms. Jackson Lee. Well, clearly, it means that out of the
Budget Committee we should be focusing on just the
infrastructure that you mentioned. It is a shame that in 2020
we are still fighting to get broadband everywhere, and for
those of us who are watching our schools open and they are
hybrid or virtual, to see people standing in line trying to
simply get laptops because they don't even have that and as
well hot spots or the hot boxes so they can have the
opportunity to have access.
Let me do a round robin question dealing with COVID-19. We
have heard a very stark admission of the knowledge of how
deadly COVID-19 was as early as February 7, 2020, if I might.
Let me ask all of you to comment how COVID-19 could have been
attacked, if I might, starting with Dr. Athey and going to Dr.
Matheny, with artificial intelligence in terms of treatment, in
terms of outreach, in terms of saving lives.
Doctor--is it pronounced correctly, Dr. Athey?
Dr. Athey. Yes, Dr. Athey. Thank you very much for the
question.
And I think the telemedicine point is super important. We
were a little slow getting started in trying to get information
to people, getting people in touch with their doctors without
broadband access and without good access to medicine. We
weren't always making good decisions for patients early.
Another thing is that actually using AI machine learning to
understand what treatments work best was actually very limited
in the United States by our disjointed medical system and the
inability to do analysis that incorporates data from multiple
sources because, as the epidemic happened, patients were being
treated in hospitals. The insurance companies only get the data
later once bills have gone out, and that is not fast enough.
So it turned out that we were just unprepared to be able to
do analysis that spanned multiple medical centers and give
real-time intelligence. We also missed opportunities to have a
more coordinated approach to clinical trials and R&D that was
really focused on getting the most information and the best
treatment decisions possible given the patient flow that we
had. There was just a lack of coordination.
And I really hope that if anything like this ever happens
again, we are prepared to be able to do the right analysis and
coordinate the studies and the research, and that just requires
really advanced preparedness and a lot of kind of air traffic
control from the federal government. And AI machine learning
can only do their work if they are given the opportunity to
access data and really influence decisions.
Thank you.
Chairman Yarmuth. OK. The gentlewoman's time has expired.
I now recognize the gentleman from New Jersey, Mr. Sires,
for five minutes.
Mr. Sires. Thank you, Mr. Chairman. Can you hear me?
Chairman Yarmuth. I can hear you fine, yes.
Mr. Sires. I want to thank the panelists for being here
today.
You know, I always think of a job as something that creates
self-worth in a person, and we seem to be obsessed with this
productivity word and, obviously, artificial intelligence
creates a lot of productivity. But you have countries like
China and you have countries like India who have such large
populations, and as artificial intelligence is more productive,
more and more people are left behind.
Do you think that these countries with such large
population will ever come to a point and they say, OK,
artificial intelligence is great, but we have passed beyond the
ability to provide jobs for the people of my country. Maybe we
should slow down this artificial intelligence that is creating
so much automation and leaving, so many people behind because,
as you know, if there is no work in a country, it leads to
unrest.
I just wonder if any of the panelists would want to address
that where a country would say, hey, let's put a little brakes
on this because our population is staying behind, is being left
behind.
Can anybody talk to that a little bit?
Dr. Acemoglu. I would be happy to. I would be happy to
comment on that.
Mr. Sires. OK.
Dr. Acemoglu. You know, I think, first of all, I completely
agree a job is much more than just productivity. I think self-
worth is important for the community, important for society. I
think these are critical. But the tragedy in some sense is
that, at least on the current measurements, we are not even
doing that well on productivity. Despite the bewildering array
of technologies all around us and all of this excitement that
goes on, we are actually enduring one of the eras in our
history where productivity growth is lowest.
This goes to underscore what I was trying to emphasize,
that it is not a question of AI versus not AI. It is a question
of how we are using AI technology. And if we are not using it
well, we would destroy jobs and all of the self-worth and
community contributions that we are talking about and also not
reap all of the benefits in terms of productivity.
I think that is exactly the sort of situation that we are
in right now, so a lot of AI goes into marginal activities,
such as self-checkout kiosks or things that humans can do very
well, then it will not bring the productivity gain. I don't
think that China is ever going to turn back from AI, partly
because they have made a huge investment in that, but also
because part of the AI's appeal to authoritarian regimes is
that it actually provides a much better monitoring system,
facial recognition, snooping on communications, control of the
internet. But those are exactly the sorts of things that are
not going to bring huge productivity gains and they are not
going to contribute to making jobs more meaningful.
But if you look at what American companies invest in, it is
not that different. We pour a lot of money into facial
recognition and monitoring aspects of AI as well. So that,
again, goes to my broader point, that I think there are ways of
making use of the AI platforms in a manner that is going to
bring much better social benefits and jobs and productivity
than we are doing currently.
Mr. Sires. So the productivity to work ratio for the United
States is 6:1, as somebody mentioned before, and in China it is
1:1. So I was just wondering, if China does not want a 6:1 or
an 8:1 productivity ratio.
Dr. Acemoglu. I would say China definitely wants that and
has made huge progress----
Mr. Sires. But doesn't that leave a lot of people behind? I
mean----
Dr. Acemoglu. Right. So----
Mr. Sires. If you reach that kind of productivity like in
the United States?
Dr. Acemoglu. Well, it may or may not. If consumption keeps
up with it and that productivity gain is broadly distributed in
society, it may not. In China, it hasn't taken that form. The
inequality has actually increased a lot, the gap between cities
and rural areas and even within cities between migrant workers
and non-migrant workers have opened up hugely.
But, sure, I think there is a huge drive in China toward
increasing labor productivity, but they are also willing to
invest an enormous amount of resources in order to monitor
these workers better, in order to monitor their communications,
the civil society participation, and other social activities,
even if those things aren't proactive because they do need to
maintain the current political system.
Thank you.
Mr. Sires. Thank you.
Thank you, Chairman.
Chairman Yarmuth. The gentlemen's time has expired.
I now recognize the Ranking Member, Mr. Womack, from
Arkansas for 10 minutes.
Mr. Womack. Thank you, Mr. Chairman, and thanks to all of
my colleagues who have participated, and to the witnesses,
thank you very much. A very interesting discussion.
I am going to start where I kind of left off in my opening
remarks, and that was about matters of fiscal accountability at
the federal level, and I don't need to tell anybody engaged in
this forum this afternoon that we are in some very difficult
circumstances right now.
COVID has exacerbated it three times more so than what we
would have otherwise had in terms of a deficit goes.
And to my colleagues on this call today, I will sound a bit
like a broken record. I am an appropriator by nature. I just
happen to be the Ranking Member of the Committee and formally
Chaired the Committee.
But as an appropriator, I am very concerned about the
escalating cost associated with mandatory spending, how much of
the federal budget it is commanding and the squeeze, as I call
it, that crowding out effect that it is having on the matters
of the discretionary budget that we appropriators are in charge
of, should be in charge of--the last few years is an exception,
but that is a whole other story.
Mr. Womack. But the fact is, that if we are going to invest
in anything in our country, if we are going to ask the federal
government to have a role in resourcing a lot of this R&D, then
it is going to face continued and escalating pressure from the
mandatory outlays that continue to consume a larger and larger
share of federal resources.
So here is my question and I am going to pose this to Dr.
Matheny, and that is, with that in mind, if we can all agree
that there is this crowding out effect, how would the federal
government prioritize spending on matters of research and
development and so forth in the AI spectrum?
And what have you seen from your federal government, if
anything, that is worked? So help me understand how we would
prioritize the spending that goes toward a more robust AI
circumstance in our country?
Dr. Matheny. Thank you. It is a great question.
I think that the federal government can most cost
effectively focus on basic research, on testing and evaluation,
and on safety and security, areas that suffer market failures
so that the commercial sector is likely to under invest.
Much of the current wave of AI research that we see right
now is due to federal investments in basic research,
particularly by the National Science Foundation and by the
Office of Naval Research, and DARPA, dating back to the mid-
1980's, which funded early work on deep learning, provided
training grants to much of the current generation of AI
researchers.
And that work in basic research really does need to
continue so that we fund the next generation of breakthroughs
that will fuel future AI systems. Equally important is the work
of the National Institute of Standards and Technology in its
bench marking and its testing and evaluation that has been
critical for actors in both the private sector and the public
sector to be able to bring their tools to have them tested on a
level playing field, understand where they work and where they
break. And just as important also has been the federal
government's investments in microelectronics.
In the 1960's, NASA effectively started our
microelectronics industry, but we also have examples of less
successful programs, very large projects, overly broad goals. I
think the Strategic Computing Initiative which ran from the
mid-1980's to the mid-1990's is an example of that.
So where the government can help is really on the cases
where the commercial sector isn't going to invest on its own,
where the goals, though, can be clearly defined, and where we
can lift up and address those market failures.
Thank you.
Mr. Womack. You bet.
Dr. Acemoglu, am I even close on the name? OK. Good. Thanks
for the thumbs up.
And Mr. Sires reiterated the point you made--I think it was
you that made the point early on in your testimony--that China
had 11 times as many people in the workforce, but they were
only two times as productive. If the American worker has proven
to be the most productive worker, I guess, on the planet says a
lot about our ability and about our capacity.
One of my concerns has been is that the pace of the private
sector in virtually every area is a lot faster than our
education system seems to be trying to deliver.
Is that a fair statement?
Dr. Acemoglu. Yes, I believe so. I think it is definitely
true that our education system has lagged behind. AI, for all
the reasons that we have discussed today, has already started
changing the labor market and it will change it even more, but
our education system, both at the university level but also at
middle school and high school level, is very backward looking.
We continue to teach in the way that we used to, you know,
for the most part, 30, 40, 50, 60 years ago. AI actually
provides--I think Susan mentioned this already--provides
tremendous opportunities for revolutionizing many of the key
sectors such as healthcare and education.
AI can be used for taking over some of the tasks that
educators do that are quite boring, such as grading, but even
more importantly, it can create a much more interactive
classroom, enable teachers to understand the specific
challenges and needs of students and cater their teaching and
curriculum to their needs in real-time. It can enrich what we
teach and how we teach it. There are already companies that
have completely transformed their training systems using AI.
So I think there are a lot of opportunities, but sure, we
are lagging behind. And it is absolutely critical, as you have
pointed out, for our success that the American worker maintains
their productivity edge over other nations, but we have not
done very well in that regard either.
If you look at an inclusive measure of productivity growth,
what the economists called total factor productivity growth, in
the three or so decades following World War II that was growing
over 2 and a half percent a year and it is around 1 percent for
the last 20 years.
So we are not really doing enough to keep our productivity
edge relative to other nations, many of whom have faster
productivity growth rates.
Mr. Womack. Mr. Chairman, my final question--and I am going
to throw this on the table. I don't know really who to direct
it to, but we have seen some challenges in recent years of
building government industry partnerships in what is to me even
more disconcerting is a lot of the companies that we are
talking about are now not only not building those partnerships,
but they are just unwilling to work with the federal government
or work with, you know, partner nations or you can pick from
the spectrum of issues.
Just last year, Google pulled out of a major AI project
called Project Maven with the Department of Defense. It is my
strong opinion that we need to see some change in this area.
Why is this occurring and what are the long-term
consequences of not having the proper relationships between
government and industry? And is it the slow pace of government
in general because we all know that we don't operate with a lot
of speed?
Dr. West. And Congressman, I would be happy to jump in on
that question. And I agree, it is important for a government to
work with industry. I think the Google thing, there was some
idiosyncrasies to that decision. Other tech companies are
embracing the role of working with the federal government, but
I do think, as part of your concern about debt and deficit
issues, we do need to think about agency modernization just
because we have to get the federal government acting much more
efficiently than it is right now just in terms of the
administration of services.
And the way to cut some of the program costs without
hurting the beneficiaries is to make the organization more
effective. And so the public sector still lags the private
sector in using AI.
Just one quick example, every federal agency should be
using AI for fraud detection. It is something that is very
common in private companies. We know there is waste and fraud
in the federal government. The AI looks for outliers, it looks
for unusual activities. Like, this is one tool the federal
government agencies should be using to try and get a better
handle on the spending side. I think that is an example of
where technology can be part of the solution.
Mr. Womack. Yes. And in the budget when I was Chairman that
we prepared for Fiscal Year 2019, a key component of trying to
do the deficit reduction was the fact that we had billions and
billions of dollars of improper payments and there has to be a
way that we can get after those without unnecessarily burdening
ourselves.
Anyway, my time is expired. Thanks, again, to all the
panelists. Thanks to my colleagues. Chris, welcome, again, to
the Budget Committee.
And Mr. Chairman, as always, I am going to yield back my
time and with regrets that you didn't get to see Authentic win
the race there on your home track, but nonetheless a good
derby.
Chairman Yarmuth. I thank the Ranking Member. I did watch
it on television. Fortunately, I tried to set up a betting
account and they were so swamped with people trying to do that
I couldn't get on, so I didn't lose anything. That was a plus.
Mr. Womack. Well, AI, if we had a little better AI
platform, we could have probably fixed that early on.
Chairman Yarmuth. Probably so. Thank you for that.
And I yield myself 10 minutes for my questions.
Once again, thanks to all the witnesses. It has been an
extremely enlightening conversation and I think a very valuable
one.
I am going to kind of segue off where the Ranking Member
was because we spend most of the hearing talking about impact
on jobs and I think that is kind of the natural topic and how
that might impact tax revenues and so forth, but we really
didn't focus much on how AI might help reduce expenses for the
government.
And I can see--I think Dr. Athey you mentioned telemedicine
and I think there is a lot of potential as you mentioned for
reducing costs in Medicaid, transportation costs, as well as
probably getting better diagnoses and drug interactions and so
forth. I think there is a lot of possibilities there.
Where might be some other areas in which there actually
could be a positive impact of AI on expenses for the
government?
Dr. Athey. Thank you for that question, Chairman, and I
think that really does pick up from the Ranking Member's
comments as well that government can be much more efficient
than it is.
Now, I would have actually been pretty scared 10 or 15
years ago to suggest governments invest more in IT because IT's
projects are--often fail in private sector, frankly, and when
governments take them out, we have a lot of problems with
procurement of large IT projects, but one of the things that I
think has been really impactful in how AI and machine learning
have been diffusing through the economy in the last few years
is the way in which IT services delivers has changed.
We are having more software as a service, we are having
more cloud computing, so that you don't have to say take on
this huge project which has huge risk and then you are kind of
locked into a software for the next 20 years, but rather you
are getting services that meet your needs that are updated
automatically and where a lot of R&D can be centralized and
focused on use cases.
So I do believe that it is a good time to start thinking
about modernizing the federal government infrastructure. And
then alongside of that, in these very common AI applications
like fraud detection was mentioned, also security.
Cybersecurity is a huge problem and, again, because of its
antiquated infrastructure, the federal government and all of
its employees are vulnerable.
And so if we can start modernizing and we can put in best
practices, we can deliver services more efficiently and
effectively.
Now, I also want to pick up on another comment that you
made, which we really didn't talk enough about today, I think,
which is that, you know, when labor is used as an input, then
that is affecting the cost of a product.
Daron and I have both mentioned that there are some cases
where the worker and the machine are sort of creating similar
cost structures, but there are other settings where investments
can really lower the marginal costs of providing services, as
well as the marginal costs of receiving services.
And actually, especially for state and local governments,
that is very true. We--people are standing in line and wasting
their time and taking off of work to get needs met and a person
is sitting behind the counter doing something where all of this
would just be so much faster and better if you could just get--
do it electronically and get your needs met.
And so while that loses a piece of employment for the
worker sitting behind the counter, maybe there is other things
that your government could be doing--more childcare, more elder
care, you know. There are other services that are under
provided where those human workers could be better deployed if
we use technology to do things where the human time is getting
wasted on both sides of the table.
Chairman Yarmuth. I appreciate that.
And I think, even though, we are focused--we are the U.S.
Congress, we are focused on the federal budget, we also need to
think about impacts on state and local budgets. These are all
tax dollars and we do have a federal system.
As I said before the hearing started off the air that this
is something I have been planning to do for about a year and a
half now. And when people would come to my office, different
groups would come to my office when they were still doing that,
I would invariably ask them at the end of the meeting, what
impact artificial intelligence is having on their profession or
their activity?
And I never forget, I had the Kentucky CPAs in the office
and they were there to lobby about tax policy, which is
understandable. At the end of the meeting I asked them, in your
professional meetings, do you ever talk about artificial
intelligence? And their eyes all opened wide and they said,
that is the number one topic at all of our meetings because
they see a dramatic reduction in the need for accountants
because of artificial intelligence.
I had the War College--people from the War College in my
office. I asked them that, and one high-ranking soldier said,
we don't think that there will ever be a battlefield decision
made by a human being again. And I am sure he was exaggerating
somewhat, but the idea was that AI can consider all the
hundreds, if not more, variables that would go into a decision
as to when or where to stage a military action.
And so particularly, you know, I talked to IBM people and
they say Watson, at least in their analysis, can now do 70
percent of what lawyers do with greater accuracy. They can read
MRIs. Watson can read MRIs and CT scans more accurately than
radiologists can.
All of these things--meaning to say, the impact is not
necessarily just going to be on the routine type of jobs; that
there are going to be some very high level jobs that are going
to be changed or eliminated, many cases, which connects me back
to the education issue, that when you look at professional jobs
that require years and years and years of education and
hundreds of thousand dollars' worth of tuition and you are
seeing the possibility that those jobs might be eliminated, how
do you think this is going to change the future of even
professional education?
Dr. West, you want to try that one.
Dr. West. I think you are right. It certainly is not just
entry level jobs that are going to be affected by AI and
automation, but higher level jobs, including the example you
gave of radiologists. Accountants should be worried. They are
exactly right because there is a lot of really good finance AI
that is out there. Financial advisers, the same thing.
So I do think that we need to keep our eye on the education
process. When I talk to young people today, I tell them one of
the most important skills they need to develop is adaptability
because they are going to face such a changing economy, a
changing workforce, and changing job needs and job skill needs
that whatever knowledge and skill they have at age 21 when they
are graduating from college, it is probably going to be
completely inadequate 10 years later. It is certainly going to
be inadequate 20 and 30 years later. So they are going to need
to constantly upgrade their job skills. It is the reason, in my
testimony I talked about lifelong learning.
I think the adult education aspect is going to end up being
as big as higher education today, so the education component is
very important.
Chairman Yarmuth. I am glad you said that. It is exactly
what I tell students when I talk to them, too. You are going to
have to be adaptable. It is the number one talent.
I promised Mr. Woodall, Dr. Acemoglu, that I would let you
answer--he asked about companies, corporations that are doing
the right thing. So in the time I have left, do you want to
expand on those that are good examples for us?
Dr. Acemoglu. Sure. I think--let me say two things.
One is in answer to Congressman Woodall's question, but
before I do that, I want to sort of build on what Dr. West has
said. I think adaptability is extremely important and it is
going to become more important.
But I also think that there is a lot of uncertainty about
which types of jobs AI is going to be more threatening to and
there is some disagreement, but if you look at the current
users of AI, which are still sort of limited, they are still
more geared toward more low-paying jobs.
And part of the reason for that is because when you look at
higher paying jobs, they involve a variety of tasks and only
some subset of those tasks can be automated. And when the rest
aren't, then the adaptable workers especially benefit a lot.
So I expect AI technology in whatever direction it goes to
add to our concerns about inequality. So that I think is a very
important thing.
When it comes to which companies are using AI--I think, you
know, there are many companies in Silicon Valley that are using
AI in extremely creative ways. I think the problem is that some
of those are not very, very good when they look--when you take
their social implications into account.
So, for instance, I think you can use AI as a sort of,
niche industry right now, but there are a couple of companies
that are working on using AI for doing test grading. That is
going to be a growth industry, and I think it is going to be
very, very useful. But there are many fewer of them that are
using AI technology for creating more adaptable classrooms.
There are a few, but I would say that is one area that is
actually very promising, but because of the complexity of the
question, I think one of the concerns is that, when I have
talked to some of those companies, they think that their
technology would not get a hold in many school districts
because it would involve hiring more skilled teachers and
school districts are not going to have the resources or the
interest in doing that.
So I think there is sort of a chicken and egg problem.
There is a lot of creativity that could be put to use AI in
very new and inspiring ways, but we may not have the
infrastructure to support that completely yet.
Thank you.
Chairman Yarmuth. All right. Well, thank you very much, and
my time is expired.
So as we close, let me, once again, thank all of you
witnesses for your time and your wisdom and knowledge, and all
the Members for participating.
And if there is no further business, this meeting is
adjourned.
[Whereupon, at 3:25 p.m., the Committee was adjourned.]
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]