[House Hearing, 116 Congress]
[From the U.S. Government Publishing Office]


                        ARTIFICIAL INTELLIGENCE
                         AND THE FUTURE OF WORK

=======================================================================

                                HEARING

                               BEFORE THE

                SUBCOMMITTEE ON RESEARCH AND TECHNOLOGY

              COMMITTEE ON SCIENCE, SPACE, AND TECHNOLOGY
                        HOUSE OF REPRESENTATIVES

                     ONE HUNDRED SIXTEENTH CONGRESS

                             FIRST SESSION

                               __________

                           SEPTEMBER 24, 2019

                               __________

                           Serial No. 116-48

                               __________

 Printed for the use of the Committee on Science, Space, and Technology

[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]

       Available via the World Wide Web: http://science.house.gov
       
       
                               __________
                               

                    U.S. GOVERNMENT PUBLISHING OFFICE                    
37-740PDF                  WASHINGTON : 2019                     
          
--------------------------------------------------------------------------------------       
       
       

              COMMITTEE ON SCIENCE, SPACE, AND TECHNOLOGY

             HON. EDDIE BERNICE JOHNSON, Texas, Chairwoman
ZOE LOFGREN, California              FRANK D. LUCAS, Oklahoma, 
DANIEL LIPINSKI, Illinois                Ranking Member
SUZANNE BONAMICI, Oregon             MO BROOKS, Alabama
AMI BERA, California,                BILL POSEY, Florida
    Vice Chair                       RANDY WEBER, Texas
CONOR LAMB, Pennsylvania             BRIAN BABIN, Texas
LIZZIE FLETCHER, Texas               ANDY BIGGS, Arizona
HALEY STEVENS, Michigan              ROGER MARSHALL, Kansas
KENDRA HORN, Oklahoma                RALPH NORMAN, South Carolina
MIKIE SHERRILL, New Jersey           MICHAEL CLOUD, Texas
BRAD SHERMAN, California             TROY BALDERSON, Ohio
STEVE COHEN, Tennessee               PETE OLSON, Texas
JERRY McNERNEY, California           ANTHONY GONZALEZ, Ohio
ED PERLMUTTER, Colorado              MICHAEL WALTZ, Florida
PAUL TONKO, New York                 JIM BAIRD, Indiana
BILL FOSTER, Illinois                JAIME HERRERA BEUTLER, Washington
DON BEYER, Virginia                  JENNIFFER GONZALEZ-COLON, Puerto 
CHARLIE CRIST, Florida                   Rico
SEAN CASTEN, Illinois                VACANCY
KATIE HILL, California
BEN McADAMS, Utah
JENNIFER WEXTON, Virginia
                                 ------                                

                Subcommittee on Research and Technology

                HON. HALEY STEVENS, Michigan, Chairwoman
DANIEL LIPINSKI, Illinois            JIM BAIRD, Indiana, Ranking Member
MIKIE SHERRILL, New Jersey           ROGER MARSHALL, Kansas
BRAD SHERMAN, California             TROY BALDERSON, Ohio
PAUL TONKO, New York                 ANTHONY GONZALEZ, Ohio
BEN McADAMS, Utah                    JAIME HERRERA BEUTLER, Washington
STEVE COHEN, Tennessee
BILL FOSTER, Illinois
                        
                        
                        C  O  N  T  E  N  T  S

                           September 24, 2019

                                                                   Page
Hearing Charter..................................................     2

                           Opening Statements

Statement by Representative Haley Stevens, Chairwoman, 
  Subcommittee on Research and Technology, Committee on Science, 
  Space, and Technology, U.S. House of Representatives...........     8
    Written Statement............................................     9

Statement by Representative Jim Baird, Ranking Member, 
  Subcommittee on Research and Technology, Committee on Science, 
  Space, and Technology, U.S. House of Representatives...........    67
    Written Statement............................................    68

Written statement by Representative Eddie Bernice Johnson, 
  Chairwoman, Committee on Science, Space, and Technology, U.S. 
  House of Representatives.......................................    69

                               Witnesses:

Dr. Arthur Lupia, Assistant Director, Directorate for Social, 
  Behavioral and Economic Sciences, National Science Foundation
    Oral Statement...............................................    11
    Written Statement............................................    13

Dr. Erik Brynjolfsson, Schussel Family Professor of Management 
  Science and Director, The MIT Initiative on the Digital 
  Economy, Massachusetts Institute of Technology
    Oral Statement...............................................    20
    Written Statement............................................    22
Ms. Rebekah Kowalski, Vice President, Manufacturing Services, 
  ManpowerGroup
    Oral Statement...............................................    36
    Written Statement............................................    38
Dr. Sue Ellspermann, President, Ivy Tech Community College
    Oral Statement...............................................    55
    Written Statement............................................    57

Discussion.......................................................    69

             Appendix I: Answers to Post-Hearing Questions

Dr. Arthur Lupia, Assistant Director, Directorate for Social, 
  Behavioral and Economic Sciences, National Science Foundation..    84

            Appendix II: Additional Material for the Record

Letter submitted by Representative Haley Stevens, Chairwoman, 
  Subcommittee on Research and Technology, Committee on Science, 
  Space, and Technology, U.S. House of Representatives...........    86

Executive Summary submitted by Representative Haley Stevens, 
  Chairwoman, Subcommittee on Research and Technology, Committee 
  on Science, Space, and Technology, U.S. House of 
  Representatives................................................    88

Statement submitted by Representative Haley Stevens, Chairwoman, 
  Subcommittee on Research and Technology, Committee on Science, 
  Space, and Technology, U.S. House of Representatives...........   100

 
                        ARTIFICIAL INTELLIGENCE
                         AND THE FUTURE OF WORK

                              ----------                              


                      TUESDAY, SEPTEMBER 24, 2019

                  House of Representatives,
           Subcommittee on Research and Technology,
               Committee on Science, Space, and Technology,
                                                   Washington, D.C.

    The Subcommittee met, pursuant to notice, at 4:02 p.m., in 
room 2318 of the Rayburn House Office Building, Hon. Haley 
Stevens [Chairwoman of the Subcommittee] presiding.
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]

    Chairwoman Stevens. This hearing will come to order. 
Without objection, the Chair is authorized to declare recess at 
any time. Good afternoon. Welcome, and thank you to our 
witnesses for joining us here today. We are all looking forward 
to your testimony. Thank you also for your flexibility with the 
later start this afternoon. I'd like to take a moment to offer 
my deepest sympathies to Majority Whip Clyburn on the passing 
of his beloved wife, Emily. My thoughts are with him and his 
family during this time of sorrow.
    We are here today to examine the role of artificial 
intelligence in shaping the work of the future. Recent 
developments in machine learning algorithms, combined with 
increasing computing power and data generation, have enabled 
rapid advances in the accuracy, efficiency, and applicability 
of artificial intelligence (AI) systems. AI systems have 
already begun to change the nature of work and the workforce. 
They are being used in manufacturing processes, medical care, 
and customer service.
    As we talk--and we will talk about this--as we talk about 
job loss that will occur as advanced technology increasingly 
affects all occupations and wage levels, companies in my 
district in southeastern Michigan are also telling me how much 
trouble they are having trying to fill the jobs they have 
available. A 2017 study by the McKinsey Global Institute found 
that approximately half of all work activities could be 
automated by technologies that are already available today, so 
we need to start having the discussion at a broader level about 
how available jobs will transform, rather than disappear, as 
specific tasks are taken over by AI systems, and the workers 
take on new job roles. The advances enabled by artificial 
intelligence also have the potential to create new kinds of 
jobs, and in doing so, elevate the standard of living and 
quality of life for many.
    Sixty-five percent of children entering elementary school 
today, in the year 2019, will ultimately end up working in 
completely new job types that currently do not exist. As the 
integration of these technologies transform work and create new 
jobs, there will be significant need to ensure we are training 
workers to succeed at all levels, from the factory floor worker 
to the radiologist. The key is ensuring that the gains from AI 
systems are shared by all Americans, increasing the quality of 
life for everyone. As we discussed at a hearing in this 
Committee in June, if our Nation leads in the responsible 
development of AI, we can help set the standards and norms the 
rest of the world will follow. That applies equally to the use 
of AI in the workplace.
    We are holding this hearing today to discuss what we do 
know, and also explore what we do not know, and the compelling 
topic of the future of work has certainly compelled many. 
Research studies, companies who are organizing and orienting 
their organizational development, academic institutions, and 
this very body, are compelled to act. As AI-powered robots 
become more common, the question we ask is, how do we ensure 
worker safety alongside these robots? Will artificial 
intelligence be routinely used to monitor workers, as some 
companies do today? How do we balance privacy rights with the 
potential productivity benefits and worker benefits these 
analyses could provide? How do we keep this data secure, and 
prevent its malicious use? And finally, how do we get a better 
understanding of the macroeconomics and labor outlook so that 
the government, companies, colleges, universities, and workers 
can all plan for this transition? It's the question hanging 
above us in this 21st century age. These are just some of the 
questions the researchers are pursuing.
    So I am greatly looking forward to today's hearing, because 
we are compelled to act, to explore, to develop good policy, to 
stand up for the value of work, what knowledge and tools, 
researchers, companies, and workers need going forward, and how 
Federal science agencies, such as the NSF (National Science 
Foundation), are helping to lead the way.
    Before I recognize Dr. Marshall for an opening statement. 
Wait, hold on 1 second. We're pausing on an opening statement. 
OK. Before we move on for opening statements, what I'd like to 
do at this time is to present for the record a letter from 
Kelly Services in support of this hearing, and I would also 
like to submit the executive summary from the 2018 report 
written by the great Mark Muro, and his team from The Brookings 
Institution, titled ``Automation and Artificial Intelligence: 
How Machines Are Affecting People and Places'', a great read 
that's recommended by many.
    [The prepared statement of Chairwoman Stevens follows:]

    Good afternoon, welcome and thank you to our witnesses for 
joining us here today, I'm looking forward to hearing your 
testimony. Thank you for your flexibility with the late start 
today. I'd like to take a moment to offer my deepest sympathies 
to Majority Whip Clyburn on the passing of his wife; my 
thoughts are with him and his family during this time of 
sorrow.
    We are here today to examine the role of artificial 
intelligence in shaping the work of the future. Recent 
developments in machine learning algorithms, combined with 
increasing computing power and data generation, have enabled 
rapid advances in the accuracy, efficiency and applicability of 
artificial intelligence systems.
    AI systems have already begun to change the nature of work 
and the workforce. They are being used in manufacturing 
processes, medical care, and customer service.
    As we talk about the job loss that will occur as advanced 
technology increasingly affects all occupations and wage 
levels, companies in my district are telling me about how much 
trouble they are having trying to fill the jobs they have 
available. A 2017 study by the McKinsey Global Institute found 
that approximately half of all work activities could be 
automated by technologies that are already available today.
    We need to start having the discussion at a broader level 
about how the types of jobs available will change rather than 
disappear, as specific tasks are taken over by AI systems and 
the workers take on new tasks.
    The advances enabled by artificial intelligence also have 
the potential to create new kinds of jobs, and in doing so, 
elevate the standard of living and quality of life for many. 
65% of children entering elementary school today will 
ultimately end up working in completely new job types that 
currently do not exist.
    As the integration of these technologies changes jobs and 
creates new jobs, there will be a significant need to ensure we 
are training workers to succeed at all levels, from the factory 
floor worker to the radiologist. The key is ensuring that the 
gains from AI systems are shared by all Americans, increasing 
the quality of life for everyone. As we discussed at a hearing 
in this Committee in June, if our Nation leads in the 
responsible development of AI, we can help set the standards 
and norms the rest of the world will follow. That applies 
equally to the use of AI in the workplace.
    We are holding this hearing today to discuss what we do 
know, but the fact is there is a lot we still do not know. As 
AI-powered robots become more common, how do we ensure worker 
safety alongside these robots? Will artificial intelligence be 
routinely used to monitor workers, as some companies do today? 
How do we balance privacy rights with the potential 
productivity benefits and worker benefits these analyses could 
provide? How can we keep this data secure and prevent its 
malicious use? And finally, how do we get a better 
understanding of the macroeconomics and labor outlook so that 
the government, companies, colleges and universities, and 
workers can all plan for the transition? These are just some of 
the many questions researchers are pursuing.
    I look forward to hearing from today's distinguished panel 
who will help us understand what we do know now, what knowledge 
and tools researchers, companies, and workers need going 
forward, and how Federal science agencies such as NSF are 
helping to lead the way.

    Chairwoman Stevens. So at this time I would like to 
introduce our witnesses. Our first witness is Dr. Arthur Lupia. 
Dr. Lupia is the Assistant Director of the Directorate for 
Social, Behavioral, and Economic Sciences at the National 
Science Foundation. He also serves as the Hal R. Varian 
Collegiate Professor of Political Science at the University of 
Michigan. Delighted to have you here on behalf of the 
University of Michigan, as well as the NSF, Dr. Lupia, and you 
also serve as the co-Chair of the Office and Science and 
Technology Policy's Subcommittee on Open Science. Dr. Lupia's 
research focuses on processes, principles, and factors that 
guide decisionmaking and learning. He earned his bachelor's 
degree in economics from the University of Rochester, and his 
social science Ph.D. from the California Institute of 
Technology, Caltech.
    Our next witness is Dr. Erik Brynjolfsson. Dr. Brynjolfsson 
is the Schussel Family Professor of Management Science and 
Director of the MIT Initiative on the Digital Economy. His 
research focuses on the effects of information technologies on 
business strategy, productivity and performance, digital 
commerce, and intangible assets. He is the author and co-author 
of several books, including ``The Second Machine Age: Work, 
Progress, and Prosperity in a Time of Brilliant Technologies.'' 
We applaud you for this milestone work that you have published, 
sir. We are delighted to have you here at this hearing, and we 
also note that you received your bachelor's and master's 
degrees in applied mathematics and decision sciences from 
Harvard University, and a Ph.D. from MIT in managerial 
economics.
    Our third witness is Ms. Rebekah Kowalski. Ms. Kowalski is 
the Vice President of Manpower Manufacturing, a role she has 
held since January 2019 throughout her long and remarkable 
career at ManpowerGroup. Her current portfolio focuses on 
developing solutions that help organizations and leaders deal 
with the implications of the shortage of skilled workers, and 
the evolution of roles and skills. She previously led the team 
that worked with MXD, a digital manufacturing institute, to 
identify how roles and skills will evolve as manufacturing 
changes with the increasing introduction of digital 
technologies, a truly profound work of primary research that 
has helped many companies orient and prepare for the future of 
work. Ms. Kowalski received her B.A. in English from the 
University of Wisconsin-Parkside.
    Our final witness, Dr. Sue Ellspermann, is the President of 
Ivy Tech Community College of Indiana. Prior to her role at Ivy 
Tech, Dr. Ellspermann was Indiana's 50th Lieutenant Governor, 
from 2013 to March 2016. As Vice Chair of the Indiana Career 
Council, she led efforts to align the State's education and 
workforce development system to meet the needs of employers, a 
continued focus for her as President of Ivy Tech. She certainly 
focuses on the cross-cutting collaboration that is so needed 
with our training centers and our employers. And Dr. 
Ellspermann earned her bachelor's of science in industrial 
engineering from Purdue University, and her master's of science 
and Ph.D. in industrial engineering from the University of 
Louisville. Absolutely fabulous.
    As our witnesses should know, you will each have 5 minutes 
for your spoken testimony, and your written testimony will be 
included in the record for the hearing. When all of you have 
completed your spoken testimony, we will begin with questions. 
Members will have 5 minutes to question the panel. And at this 
time, Dr. Lupia, we'd like to start with your 5-minute 
testimony. Thank you.

                 TESTIMONY OF DR. ARTHUR LUPIA,

          ASSISTANT DIRECTOR, DIRECTORATE FOR SOCIAL,

               BEHAVIORAL AND ECONOMIC SCIENCES,

                  NATIONAL SCIENCE FOUNDATION

    Dr. Lupia. Thank you. Good afternoon, Chairwoman Stevens, 
Representative Marshall, and Members of the Subcommittee. My 
name is Dr. Arthur Lupia. I am the Assistant Director of the 
Social, Behavioral, and Economic Sciences Directorate at the 
National Science Foundation. It is a pleasure to be with you 
this afternoon to discuss how NSF is helping our fellow 
citizens prepare for the future of work.
    Work is a vital and dynamic element of our society. Work 
powers our offices and our factories. It supports our 
communities, and our Nation. And as we can all see, work is 
changing. We know that AI and related technologies can increase 
national competitiveness by making businesses, governments, and 
social organizations more competitive and more effective. These 
technologies can also create many new careers. If these 
technologies are applied with sufficient foresight, they can 
create new opportunities for workers, and improve quality of 
life for communities across the country.
    How can we achieve a future where technological change 
benefits as many people as possible? At the National Science 
Foundation, we believe that achieving this future requires 
working together. Our Future of Work at the Human Technology 
Frontier Program treats future work, future technology, and 
future workplaces as deeply integrated and intertwined elements 
of our Nation's work-based ecosystem. In NSF's Future of Work 
approach, we collect data on worker experiences to inform 
social and behavioral research on workers and workplaces. This 
research, in turn, can guide technological development. Work 
like this can reveal new ways to empower workers, and increase 
productivity.
    Studying workers, workplaces, and technology together are 
the key to creating benefits that everyone can realize, and 
pioneering research of this kind is already underway. On the 
screen is one of the projects NSF has recently supported. This 
is a human being in an exoskeleton. Today's exoskeletons help 
human beings transport very large objects, and navigate 
impossible situations. But this project is about tomorrow's 
exoskeletons. The device that you see here is not just an 
exoskeleton of the body. It's an exoskeleton of the mind. This 
exoskeleton of tomorrow provides information to the worker 
through an augmented reality system. The system empowers the 
worker to process information, and make better decisions, with 
unprecedented speed. This type of technology is awesome, and 
it'll have impacts far beyond factory floors. Today, for 
example, the Veterans' Administration is one of the Nation's 
leading users of exoskeletons. Tomorrow's exoskeletons will 
open new opportunities for our veterans.
    NSF's Future of Work Program supports this technology by 
incentivizing developers, AI experts, and workplace specialists 
to collaborate. Working together, researchers and developers 
can increase performance, decrease injury, expand access, and 
improve quality of life in ways that just would not be possible 
if any of these groups worked alone. That's what NSF can do. To 
date, NSF's Future of Work Big Idea supports projects in a wide 
range of work contexts, including health care, power grids, 
farming, learning, scientific research, transportation, 
emergency response, and, of course, manufacturing.
    NSF not only supports fundamental research in these areas, 
but also supports efforts to bring these big ideas to market. 
For example, NSF recently unveiled new Future of Work awards 
from its Convergence Accelerator. NSF's Convergence Accelerator 
is designed to fund technology-based partnerships that 
simultaneously advance national priorities and create new 
opportunities for American workers. For example, a project 
based at the University of Michigan is examining how to combine 
research in AI, data science, and industrial psychology to find 
better ways to link workers with innovative new training and 
educational opportunities that will help them not only 
contribute, but thrive, and build amazing careers in their new 
workplaces.
    This is an exciting time for our country, and, like you, 
NSF is grateful to see our Nation's brightest minds 
collaborating on the fundamental research that will transform 
our workplaces, empower our workforce, and provide tremendous 
new sources of innovation for our Nation. So thank you for 
having this hearing today, and for the opportunity to testify. 
I'm happy to answer any questions you may have.
    [The prepared statement of Dr. Lupia follows:]
    [GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
    
    Chairwoman Stevens. Dr. Brynjolfsson? Yes.

               TESTIMONY OF DR. ERIK BRYNJOLFSSON,

            SCHUSSEL FAMILY PROFESSOR OF MANAGEMENT

            SCIENCE AND DIRECTOR, THE MIT INITIATIVE

                    ON THE DIGITAL ECONOMY,

             MASSACHUSETTS INSTITUTE OF TECHNOLOGY

    Dr. Brynjolfsson. Good afternoon, Chairwoman Stevens, 
Representative Marshall, and Members of the Committee. Thank 
you so much for inviting me to share some of the research my 
team and I have been doing. Addressing the opportunities 
created by AI is one of the most important challenges for 
government in the coming decade. Thanks to AI, some weird and 
wonderful things are beginning to happen. Cars are learning to 
drive themselves. Machines can now recognize your friends' 
faces. When you see people walking down the street talking on 
their phones, you don't know if they're talking to another 
human or to a machine, and expecting the machine to answer. 
Just last week Siri tried to join into a conversation I was 
having about interest rates.
    However, it's also critical to understand that we are very 
far from what we call artificial general intelligence, the kind 
of AI that spans the full range of human intelligence. While 
machine learning is now superhuman in many tasks that involve 
mapping a particular set of inputs into outputs, humans 
outperform machines in most other cognitive tasks. Therefore, 
we are not facing the imminent end of work, but we are facing a 
major restructuring of work. In research that I've been doing 
with my colleagues, we find that few, if any, occupations will 
be fully automated by this new wave of technologies, but at the 
same time, few, if any, will be unaffected. Instead, most will 
be transformed. For instance, the job of a typical radiologist 
consists of 27 distinct tasks. While machine learning has made 
impressive advances in some of them, like reading medical 
images, it is of little use in most of the other tasks, like 
counseling patients.
    So massive unemployment is not the challenge of our era. 
Instead, we face challenges in two other areas. One is 
delivering productivity growth, and the other is reducing 
inequality. To date, despite impressive improvements in AI, 
productivity growth has actually slowed down. Between 1995 and 
2004 it averaged 2.8 percent per year, but since 2005 
productivity has been just 1.3 percent per year. That's less 
than half the growth rate previously. So why is that? Well, the 
bottleneck is actually not the technology, but rather the lack 
of complementary process innovations, workforce reskilling, and 
business dynamism.
    The second challenge is inequality. There's no economic law 
that says that everyone will benefit from technological 
advances. As the economic pie grows, it's possible for some 
people to be left behind, even as others benefit 
disproportionately. Indeed, over the past several decades the 
benefits of economic growth have been very unequal. Not only 
has the median income barely grown since the late 1990s, but 
other social indicators have actually worsened. For the first 
time in history, average life expectancy of Americans has begun 
to fall, driven by worse mortality of less educated Americans. 
It's no coincidence that these are exactly the Americans who 
haven't shared in our economic growth.
    So my policy recommendations can be grouped into five key 
areas. The first one is to reinvent education. We need to 
recommit ourselves to investment in education. It's a field 
that the U.S. has once led the world. We also need to reinvent 
it so that we focus more on the kinds of skills that machines 
cannot match. These include creativity and interpersonal 
skills.
    Second, we need to rebalance capital and labor. As noted in 
a recent research report by the MIT Work of the Future 
Initiative, our tax code and other policies are heavily skewed 
toward helping capital, rather than labor. We need a more-level 
playing field, particularly as AI starts to affect more and 
more of the labor force. This means taxing capital at 
comparable rates, encouraging investments in human capital, 
just as we do for physical capital, and updating corporate 
governance to recognize workers as stakeholders alongside 
stockholders. We can also expand the Earned Income Tax Credit 
to boost incomes for the working poor, and use revenues from 
things like carbon taxes to lower taxes on work.
    Third, we need to invest in U.S. technological leadership. 
U.S. leadership in AI and other technologies is at serious risk 
because we have cut Federal investment in R&D (research and 
development), even as other nations have boosted theirs. 
Federal science agencies, like the NSF, working with our 
leading universities and private industry, have a central role 
in maintaining and extending America's science and technology 
leadership in AI and other areas.
    Fourth, we need to welcome high skill immigrants. A vastly 
disproportionate number of America's leaders in science and 
business are immigrants, or the children of immigrants. When I 
ask my students at MIT what was the most important message I 
should convey to you here in Washington regarding AI policy, 
they unanimously advised me to push for less restrictive 
immigration policies.
    And fifth, we need to work hard to support 
entrepreneurship. Boosting entrepreneurship can help reverse 
the stagnation of wages for the bottom half of the income 
distribution, particularly those who have been most adverse 
affected by automation. Among the policies that can help with 
this is decoupling healthcare from employment, reforming 
occupational licensing, and direct investments in teaching and 
entrepreneurship, and boosting new business formation.
    With the right policies, AI can be harnessed to make the 
next decade the best decade in U.S. history.
    [The prepared statement of Dr. Brynjolfsson follows:]
    [GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
    
                 TESTIMONY OF REBEKAH KOWALSKI,

            VICE PRESIDENT, MANUFACTURING SERVICES,

                         MANPOWERGROUP

    Ms. Kowalski. Chairwoman Stevens, Ranking Member Dr. Baird, 
and Representative Marshall, on behalf of ManpowerGroup, thanks 
for the invitation to speak today on such an incredibly 
important topic. ManpowerGroup is the world leader in 
innovative workforce solutions. Every day we connect more than 
600,000 people to work around the world in a wide range of 
skills and industries. One of our most predominant industry 
sectors is the manufacturing sector, and I oversee our 
manufacturing solutions practice. I've worked with a lot of 
companies as they are struggling to deal with the twin 
challenge of finding enough rightly skilled talent, and 
figuring out how they're going to navigate the bright new 
future that digital offers.
    Manufacturers are reporting talent shortages as they 
struggle to find the right blend of technical and soft skills. 
Our perspective is that AI, machine learning, and other digital 
technologies produce new jobs that require new skills. Some of 
those we can't even imagine yet. Our research shows that over 
90 percent of employers expect to be impacted by digitization 
over the next 2 years. Eighty-seven percent of them plan to 
maintain or increase head count. Four percent say they don't 
know. And yes, there is a small number, 9 percent of them, that 
say that they anticipate a reduction. Fully 75 percent say this 
is going to require new skills, skills that we do not currently 
have in our workforce, and skills that we can't actually even 
anticipate.
    In 2017 we released a study with MXD, which was formerly 
known as the Digital, Manufacturing, and Design Innovation 
Institute, on how digital technologies, including AI and 
machine learning, would impact manufacturing jobs. The study 
was accomplished in partnership with academia and industry, and 
identified 165 new or significantly evolved roles. Today the 
majority of manufacturing roles are in the general entry level 
population, by count. That is--those are roles like picker/
packer, assembler, operator, helper, laborer. And the 
manufacturing sector, the backdrop here, is that we are going 
to produce 3.5 million new jobs over the next decade, while at 
the same time 2.7 manufacturing workers are set to retire. Many 
of the new jobs will be in these more specialized areas, like 
technicians, testers, analysts, specialists, and that's a 
significant shift for us.
    We have the following concerns. First, employers are 
uncertain about how digitization will impact roles and skills, 
and over what period of time. Second, the ability of employers 
of all sizes to invest in upscaling falls far short of what is 
required to produce the workforce they need, both from a time 
and resource perspective. Third, the talent shortage impacts 
all types of talent, from entry level to leadership, meaning 
employers have to determine the best way to allocate precious 
dollars. That disproportionately impacts small and mid-sized 
manufacturers.
    There are several obstacles to being resourceful around 
talent attraction and upscaling. One, it's difficult for 
organizations to predict workforce needs more than a year in 
advance. Strategic workforce planning does not have as long of 
a horizon as it needs. Without enough exact match talent, we 
need to shift to hiring on potential and learnability, but H.R. 
(human resources) systems and processes are still geared toward 
finding an exact match. Third, job descriptions need to be less 
stationary, and more evolutionary, so that individuals can 
actually anticipate the need for ongoing learning and 
adaptation. And four, organizations lack sufficient funding for 
workforce training.
    An example of improved training processes is what we do 
with Rockwell Automation in our Academy of Advanced 
Manufacturing, where we take veterans and we put them through a 
12-week embedded program, and we graduate them as Certified 
Automation Technicians. They walk away with a job that, on 
average, is double what they were making when they came in, and 
the employer walks away with the talent that they need. With 12 
million manufacturing workers in the U.S., we need those kinds 
of nimble programs, many, many more of them, in order to ensure 
that people have a path to sustainable prosperity, and we need 
to start now. Don't count the humans out.
    Talent is, in fact, the most renewable resource we have on 
the planet. It is ready to learn, adapt, and thrive in new 
environments, and we need to work collectively now across 
educators, employers, and individuals to become proactive 
builders of talent to develop a workforce with the skills 
employers and individuals need to remain competitive, both now 
and in the future.
    Thank you again to the Subcommittee for the opportunity to 
share my testimony.
    [The prepared statement of Ms. Kowalski follows:]
    [GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
    
                TESTIMONY OF DR. SUE ELLSPERMANN,

             PRESIDENT, IVY TECH COMMUNITY COLLEGE

    Dr. Ellspermann. Thank you, Chairwoman Stevens, Ranking 
Member Baird, welcome, and Representative Marshall. It's really 
a privilege to be here representing community colleges today, 
and Ivy Tech Community College specifically, as we talk through 
machine learning, artificial intelligence, and particularly how 
that's affecting community colleges, and how we're working with 
industry and businesses to establish an ecosystem to address 
the changing demands. I also will speak at the end about what 
the Federal Government could do to assist in this work.
    So remember that community colleges are the most common 
type of U.S. college, with Ivy Tech being one of those, 
established in 1963 as a vocational/technical college, now the 
largest in the Nation Statewide system, singly accredited, with 
150,000 students and 18 campuses, 40 locations. But think about 
our student, who is now--the traditional student is that 
community college-like student, who is part time. Average age 
is probably 27 years old, Pell eligible, and a quarter of those 
students have dependents, children, that is, and you can see 
more in our report.
    But how will that impact us as we look at AI and machine 
learning? And what you heard from several of my colleagues here 
is that there will be some displacement, but with that 
displacement will become very good opportunities, and it's up 
to our community colleges to prepare those students, those 
employees, for the wide spectrum of industries and 
opportunities that are out there. So let me talk about just a 
few of the very concrete things that we've done, and I thank 
Ms. Kowalski for sharing some of those as well in the 
manufacturing space, but one that I'm sure she'd be interested 
in is the partnership that Ivy Tech's done with the Smart 
Automation Certification Alliance as they've developed the 
first certifications in industry 4.0, which we know will be 
factories of the future, and the kind of credentials we'll need 
in that very connected manufacturing environment.
    But at the community college level, we work with many 
partners, for instance, Sales Force, through their Pathfinders 
Program to earn Sales Force developer and administrator 
certifications. We have many certificates in informatics and 
software development at the Associate level. We work with Apple 
in their iOS systems applications. We work with Cisco, as they 
overhaul their certifications, to embed those right into our IT 
programs. With Amazon Web Services, we are developing cloud 
computing certificates, and soon to be an Applied Associate in 
Cloud Computing. All of those are staying with those industries 
and particular businesses to make sure that we're providing our 
students with the kind of skills that they will need.
    I'm going to speak to a partnership we have with industry, 
particularly our Achieve Your Degree Program, which is a 
redesign of the tuition reimbursement program, where industries 
actually pay for, at the end of that cycle, the tuition that 
that employee of theirs pursues, but we, concierge-style, come 
to the industry, that business, to enroll, to do financial aid 
eligibility, and then to ensure that the programs align with 
what the business has. In doing that, we've had a great 
partnership with our Indiana Chamber of Commerce, Statewide, 
more than 200 companies doing that. I'll just share one, with 
Cook Group in Bloomington, Indiana, where 500 of their 
employees are being skilled up, have already earned 100 
credentials in the last 3 years.
    Now, in design, we put everything, from an economist 
standpoint, into quadrants to make sure that the highest demand 
areas with the smallest supply of employees are being built up 
into those particular quadrants. We'll describe those quadrants 
more in our full report, but in doing that, we make sure that 
we are putting our focused energy in the high-demand areas, 
that we're shrinking problems that need to shrink, and that we 
are seeking equilibrium in this highly changing environment. 
And it's working. In IT we, just last year, increased our 
completions by 75 percent in a single year, and we see that 
across our programs.
    I'm going to spend my last moments talking about what we 
could do with some Federal support. You know, employers hate to 
have to pay Unemployment Insurance (UI) into that trust fund. 
Several years ago, most of our States were in a deficit. We 
were in Indiana. Congressman Baird remembers that. Today we are 
at $900 million in the black. Those funds could be deployed 
toward this work rescaling earlier than when that person is 
displaced, but when you decide on that technology, and we're 
hopeful that there will be some willingness of this Congress to 
look at making that available to a State and a community 
college system to experiment with how we could deploy a portion 
of those UI funds in these ways. We look for all kinds of 
support in reducing regulation so that we can change at the 
speed of the technologies that we're working with to ensure 
that all of our workers have those opportunities. And with 
that, I'll just thank you for the opportunity for appearing 
before the Subcommittee, and the opportunity to share the work 
of Ivy Tech Community College.
    [The prepared statement of Dr. Ellspermann follows:]
    [GRAPHICS NOT AVAILABLE IN TIFF FORMAT]
    
    Chairwoman Stevens. Well, thank you all, and at this time 
the Chair would like to recognize Ranking Member Dr. Baird for 
his opening remarks. Thank you.
    Mr. Baird. Thank you, Madam Chairwoman. I apologize for 
being late, but I do admire you for going on without me. Thank 
you.
    Chairwoman Stevens. We're a team, you know.
    Mr. Baird. So I appreciate this opportunity. I appreciate 
you waiting on me, and I'd like to make this opening statement, 
and thank you for holding this ``Artificial Intelligence and 
the Future of Work'' Committee hearing. Since the term AI was 
introduced in the 1950s, we have made some huge advances in the 
field, and thanks to critical investments by government and 
industry, universities and the United States, in leading global 
AI research and development.
    Today AI systems have been deployed in every sector of the 
U.S. economy. These technologies have already delivered 
significant benefits for the U.S. economic prosperity, for the 
environmental stewardship, and the national security. AI has 
long been a subject of interest of the House Science Committee, 
and we have held several important and productive hearings on 
this topic. In the past we have discussed how to define AI, the 
science of AI technologies, and the needs for standards to 
address ethics and potential bias. Now, this afternoon, we will 
examine AI from the prospective of the American worker.
    In order to remain a leader in AI, I believe we must 
prepare our workforce for the next generation of opportunities 
in this technology, and for our future, defined by a lifelong 
learning experience. In order to grow our economy, I also 
believe we must acknowledge and understand how AI is changing, 
and will continue to change, the jobs and lives of hard-working 
Americans. This is a large scale effort that is going to 
require cooperation between industry that was already mentioned 
here, industry, academia, and the Federal agencies, so I'm 
pleased to see that the Trump Administration is making this 
issue a priority, and recently established the National Science 
Council for the American Worker and the American Workforce 
Policy Advisory Board. American industry has responded well to 
the Administration's initiatives. Over 300 companies and 
organizations have pledged to study and expand education, 
training, re-skilling opportunities for American workers to 
gain AI-relevant skills.
    We also need to re-think how we educate future workers, and 
re-skill the workers of today, all the way from K through 12 
schools to the community colleges, the vocational schools, and 
the 4-year universities. Some leaders in the U.S. education 
system are already finding innovative ways to develop a highly 
skilled AI workforce, one of the future. We have heard about 
some of those efforts from my friend, Dr. Sue Ellspermann, 
President of the Ivy Tech Community College system in our home 
State of Indiana. Sue, so glad to have you here today. At Ivy 
Tech, Dr. Ellspermann works to address the changing demands of 
employers in the Hoosier State by providing strategic support 
and career planning for students at community colleges and 
vocational schools, and working closely with local industry. I 
look forward to hearing more about her important work in our 
community, and how it will be applied across the country.
    Over the next few months, this Committee will be working 
toward bipartisan legislation to support a national strategy on 
artificial intelligence. The challenges we must address are how 
industry, academia, and the government can work together on AI 
challenges, including today's critical workforce questions, and 
what role the Federal Government should play in supporting 
industry as it drives innovation. I want to thank our 
accomplished panel of witnesses for their testimony today, and 
I appreciate the opportunity to hear how this Committee and the 
Federal Government can support innovation and education to 
ensure a bright future for America's workers, our students, and 
maintain our leadership in AI. So thank you.
    [The prepared statement of Mr. Baird follows:]

    Chairwoman Stevens, thank you for holding today's hearing 
on ``Artificial intelligence (AI) and the Future of Work.''
    Since the term AI was first coined in the 1950s, we have 
made huge advances in the field. And thanks to critical 
investments by government, industry, and universities, the 
United States is leading in global AI Research & Development.
    Today, AI systems have been deployed in every sector of the 
U.S. economy. These technologies have already delivered 
significant benefits for U.S. economic prosperity, 
environmental stewardship, and national security.
    AI has long been a subject of interest for the House 
Science Committee and we have held several important and 
productive hearings on this topic.
    In the past, we have discussed how to define AI, the 
science of AI technologies, and the needs for standards to 
address ethics and potential bias.
    Now, this afternoon, we will examine AI from the 
perspective of the American worker.
    In order to remain a leader in AI, I believe we must 
prepare our workforce for next generation opportunities in this 
technology and for a future defined by lifelong learning.
    In order to grow our economy, I also believe we must 
acknowledge and understand how AI is changing and will continue 
to change the jobs and lives of hard-working Americans.
    This is a large-scale effort that is going to require 
cooperation between industry, academia and federal agencies.
    So I am pleased to see that The Trump Administration is 
making this issue a priority and recently established the 
National Council for the American Worker and the American 
Workforce Policy Advisory Board.
    American industry has responded well to the 
Administration's initiatives. Over 300 companies and 
organizations have pledged to study and expand education, 
training, and reskilling opportunities for American workers to 
gain AI-relevant skills.
    We also need to rethink how we educate future workers and 
reskill the works of today, all the way from K-12 schools to 
community colleges and vocational schools, to 4-year 
universities.
    Some leaders in the U.S. education system are already 
finding innovative ways to develop a highly-skilled AI 
workforce of the future.
    We will learn more about some of those efforts from one of 
our witnesses today, my good friend, Dr. Sue Ellspermann, 
President of the Ivy Tech Community College system in our home 
state of Indiana.
    At Ivy Tech, Dr. Ellspermann works to address the changing 
demands of employers in the Hoosier State by providing 
strategic support and career planning for students at community 
colleges and vocational schools and working closely with local 
industry.
    I look forward to hearing more about her important work in 
our community, and how it can be applied across the country.
    Over the next few months, this Committee will be working 
towards bipartisan legislation to support a national strategy 
on Artificial Intelligence.
    The challenges we must address are how industry, academia, 
and the government can work together on AI challenges, 
including today's critical workforce questions, and what role 
the federal government should play in supporting industry as it 
drives innovation.
    I want to thank our accomplished panel of witnesses for 
their testimony today.
    I look forward to hearing how we can support innovation and 
education, to ensure a bright future for America's workers and 
students and maintain our leadership in AI.

    Chairman Stevens. If there are Members who wish to submit 
additional opening statements, your statements will be added to 
the record at this point.
    [The prepared statement of Chairwoman Johnson follows:]

    Thank you, Chairwoman Stevens and Ranking Member Baird, for 
holding this hearing. I would also like to welcome this 
esteemed panel of witnesses and thank each of you for 
accommodating the rescheduling of today's hearing. We are here 
today to discuss an urgent challenge facing the country. 
Artificial intelligence is a rapidly advancing, sophisticated 
technology that promises to transform the way we live and work.
    As Chairwoman, I take seriously the responsibility 
entrusted to this Committee to support the nation's research 
and innovation enterprise for the benefit of society. We are 
increasingly feeling pressure from our global competitors, 
particularly in the case of AI. As countries like the United 
Kingdom, Germany, and China invest heavily in this technology, 
there is a strong sense of urgency to race headlong toward 
technological maturity and widespread adoption.
    I want to urge caution. We must take the time to draw upon 
lessons learned from past technological disruptions, assess the 
opportunities and potential risks, and implement a coordinated 
national strategy to ensure the benefits of AI are enjoyed by 
everyone. We are here to explore one of the primary concerns 
associated with AI - its potential impact on the workforce. 
Many Americans are understandably worried that AI-driven 
automation and robots will make their jobs obsolete.
    Research has a critical role to play in informing how AI is 
integrated into the American workforce. Research can help 
employers understand the benefits and risks of this technology. 
Just because it seems like a task can be performed by an AI 
system, does not mean it can or should be, at least not without 
a human still in the loop. Research can also improve our 
understanding of the human-technology relationship. This can 
inform decisions regarding how best to integrate AI into the 
workflow so it can both complement and enhance the value of the 
worker. Research can advance the development of effective 
practices for retraining the current workforce and for ensuring 
workers have the flexibility to be lifelong learners. Research 
can provide students and those pursuing a career change with a 
clear understanding of emerging industries and occupations, so 
they can chart an education path best suited to their goals.
    Artificial intelligence holds immense promise to spur 
economic growth and make our lives easier. We are at a critical 
point in the development of this technology, and we must ensure 
we have the research knowledge base necessary to maximize these 
benefits for everyone.
    I look forward to today's testimony and discussion and I 
yield back.

    Chairwoman Stevens. Fabulous. At this time we're going to 
begin the 5-minutes of questioning, and the Chair will 
recognize herself for 5 minutes.
    Dr. Lupia, in your testimony, you discuss a recent award 
made to the University of Michigan to support research on how 
humans and robots are working together in construction 
environments, and you stated that, despite recent advances in 
robot functionality, many fundamental questions in robot 
interaction remain unanswered. Do you mind elaborating on that 
a little bit further, and also, could you touch on some of the 
major social science research questions regarding human/robot 
interaction, and where we need to go from here?
    Dr. Lupia. Thank you for that question. As discussed in the 
opening statements, there are things right now that AI and 
robots can do that humans can't do, but there are many things 
that humans can do that robots can't do. And when we're 
thinking about the workplace of the future, particularly its 
impact on workers and workplaces, you know, there are these 
fundamental questions about what the two groups know now, and 
what can we expect them to know in the future, to empower 
workers.
    So I think about farms, for example, right? So I grew up on 
a farm, and so, when I was a kid, people milked cows. And, if 
you've ever done that, it's not the most fun thing. But now 
they have robots that can milk cows, so if you think about 
how--just--a farm that's pretty simple, there are things that 
people can do that robots can't do, and things that robots can 
do that people can't do. And so, through a number of grants, 
we're trying to help factories, farmers, offices, and so forth 
think through, ``How do you make workplaces more efficient?'' 
``How do you make them more effective, with this set of 
evolving skills?''
    Some of it requires trust, right? So if we're going to 
automate a manufacturing process, the worker has to trust the 
robot, or the machine. And trust is a great thing, unless the 
robot's about to do the wrong thing. And so you've always got 
to have an override capacity. What we're trying to do at NSF is 
bring large groups of people together to understand, at a 
pretty fundamental level, when is the trust relationship going 
to work, when is it going to fail, and as robots get better at 
things, how does that change how we should organize the 
workplace? So that's the fundamental question.
    It takes understanding humans, because if you press the 
override button at the wrong time, you can disrupt the process. 
If you wait too long, unintended consequences can happen. So 
understanding the human/robot interaction is really critical to 
all of the progress we want, from manufacturing, to farms, to 
offices of the future.
    Chairwoman Stevens. To be successful. And, Ms. Kowalski, 
you might have given us the line of the day, which is don't 
count the humans out. And you also, in your testimony, 
discussed the rapid change in skills being sought by employers. 
And, you know, in terms of how we think about job descriptions 
to account for this rapidly changing marketplace for skills, 
and also promote a mindset of supportive, continuous skill 
development, how do we do it all? How do we bring that 
together?
    Ms. Kowalski. So I think there's a few things. One is just 
determining that this is what we have to do, right? It's a 
decision that we have to make, that we cannot allow the 
workforce to stay still, that there is no grassy plateau on 
which we'll all be able to stretch out when transformation is 
done. It will be an unending climb, and evolution and 
adaptation, which we're very good at as human beings, right? 
But the way that we approached education and employment was we 
educated to the job. People came into an exact match 
environment, and then they made progressions up the ladder 
based on merit. We haven't seen something come in that acts so 
rapidly.
    Think about automation, and the--it was about 15 years 
playing out in the last cycle. We're talking about something 
that's going to play out, by this research, in 3 to 5 years 
that's unfolding now. And it will get faster, and the peaks and 
troughs will get steeper, and so how we get people attenuated 
to that shift, that starts all the way back in K-12, and moves 
all the way through--and in employment. And the hardest thing 
is going to be taking the people that are currently employed 
and helping them understand they haven't done anything wrong. 
They are hardworking, they've been doing a great job, and these 
are the new set of skills that they have to assimilate, and 
there has to be a new contract, right? And that contract is one 
of you put in for continuous adaptation and evolution, we'll be 
right there to meet you with the resources.
    You know when we were good at doing that? Was in the 1950s 
and 1960s, when we hired on potential. We built whole companies 
hiring on potential for jobs we didn't even know what they were 
going to look like, and people got used to making the 
progression, and having a partnership with employment and 
educators in order to do that. And it's going to take a system 
to do it.
    Chairwoman Stevens. Well, I am just at time, so--you can 
tell we're in a rich topic area. So I'm going to yield back the 
remainder of my time, and I am going to recognize Dr. Baird for 
5 minutes of questioning. You've got this, Dr. Baird.
    Mr. Baird. Thank you, Madam Chair. And, Dr. Ellspermann, 
I'm sure that you recognized I was probably going to start with 
you.
    Dr. Ellspermann. Thank you.
    Mr. Baird. The thing that, and I know you spent a lot of 
time in this area, and thinking about it, but the needs of the 
industry today, compared to the future, and this technology 
that we're discussing today, is changing so fast because of 
quantum computing, and that sort of thing. So I guess, in other 
words, how do you feel, or how do you see Ivy Tech balancing 
that need for today, and then in the future? Kind of give us 
some feel what you think that might look----
    Dr. Ellspermann. Very good. So, actually, 3 years ago, the 
General Assembly in Indiana understood how important it would 
be that we would be work-forced aligned as a system, and 
actually required that, in addition to having a provost I have 
a Chief Workforce Officer, which makes sure this alignment 
happened.
    So I alluded in my comments to this way that we classify 
all of our programs, because we know it's a moving target, we 
know that there is BLS (Bureau of Labor Statistics), and Emsi, 
and other good data out there, economically, to project the 
future. We know that broadly, but it's not accurate at the 
local level, so we take that, and we let the local industry 
work with us to look at what's coming, what is the real demand, 
and then we size our programs on every campus, every program, 
to be that right size.
    And that's where our quadrants, that quadrant one of--
quadrant one is where we focus. It is those high-demand, low-
supply, not enough students to fill that work, and making sure 
we're building those programs. We have limited enrollment 
programs that we have to push on. We have programs that have to 
shrink so that they are the right size, or maybe discontinue, 
and then finally equilibrium. That work is working. It is 
working across our State. We can take the local data to 
understand that maybe the economic data is not quite accurate 
to what the local needs are, and we could shore up, and we 
could shrink, and we do that in a very rapid way.
    What becomes challenging is the support at the Federal 
level to get those kind of programs, when you need new programs 
stood up, to quickly stand those up in 3 to 6 months so that an 
employer gets the kind of skill set that they need. And so we 
look to any support we can get with our U.S. DOE (Department of 
Education) to quickly approve programs. But, as I shared, 
looking at new ways to anticipate, when we know these changes 
are coming, and we know there's a higher demand, how do we 
identify that employee at risk early, even if it is just a 
year, or a year and a half in advance? We can then begin 
skilling before that individual is out of a job, unemployed, 
which is, for many, much more than just about being out of a 
job. It is psychological impact. It is a feeling that a trust 
has been broken with that employer, so how do we proactively 
work with them?
    And whether we use Unemployment Insurance as a part of that 
trigger, we need to change that mindset to create that contract 
again between employer and employee. And I believe the 
community colleges, at least Ivy Tech, is working very hard to 
get there, but I know we will be the front lines for most of 
employers as they look to scale up their employees, and it's 
our job to be as rapid as we can.
    Mr. Baird. Thank you. Dr. Lupia, I'm glad to hear you came 
from a farm, and I couldn't help but say--several of you 
mentioned the importance of the human factor. I think you 
mentioned that. But I just want you to know that those old cows 
have a vested interest in how well these--but anyway, I thought 
maybe you might want to elaborate--I've been fascinated by 
NSF's convergence accelerators since Director Cordova spoke 
about them in this Committee in May. Would you mind elaborating 
on how this new approach to research will improve our 
understanding of the future of work, and enhance the lives of 
American workers?
    Dr. Lupia. Absolutely, sir. Thank you for asking that 
question. The convergence accelerators really build on the 
traditional NSF approach. So in NSF, we fund all of science, 
but the idea with the Convergence Accelerators is, from the 
beginning, you bring in other partners, people in the room, 
who, if great ideas emerge, they can bring them to market. So 
the Convergence Accelerators have really been an exciting way 
to think about how to take amazing collaborations and bring 
them to market.
    So I'll give you one example, because we just started 
funding these things. One has to do with re-skilling the 
workforce, and it is funded, coincidentally, at Purdue 
University, and it focuses on apprenticeships. So both of my 
grandparents were in the trades, and the way that you've 
learned a trade for 100 years is through an apprenticeship. So 
that takes a number of months, and you follow somebody around, 
and you learn the trade. But for a small company it's really 
expensive to take one of your workers and have them do an 
apprenticeship for 6 months. So one of the Convergence 
Accelerators is a project built around using technology to do 
apprenticeships at scale.
    So imagine we could take what a master plumber knows, or a 
master technician, or someone who runs a computer, and we can 
follow them around, and then create scalable, low-cost ways to 
distribute this information to everybody. And one of the ways 
you can do that is through having simulations. So instead of 
one person shadowing the expert, you can build a simulation 
where 50 people can have a virtual reality experience of 
shadowing the expert. And so it has a lot of the benefits of 
the traditional apprenticeship, and of course you still need 
the one-to-one contact, but this is a way to really make that 
happen at scale. And, again, if you're a small company, if you 
can go to a community college, or somewhere else, and get this 
type of training, it's a real game changer.
    So the idea here is apprenticeship, lower costs, improve 
speed and reliability, minimize errors, and this is something 
that the Converge Accelerator, I think, can really do to help 
companies across the country.
    Mr. Baird. Thank you.
    Chairwoman Stevens. At this time the Chair would like to 
recognize Dr. Marshall for 5 minutes of questioning.
    Mr. Marshall. All right. Thank you, Chairwoman. Dr. 
Ellspermann, I'm a community college graduate, my wife, 
community college graduate, huge fans of my community colleges. 
The technical colleges can quickly pivot to the job needs of my 
community, and I think that's where the rubber meets the road. 
How do you measure success? What are you measuring to say, 
we're being successful in our technical college that you run?
    Dr. Ellspermann. We measure success the way most Americans 
do, by wages. We actually measure the wages of our graduates 1 
year out to see what they're making. Our goal is that 80 
percent of all of our graduates will make above median wage 1 
year after completion. We're at 45 percent today, we were at 38 
percent 2 years ago, and we're marching our way--but we think 
that is one fair way to do that. In addition, too, we also hold 
ourselves accountable to those four quadrants. We want 80 
percent of our programs to be in equilibrium, meaning we're 
roughly producing the number of graduates needed for our 
community.
    Mr. Marshall. And are you measuring their debt when they're 
leaving too?
    Dr. Ellspermann. We have minimal, you know, community 
college debt is kind of the best kind of debt. It's, like, 
under $10,000. It's the way to do college. But we do measure 
debt, and we do measure what those students have, and always 
are looking for ways to continue to reduce that.
    Mr. Marshall. I'm not sure how long you've been at Ivy 
Tech, but what are you doing differently today than 1 year ago, 
than 3 years ago, or 5 years ago?
    Dr. Ellspermann. We have reinvented how we're delivering. 
So we've gone from traditional 16-week courses to 8-week 
courses because, guess what, adults do better in that format. 
There's higher pass rates, lower drop rates. We've just 
redesigned our online education. We're one of the largest 
online educators in the country. We know we have to do that 
better because, guess what, single moms need to be able to take 
courses online, and they have to be as good as the face-to-face 
delivery. So in that redesign, we are looking at all of the way 
we do our work to align better to industry, and to deliver in 
the best way for our students. And there's much more to do, 
Congressman.
    Mr. Marshall. So certainly, as an obstetrician, you're 
hitting on exactly who I'm thinking of, that single mom who 
maybe could get her auntie, or her sister, to come in and help 
with the kids for 6 weeks or 8 weeks, but it's hard to get them 
to commit to 18 weeks. One of the things that we're certainly 
looking at is using Pell Grants in a non-traditional situation, 
what you're describing. Hopefully we can make some progress 
there at some point in time.
    Dr. Ellspermann. Thank you.
    Mr. Marshall. So we have the NSF person here, Dr. Lupia, as 
well. What would you tell him? How could NSF work better with 
community colleges and technical colleges? What ideas out there 
are outside the box that you wish we could get better engaged 
with NSF?
    Dr. Ellspermann. I would say certainly in helping us to 
adopt that technology early. We are not funded at the levels of 
research institutions, as you might guess, so keeping our labs 
up to date with that front-edge technology at the same time 
industry's getting it, not a generation later. We really need 
to have it early. Certainly Perkins helps on that front, but 
that cycle of rapid change is so much quicker than it was 
generations ago that we have to be able to refresh our 
equipment every year, two, or three, which there's probably a 
partnership to be built there.
    Mr. Marshall. OK. Dr. Lupia, any return thoughts, or 
comments?
    Dr. Lupia. We are so grateful for the work that your 
organization does, and part of our Future of Work Project is 
really to try and make this information and these 
collaborations happen a lot earlier. So the scientific approach 
is, there's a relationship between jobs and skills. Most jobs 
take a whole bunch of skills, and as jobs evolve, some of the 
skills that we have now will still be relevant, but there will 
be these other new skills that you can use.
    So we are working with government, industry, and a whole 
range of researchers to try and project, ``How are skills and 
jobs likely to evolve?'' If we can figure that out, and put 
that into data bases, and match it to jobs as they're evolving, 
then our partners can make that data available to everyone--
because that's the idea, right? We have projects in several 
States--Georgia, West Virginia now--where we're collecting data 
from them, and then trying to push out real-time and usable 
data about how jobs are likely to change. This can produce 
really great efficiencies, because now we can tell community 
colleges and others these are the skills that employers need 
now, these are the skills they're likely to need 6 months from 
now, 12 months from now, 24 months from now.
    And with that type of data you not only get these 
efficiencies, now you have this possibility someone can go to a 
college and not just get the next job, but be able to be given 
the skills that can help them build a career, that can take the 
next two or three steps in their life. So we want to be a 
tailwind to them, and very supportive.
    Mr. Marshall. Sounds good. I'm going to start my Community 
College Caucus here someday. I need to do that. Thank you so 
much for being here, and I yield back.
    Chairwoman Stevens. Great, thank you. And at this time 
we're going to begin a second round of questions.
    Dr. Brynjolfsson, as you've kind of defined the two urgent 
economic challenges around lack of productivity growth and too 
much inequality, and then gave us a list of pretty cogent and 
solid recommendations on how to address those, do you mind 
weighing in a little bit around some of the ethical 
considerations that come up on this topic, and how those either 
might be urgent right now, or might become more urgent as we 
move forward?
    Dr. Brynjolfsson. Absolutely. I think those are some of the 
most urgent challenges. They're a little outside some of the 
economics, but some of them also have an economic implication 
as well. Machine learning systems have been remarkable at 
helping us make all sorts of decisions, but one of the things 
we've also discovered is that they're only as good as the data 
that go into them, and oftentimes machine learning systems that 
are trained on decisions that humans made end up perpetuating, 
or even amplifying, the biases that we often have. So when it 
comes to hiring, or making credit loan decisions, or who gets 
parole, if the humans who are making those decisions have a set 
of biases, those are going to be captured by the systems and 
repeated. So there have been a number of academic studies 
that--these are one of the challenges.
    There's both a challenge and an opportunity there. Part of 
the challenge with machine learning systems, particularly when 
they're using deep neural net technology, is that it is 
difficult to understand what's going on inside the black box. 
They capture data, sometimes from thousands or millions of 
examples, and they spew out a recommendation, and it's hard to 
know exactly why, and that makes it challenging to second guess 
it and say, wait a minute, this may not be right.
    But the opportunity is that we can use techniques like one 
called a Turing Box, where you have repeated sets of inputs, 
with different characteristics going in, and sets of outputs 
coming out, and you start learning what kinds of biases the 
machine may have inadvertently picked up, and you can correct 
those in a way that may actually, ultimately, I think, be 
easier to correct than our own human biases. Because, after 
all, it's not like humans are perfect either.
    So I wouldn't necessarily rule out using machine learning 
systems for some of these challenges, even when they are 
imperfect, but we should put very high on the agenda better 
understanding of some of the ethical and other biases that they 
can create.
    Chairwoman Stevens. And, Ms. Kowalski, coming out of your 
taxonomy that you helped to lead with MXD, do you mind just 
chiming in on some of the job roles that you identified that 
might be pertinent to some of the points that Dr. Brynjolfsson 
just talked about?
    Ms. Kowalski. Yes. It's a great question. There are five 
that I think really, really pop out of the work. One is what we 
call the digital era enterprise ethicist, and that's a 
conceptual title, of course, no one puts that out there, but it 
was, you know, an individual success profile of a role of who 
gets to make those decisions. Who makes the call? Who says how 
far is too far?
    Traditionally these decisions have been kind of bandied 
about, maybe IT owns this, or Risk owns it, or Legal owns it. 
Well, now, the way organizations are built in the digital era, 
it does not land neatly in one of those silos, it spreads 
across. And so where the buck stops actually is in a place 
where no one ever imagined it. And so there are--you made a 
comment earlier about how processes haven't caught up, so 
that's decisionmaking processes, that's organizational 
structures. It's a recognition that there's distributed 
decisionmaking more and more now in organizations, and we still 
have an end-of-year code-of-conduct compliance, you know, mind-
numbing 2-hours of training that we take that don't actually 
get to can you identify the decisionmaking framework that your 
organization uses for developing new products, solutions, or 
making decisions around human beings? That's a fundamental 
issue that has to be dealt with now.
    A couple other things, in terms of just roles that you're 
going to see popping up, obviously an organization only has one 
ethicist like that, but does have to establish the framework 
that supports it, but some of those specialist roles, like the 
machine learning specialist, the collaborative robotic 
specialist, the autonomous mobility engineer, right, how do you 
make sure that, you know, people of different ethnicities are 
recognized by that autonomous vehicle, right? How do you make 
sure that your H.R. systems are wired not to filter people out, 
but actually to bring people in, based on potential?
    So those are some of those roles that we see coming up 
across all organizations, and obviously a few of those are 
quite specific to the manufacturing sector. And it's important 
that we figure these out, because what I see right now is a lot 
of organizations just trying to spread that responsibility out 
without actually recognizing that those need to be defined 
disciplines.
    Chairwoman Stevens. As we talk about technical talent, and 
the push for the hard-skilled trades, and the work that we see 
out of our community colleges, and the push for people to go 
into apprenticeship, and other training programs, we still feel 
the need to train for analog, but also embrace the soft skill 
digital. And I'm slightly over time, but with just the 
remainder that I'm going to steal here, I'd love for each of 
you to just comment on this shift here, and the balance of the 
soft with the hardnosed technical skills that are still 
required in many jobs. And, Dr. Ellspermann, if you want to 
start, we'd certainly----
    Dr. Ellspermann. I'd be happy to. We recognized 3 years ago 
that we weren't doing enough to prepare students to be 
successful in the workforce: Number one, making the right 
decisions in the careers, being prepared for the world of work, 
because not every student anymore comes to us already with some 
prior work experience, and that they would be successful so--
building that in, so we are in the midst of rolling out what we 
call our Career Coaching and Employer Connections, which 
ensures every student, when they begin with us, begins building 
a career action plan, which includes work and learn experiences 
in industry to build some of that kind of real-world work.
    We build in, certainly, soft skills throughout the 
curriculum, but those skills are learned best on the job, 
making sure every student has that experience before they get 
out there. But it is an early and often experience, meeting 
with employers being out there, interviewing, understanding 
what's expected. And we know there's a lot to be done that 
we've never been really asked to do in the past, but is 
required by our industry, and know that that's a part of the 
future.
    Ms. Kowalski. So I'll pick up on this theme of moving from 
analog to digital roles. So, if you were to look at the 
research that we have, you'd see that 28 percent of those 165 
new or highly evolved roles are sitting on the production 
floor, and what we estimated was about 1 to 2 years of building 
up that talent that would prepare them to take on progressively 
more digital roles.
    Because at the heart of it, the shift is really from doing 
things physically, physical operations, to accomplishing those 
operations through systems and technology. So you see a lot 
more skills like quantitative, tech-assisted, optimization-
focused, integrative, mobile, virtual, and remote. That wasn't 
in the lexicon, really, 5 years ago, even 3 years ago. You 
know, organizations that were starting to talk about it were 
the OEMs (original equipment manufacturers), for instance, that 
participated in this study. Now it's spreading throughout the 
supply network, and we have quite a task in front of us to gear 
people up, because right now they'll have to bridge from those 
more tactical analog roles into the transitional. So 
organizations have to keep a foot planted firmly where they are 
now, and reach for the future.
    Dr. Brynjolfsson. Thank you for that question. I think this 
is a very important issue, about the balance between hard and 
soft skills. I teach at the Massachusetts Institute of 
Technology, so certainly I have an appreciation for the 
importance and value of hard skills. There are a number of 
technical capabilities that our workforce is lacking and that 
we need to supplement. In some cases, they can be compensated 
very highly. But I also want to stress that soft skills are 
increasingly the ones that are less automatable, and therefore 
more humans will be needed to do those softer skills. They 
often have a longer span of relevance and usefulness.
    In the science article that I included as background, we 
created a framework for which tasks are suitable for machine 
learning. And, indeed, the ones that were less likely to be 
automated were many of the softer skills, involving creativity 
and interpersonal skills, persuasion, caring, coaching, 
leadership, and teamwork. These are things that are very 
important in the workforce, and I also think that there are 
opportunities to teach them, not just on the job, but by 
reinventing and reorganizing our educational curriculum. And a 
research agenda to better understand the kinds of skills that 
are needed going forward, I think, would be a useful supplement 
to be able to map our strategies, both in education and 
workforce training, going forward.
    Dr. Lupia. I'd just like to state a principle and an 
example. One of the overarching principles for this problem is 
the idea of values-based design. So when you build a new 
technology, oftentimes we're thinking about the products, and 
we're not thinking about the people. And so you don't think 
about the people, and the workers, and the consumers, until the 
end of the process, when the unintended consequences and the 
inefficiencies are already built in. A lot of our recent 
misadventures with Big Tech, I think, are an example of not 
thinking about the people at the beginning.
    So now, when we think about the future workforce, with 
values-based design, we're thinking about the people in the 
workplace, and how they're going to interact. If you think 
about that--starting at time one, when you start to build the 
code, when you start to write the algorithms and so forth, 
there are all kinds of efficiencies that you can realize later 
on. And one of the efficiencies, with respect to the workforce, 
is personalized practice. Because once we think about how the 
new technology, and the new workplaces are going to affect 
people, now we can start to understand the set of skills that 
are going to be needed, and we can start to design personalized 
education so that people can learn efficiently the skills that 
they will need in this new place. But if you start with values 
at the beginning, you get to those outcomes.
    And in the point of practices, NSF is already trying to 
help support this through its Advanced Technological Education 
Program, or ATE. There are hundreds of community colleges and 
48 ATE centers around the country that are really preparing 
students for STEM (science, technology, engineering, and 
mathematics) and the skilled technical workforce. We've got 17 
million Americans in the skilled technical workforce now that 
are in the workflow. They're building the machines, and 
maintaining the computers, and so forth, and the ATE Program is 
really meant to encourage and improve the training of science 
and engineering technicians at both undergraduate and secondary 
levels. So the things we're doing right now are things like 
ATE, but the future benefit really comes from thinking through, 
you know, what are the human impacts of technology?
    Chairwoman Stevens. Thank you. And I'm lucky that my 
colleague likes me, because I spent some of that liking capital 
going slightly over, but it was really to hear from all of you, 
and to have your expertise. So, at this time, I'd like to 
recognize my good friend, Dr. Jim Baird, for 5 minutes of 
questions.
    Mr. Baird. Thank you, Madam Chair, and my question now is 
going to be directed at all of you, at some point here. But, 
you know, online, you know, I have grandchildren that can use 
these faster than they could talk, almost, and so my question 
relates to that, in a way. We're using online courses for both 
formal and informal education, and so I guess the question is 
this: Do we have any research that tells us what online 
courses, and how to make those effective? And then also, how do 
online courses, and what you're doing--and AI relate to STEM 
education? We're carrying a bill about the STEM careers, and so 
on. So I'm going to start with--at your left, my right, and 
move that way, go ahead. Thank you.
    Dr. Ellspermann. Congressman Baird, let me just say that I 
think we realize that online education is here to stay. It's 
not going to take over all of education. It's not the best way 
for all of education. It's not the preferred learning style for 
many. But we know, as I shared earlier with that single mom, 
she's got to have that opportunity to learn. So we have to--as 
educators, it's our responsibility to improve it constantly. 
It's come through many iterations. It'll go through many more, 
but it'll also be hybrid, and augmented, and many things that, 
as technologies we're talking about here today, ever greater 
enables us to make that online experience more real, more 
virtual, more--in the way that that learner wants to learn it.
    But I think we understand, as community colleges, we have 
to lean in, and it's not an either/or, it's an and, it's a 
both, and we need to continue to evolve. So we study, we know 
we have a gap between our face-to-face and our online learning. 
It's double digit right now, which is not acceptable, so our 
goal is to eliminate that gap. That will be one measure of 
quality, but we will continue to look for ways to make that 
experience better for the online learner.
    Ms. Kowalski. So I would agree with my co-panelist here 
that it is a both/and. We have a number of occupations that 
employers won't accept a fully virtual experience for, so they 
require some sort of hands-on. I'm not going to let you touch 
an aircraft wing unless you have actually touched an aircraft 
wing before you come into my hangar, thankfully, right. And yet 
the promise of this is pretty profound.
    So if you think back to the statistics that I shared 
earlier, in terms of the gap that we have facing us in 
manufacturing right now, the only way to close it is to become 
incredibly resourceful about who we bring in from the 
sidelines. Women are certainly one untapped resource, but what 
about people with physical and cognitive disabilities? Some of 
the greatest advancements made in digital technologies actually 
allows them to participate. The exoskeleton Dr. Lupia shared 
before is a marvelous example of how we can bring people in 
who, before this, have never even imagined actually having the 
ability to participate in workforce.
    Strictly in online education, and kind of what we think of 
as the standard, this is part of how ManpowerGroup is helping 
our associates upskill. We're offering all of our associates 
access to free education so that they can move up, with this 
idea. And just to validate what you were saying earlier, 6 to 8 
weeks, that's the ability of an individual who's working full 
time, sometimes two jobs, and raising kids. So it opens up more 
opportunity than we've ever seen before, but it's not going to 
be the only way that we can educate, because there are some 
things fundamentally that require hands-on.
    Dr. Brynjolfsson. Thank you for that question, Dr. Baird. 
At MIT we've been doing a lot with online education for quite a 
while. One of the first big courses that we did was an online 
circuits design course. A couple hundred thousand people took 
it. Anant Agarwal organized it. One of those students was 
actually in Mongolia, and got a perfect score on it. It turned 
out to be a 17-year-old boy, and it was someone who wouldn't 
have been reached otherwise if there weren't this kind of 
technology. MIT went ahead and admitted him to the regular 
program, and it was somebody we probably wouldn't have found 
otherwise.
    We have put all of our regular courses online through the 
Open Courseware for free. People can just access and read them. 
In fact, you can see my syllabi, and see my lecture notes, and 
problem sets. There's also an online system called edX. It's a 
consortium of universities--it started with MITX, then Harvard 
and others joined--that coordinates course materials to have 
them in a little more structured way so that you go through a 
curriculum. And these are what we call MOOCs, massive online 
courseware. I think there was an early wave of hype and 
excitement about them, you know, taking over, and doing all 
sorts of things. It worked very well in some areas, like the 
Circuits Course. It didn't work so well in others.
    It's certainly not a silver bullet, but I think there are 
four things that we've learned. One is that, for many 
applications, you can get enormous scale, and much lower cost, 
than we could've previously. Second, one of the unexpected 
benefits was an ability to personalize. People learn at 
different rates, and there's different media that work better 
for other people, and you can have things extremely customized, 
and even personalized, and we're learning how to do that 
better. Third, it often makes sense to do a hybrid system, 
where you have people meet in person, particularly for some of 
the softer skills we were talking about. We often combine where 
people physically get together, know their classmates, do 
things together, then work separately online, then come back 
together, which is actually how a lot of workforce works as 
well, after all.
    And then last, but not least, in fact, probably most 
importantly, I think that the biggest lesson is that there is 
no one best way of doing online education. What we need to do 
is continually experiment and test. The success of a lot of 
tech companies has been this approach of A/B testing, 
constantly trying a new product, seeing if it works with 
different subsets of people, and we've very much taken that to 
heart with our online course offerings, and companies like 
Coursera, Udacity, have been very successful in trying things. 
And sometimes they work, sometimes they don't, but it's an 
attitude of experiment testing. So your question was spot on, 
what is the research showing what is working, and what isn't 
working? And there's a whole set of things that have failed 
miserably, another set of things that have succeeded. But I 
think we're still in very early days, and the digital approach 
allows you to gather data at a scale, and cost, and speed that 
just can't be matched in other ways.
    Dr. Lupia. Well, thank you for asking that question. At NSF 
there's a foundation-wide effort to really support basic 
research on how to develop, evaluate, and improve online 
learning structures. One common way of doing it is you collect 
a lot of information about the types of things people need to 
know, you correlate that with information about the types of 
tasks that they may be asked to do, you integrate that with 
information about curricula, and how people are doing in 
learning environments, and you take all that data together, and 
then you can really evaluate not just what does somebody 
remember after they take a test, but what can they do 6 months 
later? So there's all kinds of projects like that being funded 
at NSF, from trucking to farms and there's even one for 
veterans. So the idea is, you know, how do you structure 
curricula to help veterans who want to get into STEM pipelines, 
because veterans have special abilities, and sometimes special 
challenges.
    I guess the biggest headline, in terms of what we've been 
doing recently, is--about a year ago the Boeing company gave 
$10 million to NSF to try and really boost activity in this 
field. And, within the last few weeks, we have announced five 
new awards to study open source learning platforms to try and 
train and re-skill workers at a larger scale, and these were 
just announced. It's going to be done at the University of 
Southern California, Purdue, Northeastern, Colorado School of 
Mines, and Oregon State University. They're all getting a 
couple million dollars to test some really big ideas they have 
in different ways. So it's, like--what is it, ``coopetition,'' 
or something? They're doing it in different ways, but they're 
all going to be able to learn from each other.
    And I think this is, you know, our approach is to fund a 
lot of different innovations in the hope that some of them 
figure out something really innovative that can be spread all 
over the country.
    Mr. Baird. Well, thank every one of you, and thank you, 
Madam Chair, for letting us have that amount of time.
    Chairwoman Stevens. Well, before we bring this hearing to a 
close, it is evident that we are having a hearing with giants, 
in terms of the expertise of our witnesses here today. And it 
was not shared, but the new Dems have a Future of Work 
Taskforce that Congressman Bill Foster chairs, and I'm a part 
of. Some of our colleagues who do not sit on the House Science 
Committee, we will be sharing with them this testimony here 
today, all of your testimony, and the questions.
    And certainly we find ourselves in a profound, and 
exciting, and sometimes perplexing moment, and so your expert 
testimony will guide our Committee going forward, and help us 
to embrace some of these challenges, turn them into 
opportunities, and continue to push forward in a measured and 
data-driven way, and in a way that really respects where our 
economy is heading, and can head, and how we push to continue 
to support the workforce of the future.
    So thank you all so much for coming to Washington today, or 
taking some time to come to the Science Committee to join us 
for today's hearing. This record will remain open for 2 weeks 
for additional statements from Members, and for additional 
questions that the Committee may ask of the witnesses, and of 
which we are expecting. So, at this time, our witnesses are 
excused, and this hearing is now adjourned.
    [Whereupon, at 5:22 p.m., the Subcommittee was adjourned.]

                               Appendix I

                              ----------                              


                   Answers to Post-Hearing Questions


                   Answers to Post-Hearing Questions
Responses by Dr. Arthur Lupia
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]


                              Appendix II

                              ----------                              


                   Additional Material for the Record


            Letter submitted by Representative Haley Stevens
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]

      Executive Summary submitted by Representative Haley Stevens
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]

          Statement submitted by Representative Haley Stevens
[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]

                                 [all]