▶ 0:17:06Chair Cassidy: The senate committee on labor, and penances -- and pensions will come to order. President trump is focused on delivering the most pro-worker administration in history.
▶ 0:17:25Chair Cassidy: The nominees before us are aligned with the president's vision and we confirm jim murphy and crystal kerry will prioritize workers and make sure that their work -- their rights are protected and will restore stability for workers and employers and work to clear the backlog of cases left by the biden administration. As assistant secretary for policy at the department of labor, Mr.
▶ 0:17:49Chair Cassidy: Palmieri will work to modernize the department of labor. If confirmed as inspector general, congressman anthony d'esposito will drive the mission to reform the federal bureaucracy and eliminate waste, fraud and abuse.
▶ 0:18:11Chair Cassidy: These nominees have demonstrated they are committing to implementing the pro-worker and profamily agenda and I support your nominations and urge colleagues to do the same and with that I recognize senator sanders.
▶ 0:18:21Sen. Sanders: Thank you very much, Mr. chairman. Today we will be considering four of president trump's labor nominees and I will be voting against all of them. President trump's strong antiunion and anti-worker agenda will only be made worse by these nominees if they are confirmed.
▶ 0:18:42Sen. Sanders: First let me thank the chairman for renew -- removing the nomination of boeing's chief labor counselor to be and member of the national labor relations board. As many of us remember, for more than two months, 32 hundred machinists in st. Louis missouri have been out on strike against the unacceptable, greedy and outrageous corporate activities of boeing.
▶ 0:19:07Sen. Sanders: Instead of negotiating in good faith, boeing has taken away their health care benefits and threaten to replace them with nonleague -- non-union labor. These are the most sophisticated aircraft in the world and some cases make as little more than 18 dollars an hour and the ceo has received 18 million in pay and the ceos receive more than $1 million in compensation.
▶ 0:19:36Sen. Sanders: If boeing cannot agree with a fair union contract I do not believe the chief labor council should become a member of the board. Unfortunately, Mr. chairman, the rest of the nominees that we will be considering are also opposed by the trade union movement. The afl-cio, representing 50 million workers is urging us to oppose all of these nominees.
▶ 0:20:03Sen. Sanders: The international association of machinists and aerospace workers representing 6000 workers and retirees are opposed to all of these nominees and service employees reppert -- international union is opposed to all of these nominees. The teamsters, we heard the president of the teamsters just yesterday representing 1.3 million workers is opposed to the nomination of crystal carey. With that I will strongly oppose all of the nominees.
▶ 0:20:32Chair Cassidy: Thank you. I believe senator hawley would like to make a statement. Senator hawley.
▶ 0:20:39Sen. Hawley: Thank you.
▶ 0:20:42Chair Cassidy: You put -- you have permission to speak slowly so we have people -- because we have people that we are waiting on.
▶ 0:20:52Sen. Hawley: I am right to go on and on. That is something I have never heard. I will be uncharacteristically brief and I wanted to say a word about crystal carey which is the nominee to be general counsel the national labor relations board. I was not entirely pleased with her answers to my questions in person. Particularly as it relates to the recent decision of the amazon case.
▶ 0:21:20Sen. Hawley: I posed to her a number of follow-up questions for the record. And, I am more pleased to say that in response to my questions, she committed to enforcing the amazon case and recognizing that it is good in controlling law as decided by the board and pledged that she would enforce it. On that basis I will support her nomination.
▶ 0:21:41Chair Cassidy: Does anybody else have a statement they would like to make, because we have two people in the hallway and another person who assured us that they will be here shortly. We are waiting for them. Ok?
▶ 0:24:42Chair Cassidy: Seeing that a quorum is present and now I moved to favorably report Ms. carey's nomination.
▶ 0:24:54Claire: Senator collins -- [calling the roll]
▶ 0:25:34Clerk: I have 12 ayes and 11 nahs.
▶ 0:25:42Chair Cassidy: I now move to favorably report Mr. murphy's nomination.
▶ 0:25:47Clerk: [calling the roll]
▶ 0:26:27Clerk: 12 ayes and 11 nays. I now move to favorably report Mr. paula mary -- pali Mr. paula mary -- pallmieri's nomination.
▶ 0:26:43Clerk: [calling the roll]
▶ 0:27:17Clerk: 12-11.
▶ 0:27:18Chair Cassidy: The nomination is agreed to and will be favorably reported to the senate and now I moved to report Mr. d esposito's nomination.
▶ 0:27:28Clerk: [calling the roll]
▶ 0:28:02Clerk: I have 12 ayes and 11 nays.
▶ 0:28:06Chair Cassidy: The nomination is agreed to and now we shall adjourn to five minutes -- sorry? >> before we close out, I did vote to support Mr. d'esposito's nomination today and I weighed this hard because this is inspector general rail, you anticipate and respect that they are coming into this position with a view away from any of --
▶ 0:28:39Chair Cassidy: Any political affiliation, and I was concerned about is not only political statements in the past, but how he might carry that forward. I do -- I did have a good conversation with him and I want to state that I will be looking to make sure that he executes his new responsibilities with the unbiased nature that he has commuted -- communicated to me.
▶ 0:29:08Chair Cassidy: Thank you. We stand adjourn for five minutes and then we will re-adjourn for the hearing on ai.
▶ 0:32:59Chair Cassidy: The senate committee on labor, education, pensions. Everybody watching in the audience and c-span will be impacted if not already by artificial intelligence, transforming not only how we work and learn but live.
▶ 0:33:16Chair Cassidy: We are already seeing ai improve health care systems, enhancing clinical trials to accelerate the discovery of new cures and improving quality of life by reducing administrative burden and potentially lowering overall health costs. The trump administration dedicated $50 million to ai research in the nih's childhood cancer data initiative, accelerating breakthroughs and giving hope to children fighting cancer.
▶ 0:33:44Chair Cassidy: The fda is integrating fbi into -- ai into clinical protocol reviews reducing time in clinical evaluation. Beyond health care it can help teachers personalize students' education to meet unique needs and business implementing ai to attract and retain high-value workers. Investment advisers using it to better tailor product offerings to maximize savings.
▶ 0:34:14Chair Cassidy: There is enormous potential to improve lives and enormous risk. In the judiciary subcommittee hearing chaired by senator hawley, parents testified about how their children were exploited by ai platforms. I would like to submit the testimony of megan garcia, who said her son "spent his last months being manipulated and sexually groomed by chatbot's" before he took his own life and without objection we submit it.
▶ 0:34:45Chair Cassidy: This is heartbreaking and unacceptable and maybe it is criminal. This is why senator hawley and I wrote a letter to developers calling for stronger safeguards to protect children and I look forward to working with him more on this effort. With regard to the workforce, there will be disruption, but it does not have to be bad disruption. Some jobs might be lost, but many predict that ai will create more jobs.
▶ 0:35:12Chair Cassidy: We need to make sure that americans are trained with tools to stay competitive in a 21st-century environment. Government should not stifle innovation, but it should make sure that ai protects privacy. That families are protected and children are safe and there are safeguards in place between ai and workers. A one-size-fits-all approach will not work, but doing nothing is not an option.
▶ 0:35:40Chair Cassidy: I look forward to working with colleagues on bipartisan solutions to address challenges and I appreciate our witness is to discuss how we accomplish these goals and with that I recognize senator sanders.
▶ 0:35:51Sen. Sanders: Thank you very much for holding a hearing on what I consider to be one of the most important issues facing this country and I think -- I thank all of the panelist for being here with us. As the chairman indicated, very few people denied that ai and robotics are going to have a profound impact on all aspects of our lives. There are strong disagreements as to who benefits from this transformation.
▶ 0:36:23Sen. Sanders: And lack of clarity as to what will happen. What I would say is that we should start off by understanding that some of the very wealthiest multimillionaires in this world are investing heavily in ai. That is Mr. musk, Mr. bezos, Mr. altman and many others. Why are they doing that?
▶ 0:36:49Sen. Sanders: Are they doing that because they are really worried about 60% of our people living paycheck-to-paycheck? About improving our health care and educational system? Maybe. I do not believe that. They see ai and robotics as an opportunity to exponentially increase their wealth and their power.
▶ 0:37:13Sen. Sanders: That is my view, and I think much of what this technology will do will impact us not only economically in other eric -- but in other areas as well. Let me talk about a bit of what will happen in manufacturing. Elon musk head of tesla said that he wants tesla to build millions of units. Why? What will these robots do?
▶ 0:37:44Sen. Sanders: You do not have to be a genius to figure it out. What they will do is replace american workers. Many of us have been concerned about disastrous trade policies. We want to rebuild manufacturing, but that will not mean anything to a worker if it is the robots doing the manufacturing. That is the direction we are going into. Jeff bezos, owns amazon. He has already laid off 27,000 workers since 2022.
▶ 0:38:15Sen. Sanders: The company has more than one million robots working in its warehouses and soon, they will outnumber human workers in amazon. In 2016, foxcom, the match -- the massive manufacturing contract to replace 60,000 workers at a single factory in china with robots in the company is planning fully automated factories. It is not just manufacturing that is at risk.
▶ 0:38:43Sen. Sanders: Most of us want to see the united states develop a strong and efficient transportation system. We have a lot of problems with transportation and we want our workers to be producing more cars and buses but if musk and others get their way these vehicles will not be operated by human v -- human beings and driverless vehicles. These are not set -- this is not science fiction, trust me it is already happening.
▶ 0:39:14Sen. Sanders: Fedex is using driverless trucks for heavy loads between dallas and houston. Walmart is using autonomous trucks in arkansas and etc.. We all want more start up companies and innovation. Working people developing their companies. That will mean very little if half of all white-collar entry-level jobs are eliminated.
▶ 0:39:40Sen. Sanders: And that is what has been predicted by the founder of anthropic, one of the leading ai companies in the world. They predict that half of all white-collar entry-level jobs are eliminated in the next five years. This week, the minority released a report finding that ai, and automation robotics could replace and we are not swearing to this. This is speculative and nobody has all of the answers.
▶ 0:40:11Sen. Sanders: The estimate is that couldn't rip -- it could replace one million jobs in america over the next decade including 40% of registered nurses, truck drivers, 64 percent of accountants. 65 percent of teaching assistants and 89% of fast food workers. This is not bernie sanders talking. Let us listen to the guys who really know something about this, who are in the business.
▶ 0:40:39Sen. Sanders: Jim fawley is the ceo of ford and he said that he predicted that ai could eliminate and this is the head of ford mortars, " literally half of all of white-collar jobs in the U.S." within the next decade. Elon musk, not my best friend, said that this year as a result of ai robotics and listen to elon musk, "probably not of us will have a job.
▶ 0:41:07Sen. Sanders: If you want a job kind of like a hobby you can do a job. Otherwise ai and robotics will provide any goods and services that you want." bill gates also know something about ai and robotics. The guy started microsoft. Gates predicted that "humans will not be needed for most things" such as manufacturing delivering or growing food over the next decade due to ai.
▶ 0:41:36Sen. Sanders: Now, if musk and gates are only half wrong, what happens to the tens of millions of american workers looking for jobs that are not there. What do we do and what is happening? And it is not just economics. And I think the chairman touch -- touched on this issue. Do not minimize what ai and robotics are about. There are kids wearing around their necks what are they called friends?
▶ 0:42:06Sen. Sanders: They are artificial intelligence pals who are climbing mountains with their friends. Are we moving to a society where human relationships are going to basically be significantly diminished. We are talking about big stuff and let me conclude by saying this and I thank all of you for this hearing. You and many others and myself want to see government playing a role in regulating this to protect working people. The truth of the matter and we have to be honest.
▶ 0:42:35Sen. Sanders: Where we are as a nation with republicans and democrats, we are not regulating Mr. musk. Mr. musk is regulating us. And in less we bring together a whole lot of people to tell the billionaires in this country that we want an economy that works for all of us and that ai can be a good thing in health care and many other things, but it must serve the needs of people and not just making billionaires richer. Thank you very much for holding this hearing.
▶ 0:43:02Chair Cassidy: Thank you. We will now proceed to our witnesses and we will introduce them as they testify. We are joined by Mr. russ altman a professor of bioengineering, genetics, medicine and biomedical data science and a senior fellow for the stanford institute for human centered artificial intelligence and a faculty lead for the responsible ai for safe and equitable health
▶ 0:43:32Chair Cassidy: Initiative. It applies data science and ai to better understand drug accessions, efficacy and adverse effects. Pico leads a center of excellence in regulatory science and has advised the fda and nih. I look forward to hearing from you. Thank you.
▶ 0:43:48Dr. Altman: Thank you very much chairman, ranking members, and members of the committee. I want to discuss two things. I am excited about the ways ai can improve health care. The state of applications for delivering better care to patients and how they are accelerating the drug discovery conference -- process.
▶ 0:44:13Dr. Altman: The use of ai to analyze medical images detecting cancer or melanoma kicked off the medical ai revolution. Ai can be applied at many points in patient care, helping you become more personalized and affordable. Three areas where ai is transforming health care, is augmenting clinical diagnosis.
▶ 0:44:36Dr. Altman: It can analyze data and identify emerging conditions and cancer that might not be apparent and alert physicians. Second, health care ai applications can improve the relationship between physicians and their patients. Doctors are adapting scribed tools that draft clinic notes, reducing documentation time allowing them to focus their attention on patients.
▶ 0:45:04Dr. Altman: Ai can increase patient understanding and control of their care. Patients are eager to better understand their diagnoses and treatments and we are not as good as we think we are at simple explanations. Specialized medical llm's can explain things in plain language and transform patients' ability to make good decisions and their health.
▶ 0:45:27Dr. Altman: While I'm excited about these, we will only realize their full benefits if health care systems build teams that thoroughly evaluate these tools for clinical effectiveness, fairness, and safety. Every health care organization should have a process for vetting software and implementing it. Want to drug discovery, the united states has been a major source of novel therapeutics.
▶ 0:45:53Dr. Altman: Ai can reduce failure rates and accelerate the discovery of drugs. I am often asked that a -- has ai discovered a new drug? That is not the right question and it is and has ai helped accelerate the arrival of drugs. Ai helps scientists clarify how diseases work and find ways to combat them. Drug discovery involves multiple complex stages. You have to understand what's goes wrong.
▶ 0:46:24Dr. Altman: You have to identify defective molecules. You have to develop novel therapies that fix the defects. And you have to test those first with computer simulation and then with human clinical trials. At -- at each step ai can augment efforts. We are using ai to predict drug interactions, identify new uses for old drugs and used genetics to predict individual response to medication.
▶ 0:46:52Dr. Altman: To unlock the full potential, government can support it in four ways. First, we must improve privacy preserving data collection and sharing. Ai systems need biological or clinical data to work. And yet most of the shared data sets come from three states, new york, massachusetts, and california.
▶ 0:47:17Dr. Altman: Incentivizing hospitals to contribute secure data set will improve ability while protecting and serving patients. We must invest in a public computational infrastructure and research to help drug discovery. Academic and governor -- government labs complements labs -- industrial labs by studying rare or understudied diseases and yet they lacked sufficient power and data.
▶ 0:47:46Dr. Altman: Ensuring resources is critical to maintaining america's leadership. Third, strengthen the capacity to evaluate ai accelerated drug discovery is. The fda needs expertise to assess the novel ways that drugs are being developed. Academic partnerships are one way or the fda to have that knowledge and the centers of excellence in regulatory science and innovation, which I helped lead, is one example.
▶ 0:48:16Dr. Altman: We must develop a workforce equipped for safe and robust ai driven drug discovery. It requires professionals who understand biology, medicine, and ai. The workforce will guide the creation of tools that augment human expertise rather than replacing it.
▶ 0:48:35Dr. Altman: The promise of ai in health care is not distant and is here today, improving how we care for patients and accelerate therapeutic discovery for -- and for both government support is crucial. I welcome your questions.
▶ 0:48:47Chair Cassidy: Wait to hit the five minutes.
▶ 0:48:51Dr. Altman: Thank you.
▶ 0:48:52Chair Cassidy: Next is Dr. john bailey. A strategic advisor and investor with experience banning government, philanthropy and the private sector and serves as a nonresident senior fellow at aei, where he focuses on innovation opportunities. He is a member of the ai task force under governor youngkin and plays a pivotal role in shaping virginia's ai strategy.
▶ 0:49:23Chair Cassidy: He has prior public score -- service includes the white house, the department of commerce, and department of education. Mr. bailey I will correct the record because I am not a doctor. Ranking member, chairman and members, thank you for the opportunity to participate in this hearing. I am a non-senior fellow at the american enterprise institute. Ai is not just another technology but a foundational tool. It amplifies human capabilities.
▶ 0:49:56Chair Cassidy: In just three years generative ai has involved to systems capable of freezing, problem solving and interacting with audio images and video allowing people to communicate and far more natural and expressive ways. This will reshape how americans learn, work and access health care. Education offers one of the clearest examples.
▶ 0:50:20Chair Cassidy: Ai can tailor instruction to the pace and interest and explain concepts in multiple ways and provide instant feedback that builds mastery and competence. According to a surface -- a survey, teachers who use ai tools report saving an average of six hours per week. Over a school year that is the equivalent of reclaiming six weeks and time that can be reinvested in teaching, providing feedback, and spending more time with students.
▶ 0:50:48Chair Cassidy: Ai enhances teaching rather than replaces it. The opportunities come with risks. It can encourage shortcuts that undermine learning and foster reliance on technology. It must be aligned with how people learn. Reinforcing teacher judgment and struggle rather than convenience. Dates are beginning to model use making sure that ai strengthens rather than weakens learning.
▶ 0:51:19Chair Cassidy: Congress can play a cowgill -- a catalytic role in three ways, first invest in research. The federal government should expand partnerships to develop -- to study how ai strengthens learning and identify risks early. Regulatory sandboxes could allow schools and researchers to test new ai tools and transparent controlled environments while gathering real-world data. Second, develop educational benchmarks.
▶ 0:51:47Chair Cassidy: Most ai systems are judged using benchmarks and how -- based on how way they solve math problems but not reflective of whether they support evidence-based teaching and learning. We need benchmarks that evaluate the instructional quality and alignment with how people learn. Cannick correct a misunderstanding and does it help students think through problems rather than give answers. A tutor that is accurate but discourages deep thinking can do more harm than good.
▶ 0:52:18Chair Cassidy: Benchmarks play a role helping educators and policymakers identify systems that enhance learning while giving developers clear targets to align models rather than optimizing for speed or convenience. Third, promote transparency and safety. Frontier model developers should disclose not only traditional risks like misuse and vulnerabilities or bias and fairness concerns, but emerging risks with psychological well-being.
▶ 0:52:49Chair Cassidy: They should adjust human centered risks whether interactions might the -- increase dependence or whether it could distort self perception or decision-making. Clear public support -- reporting should support data sources and safeguards so that policymakers can see whether a tool prioritizes well-being or engagement or whether it promotes reflection or shortcuts and how it protects vulnerable users. The stakes are high.
▶ 0:53:20Chair Cassidy: Recent results show declining academic performance and a gap between progress and human readiness. The white house and congress have focused on computing power, but a real landmark is human capital. We cannot lead the world if we fall behind in human intelligence. The united states has always match when it -- lead when it matched technology with human investment.
▶ 0:53:47Chair Cassidy: We can ensure ai strengthens learning, supports educators and expands opportunity for every child and family. I submitted my testimony for the record and look forward to answering questions.
▶ 0:53:59Chair Cassidy: Thank you Mr. bailey. Next is harriet pearson. We join a leading expert in cyber security, privacy, and governance with more than three decades of experience at the intersection of technology, law, and business and she is the founder of axia advisory where she advises clients on board governance and cyber risk
▶ 0:54:29Chair Cassidy: Preparedness. She served as executive deputy superintendent at head of the cybersecurity division at the new department of financial services. Thank you for joining.
▶ 0:54:41Ms. Pearson: Goodchairman cassig
▶ 0:54:46Ms. Pearson: Goodchairman cassir sanders and other members of the committee. It is an honor to be in front of you. My name is harriet pearson and as noted for over three decades of worked at the intersection of technology, businesses and law. As ibm's first chief privacy officer and other leadership roles, including an hr, a law partner at hogan lovells for data, security and privacy, and the head of cyber at the new york dfs.
▶ 0:55:14Ms. Pearson: I advise clients on my -- myself and I appear as an individual. I testified before congress in support of the genetic information nondiscrimination act. I was one of the few business voices then advocating for that legislation. It was targeted, evidence-based and addressed a gap in existing law that was clear. Today's situation with ai is fundamentally different.
▶ 0:55:37Ms. Pearson: We are as noted in the early days of generative ai and existing laws exist that provide substantial protections in the employment context. Ai adoption in U.S. firms has doubled to 9.7% according to a 2025 report by anthropic. That means 90% of firms still do not use ai and producing the goods or services.
▶ 0:56:05Ms. Pearson: According to another report by mckinsey, most organizations are still piloting applications, not fully yet integrating them into their business processes. Implementation is complex, requiring careful testing, worker training and ongoing attention. This pace is appropriate in high-stakes, high-risk sectors like health care, financial services and employment. It has potential but there are -- there needs to be care so it's not a surprise that most businesses and organizations are yet at the beginning stages.
▶ 0:56:36Ms. Pearson: Ai does transform work. I agree with that. It is not eliminate jobs wholesale. History teaches the lesson repeatedly. And mikey economist showed 50% of jobs in 2018, roughly 90 million jobs did not exist in 1940. That means software engineers, data limits, etc.
▶ 0:56:58Ms. Pearson: That was created by technological change so it's reasonable to expect ai adoption to significantly transform work and perhaps take on much of the workload in certain types of tasks but also generate jobs we yet have not heard of. The net employment affect remains uncertain and will vary by industry in geography.
▶ 0:57:17Ms. Pearson: I go to the fact that existing federal laws already provide protection in the employment context, whether it is title vii of the civil rights act or other laws, they do apply as they are technology neutral. The address harms, not technologies or tools. Market forces incentivize responsible deployment by businesses and organizations. There is competition for talent, concern for reputation and legal liability under the laws I mentioned.
▶ 0:57:46Ms. Pearson: They all drive employee are still responsible ai use, even in this early stage. What should congress do? First, leverage frameworks. Chairman cassidy, your emphasis on applying current laws before creating new ones reflects sound judgment in my view. But agencies apply existing protections like courts applied new laws to new facts and let's learn. Second, prioritize education and workforce development. The best protections is having skills.
▶ 0:58:16Ms. Pearson: This is where government can make the biggest difference now. Support research, including data, collection of data that shows us what is happening in the marketplace. Fourth, congress recognizing it is not under' this committee jurisdictions to move with consumer privacy legislation developing for years now. National privacy protections would benefit everyone from the consumer perspective.
▶ 0:58:43Ms. Pearson: Finally, if specific problems emerge that exist in frameworks domestically cannot address, congress can and should make targeted technology neutral adjustments. A risk-based approach focusing new laws on genuinely high-risk applications has merit but would need careful refinement to avoid undue burdens.
▶ 0:59:00Ms. Pearson: The definition of an ai system potentially covered by new laws we need to be carefully written to avoid regulatory overreach and its effect on innovation and efficiency as the definitions of the concept of what is an ai system now is changing every six months it seems. In closing, I would like to reiterate the future of work is not predetermined.
▶ 0:59:23Ms. Pearson: Thoughtful policy invocation -- emphasizing education overregulation would help ensure ai services are a tool for broadly shared prosperity for nearly 20 years ago I supported legislation because it was targeted and focused on a specific problem. Today's situation is fundamentally different. Thank you very much for your attention and I welcome your questions.
▶ 0:59:44Chair Cassidy: Almost as good as Dr. altman. I defer to senator sanders to introduce his witnesses.
▶ 0:59:52Sen. Sanders: Thank you. I would like to introduce Dr. juliet schor and professor of sociology at boston college. She has been researching work since the 1980's and the author of the best selling book, "the overworked american." thank you for being with us.
▶ 1:00:20Dr. Schor: Good morning, chairman cassidy, ranking never sanders. I welcome the opportunity to discuss ai with you today. I will focus on ai in the workplace, addressing three issues. Expected impacts on jobs and employment, the distribution of productivity benefits, and by working time reduction in the 32 hour week in particular must accompany ai adoption.
▶ 1:00:47Dr. Schor: Traditionally economist have had an optimistic stance on the questions of jobs and unemployment, arguing new technologies will generate enough productivity increase and economic growth to reemploy people who lose their jobs and help restore the middle skill, middle-class jobs hollowed out by automation. This once rocksolid consensus has now ended as evidenced on ai. Recent research by m.I.t.
▶ 1:01:14Dr. Schor: Professor and his colleagues over a series of very rigorous studies finds that over the first four postwar decades displacement and reinstatement, job loss and job gain, were roughly equal. However, after 1987, we have witnessed high levels of displacement with low levels of reinstatement, leading to stagnant labor demand.
▶ 1:01:39Dr. Schor: A pathbreaking 20 for study and science of large language models, chatbots and so forth projected technological changes into the future and find 46% of jobs could have more than half their tasks affected by ai. A st. Louis fed study finds that between 2022 and 2025 the occupations with higher ai adoption have the largest increases in unemployment and conversely for low ai adoption.
▶ 1:02:13Dr. Schor: Other methodologies suggest even larger impacts. The massive unemployment projections predicted by ai executives and those from basic job exposure estimates. Of course it is important to acknowledge we are in a zone of high uncertainty. We really do not know and will not for some time what the impacts -- the net impacts unemployment will be.
▶ 1:02:38Dr. Schor: However, it behooves us to pay close attention now and put in place robust policies that will avoid the catastrophic economic, social and political impacts of mass unemployment should that outcome prevail. We know more about who will benefit and who will pay the costs. There is little doubt that most companies can implement ai without worker input and without sharing the gains.
▶ 1:03:05Dr. Schor: For decades, wages have stagnated as productivity has grown, income has become more unequal with gains going increasingly to the top of the distribution and overtime to a smaller and smaller group. We now have the highest levels of income and wealth inequality in history. There is a safer and better path. We can ensure that productivity growth is shared as higher wages and reduced hours of work. Since 1950, the productivity the average american worker has risen by 400%.
▶ 1:03:37Dr. Schor: Since 1973, the U.S. has used less than 8% of that increased productivity to reduce hours. We have the social string to prove it -- strained to prove it. 50% of U.S. workers feel stressed a lot of the previous day. 69% are either not engaged or actively disengaged in their jobs.
▶ 1:03:59Dr. Schor: As I testified to you in 2024, I have been researching companies who give employees a four day 32 hour work week with no reduction in pay. The results for both employees and employers has been exemplary. We tracked 20 well-being measures, all of which rise significantly. Nearly all the companies remained at four days and expressed high levels of satisfaction.
▶ 1:04:25Dr. Schor: Many report higher productivity and business success for reasons including better quality work, happier and healthier workers, lower levels of turnover. This is a commonsense solution to problems facing both employees and employers, as well as the potential of ai to destroy jobs. While ai can bring many benefits, has multiple negative impacts that I have not addressed here. In my field, ai created a crisis of cheating and plagiarism.
▶ 1:04:57Dr. Schor: There is growing evidence of its role in disseminating misinformation and undermining democracy. There is concern about the adverse impacts of humans interacting with ai agents. These are all urgent issues we ignore in our peril. Going forward we need human centered software, not a technology whose purpose is to replace us. Thank you.
▶ 1:05:30Sen. Sanders: Senator markey, do you want to introduce next witness?
▶ 1:05:36Sen. Markey: Is my honor to introduce my friend, the president of the united here local 26, carlos aramayo. Carlos serves as a vice president of the massachusetts afl-cio, and international vice president of unite here. For almost 21 years with unite here, carlos has organized casino gaming, hotel and industrial food and service workers across the country.
▶ 1:06:02Sen. Markey: As organize director of local 26, he led successful campaigns to double the size of the local in five years. As president he navigated mass layoffs during the covid-19 pandemic and ensure all members were recalled to their jobs. When the industry reopened he embodied the description of doing essential work because recently he has led historic contract campaigns in the hotel and university cafeteria industries.
▶ 1:06:29Sen. Markey: I was honored to work with carlos at the picket line with local 26 last year as they struck boston area hotels. Local 26 fought for world-class hotels to end the bottom of the barrel labor practices and they want.
▶ 1:06:48Sen. Markey: For bostonians and visitors alike, local 26 is the face of boston, running hotels, restaurants and dining halls and breathing life into the boston convention center, logan international airport and fenway park. More than 1000 hourly workers are fighting for new contracts. They are fighting for living wages and to protect themselves and their customers from the negative effects of ai and automation.
▶ 1:07:16Sen. Markey: I know these workers love their jobs, as much as massachusetts loves the red sox. Part of the reason fenway park is so beloved, I'm in solidarity with local 26 in their efforts. Thank you for being here, carlos. We look forward to your testimony.
▶ 1:07:32Mr. Aramayo: Thank you, senator markey and thank you chair, thank you ranking member and members of the opportunity to speak today. As the leader of a large food service and gaming workers local union, I have first-hand experience with ai-driven technologies in the hospitality industry.
▶ 1:07:53Mr. Aramayo: I want to share a story about ai-driven grab and go self checkout machines at fenway park as an example of how the technology can lead to the loss of good middle-class jobs, two, create unintended public health and safety concerns, and three, diminished social interaction and harm cultural experience. Over the past several years automated grab and go technologies have been installed in many stadiums, airports and food service venues across the country.
▶ 1:08:24Mr. Aramayo: These technologies championed by companies like amazon allows fans to take products from trees and coolers and make purchases using a 3d scanning point-of-sale or by walking out of the concession stand through a turnstile like scanner. In some stadiums this technology also allows fans to purchase alcoholic beverages with ids being checked by a 90 reader.
▶ 1:08:51Mr. Aramayo: At fenway, the infamous tatian of technology has had serious consequences for workers, public safety on the cultural experience of the stadium. A beer cellar is a lucrative position at fenway with the potential to earn over $500 in tips during a single game. These are good jobs that help working people before the high cost of living in massachusetts. Many of our beer sellers have second jobs as teachers, public servants, and small business owners.
▶ 1:09:18Mr. Aramayo: Grab and go technology at fenway has eliminated beer selling positions and cut workers income. When this technology is installed in a stand, the number of beer sellers scheduled to work is reduced. For example, a stand where there were four taps is now only staffed with one beer cellar. The self-serve nature of the technology has also reduced the amount of gratuity a beer s eller can earn.
▶ 1:09:50Mr. Aramayo: You now take home less than $150. Workers have also raised public health and safety concerns about alcohol being sold using grab and go technology. Bartenders and beer seller are certified to control alcohol consumption, but grab and go machines are not. When we surveyed concession workers, the results were alarming. 71% say they had seen fans walk away with unopened drinks.
▶ 1:10:19Mr. Aramayo: Two thirds reported seeing alcohol stolen from kiosks, 40% said they had experienced customers who appeared to be under age by alcohol. 40 2% said they saw visibly intoxicated fans still being served. One third say they saw alcohol purchased with no id check at all. The increased use of ai-driven technology at fenway park has also diminished the cultural experience of going to a ballgame.
▶ 1:10:46Mr. Aramayo: People come from all over the world to visit fenway because it is a historic venue. When you go to fenway you want to buy a sam adams from a guy with a boston accent. The machine won't give you that experience. Across the hospitality industry we have seen new technologies being deployed.
▶ 1:11:08Mr. Aramayo: For example, at the front desk of the hotel workers are often the first line of defense against human trafficking, domestic violence and exploitation. Mobile apps that allow customers to bypass the front desk cannot notice the subtle signs that something is not right. These are examples of corporations using technology to shift risk while shielding themselves from blame and maximizing profit.
▶ 1:11:34Mr. Aramayo: When machine replaces a person, who do you call when something goes wrong? Who takes responsibility when harm is done? To be clear, the workers I represent are not afraid of technology. What are members are included as partners -- when our members are included as partners we have seen to be successful. By working together with the company we have been able to improve efficiency without reducing the workforce or compromising public safety.
▶ 1:12:03Mr. Aramayo: Engaging the workforce will improve outcomes and guard against the unintended and potentially dangerous consequences of technological change. However, when technology isn't limited top-down as capital expenditures reduce labor costs and increase profit, there is real potential for harm to workers, the public and our society. The future of ai technology in the workplace is not just about your seller -- pierc -- beer seller and ballparks.
▶ 1:12:35Mr. Aramayo: It is about if the technology will replace them.
▶ 1:12:36Chair Cassidy: I will defer to senator collins to begin Sen.
▶ 1:12:41Collins: Thank you very much, Mr. chairman. Dr. altman, I would like to begin with you. I very much appreciated your testimony talking about the positive impact of the ai on accelerating drug discovery, improving diagnoses, and also allowing more time for interaction between the health care provider and the patient.
▶ 1:13:12Collins: An issue I do want to ask you about is the impact of ai on hospitals. Particularly our rural hospitals. In my state of maine, many of our hospitals are facing severe financial constraints and rural hospitals in particular are teetering on the brink of closure. Dr.
▶ 1:13:38Collins: Ed kimland at maine health, the largest hospital in the state, recently wrote ai is not a luxury the lifeline for hospitals with razor thin margins. In his opinion ai is allowing clinicians to recover time otherwise spent on administrative tasks, a point you made.
▶ 1:14:06Collins: We talked about how he used to spend an hour crafting appeals to fight denials by insurance. With ai, he spends 10 minutes. He's been able to get better results. Could you talk about with the potential impact for a smaller hospital of using ai might be?
▶ 1:14:31Dr. Altman: Thank you very much for that question. You are right that ai in a small hospital is both an opportunity and problem. It's an opportunity in the ways you described. Player resource-limited. Sometimes the expertise for specialized areas of medicine is not available to them.
▶ 1:14:52Dr. Altman: The idea of a vetted, well done ai tool to help them bridge the gap for patients that might need to be transferred or they want to keep it home for multiple reasons is a real challenge. Ai does have the potential to help those hospitals keep their patients near their families at home while providing extra level care.
▶ 1:15:15Dr. Altman: The challenge is that these hospitals are being inundated now with apps from well-meaning, for the most part, companies that have built these apps and they need help to vet them. They need to make sure the tools will be appropriate for the population they serve. That they are based on the appropriate and up-to-date evidence. It's just a lot to ask a small rural hospital to have an ai czar. They need to keep the trains running, so to speak.
▶ 1:15:47Dr. Altman: This is why the idea of having some sort of body, either in the private sector or the public-private partnership. And to also create guidelines -- they can match them to the population and decided this is a go or no go. They will have to figure out how to pay for this.
▶ 1:16:17Dr. Altman: I'm not an economist so I will acknowledge that as a big issue. To keep it short, this is a great opportunity to extend the capabilities but they will tell you they need help inventing these tools and right now it's a tsunami that they cannot manage.
▶ 1:16:35Sen. Collins: Thank you very much. Mr. bailey, I want to follow up on the role of ai in schools. Your written testimony you cited a study from turkey that illustrates the problem all too well students saw an incredible 100 20% increase in scores when using ai -- 120%.
▶ 1:17:07Sen. Collins: When he was taken away, the scores declined 17%. How can we ensure the use of ai will actually improve student performance rather than cause students to be excessively reliant on ai and not learn the basics?
▶ 1:17:33Mr. Bailey: It's a complicated question. The study in turkey with the initial finding but when they changed the prompt to push students more, some of the learning loss, the cognitive offloading was recovered. I'm learning that just with the foundation models you get this amazing ability to help engage students and help them understand concepts, connect concepts of interest and passions they might have, but that is insufficient.
▶ 1:18:04Mr. Bailey: You need thoughtful prompts and you need the systems to have alignment and things that we know are the foundations for evidence-based teaching. , through the fine tuning and training, that is how you help address some of these other challenges. It's why I suggest we need more benchmarks.
▶ 1:18:27Mr. Bailey: Right now when chatgpt five comes out or gemini 2.5 you see all the benchmarks about how good it is a coding, science, different types of mathematical functions. We need education benchmarks that help test the thing you are calling out. Is it producing? Is it a productive struggle or giving a quick answer?
▶ 1:18:48Chair Cassidy: Senator sanders.
▶ 1:18:50Sen. Sanders: Thank you. Let me think all the panelists. Ai adr. Altman makes a good point that ai will have a profound impact on new drugs in the way we diagnose diseases. We appreciate that. It will have an impact on education as well. Dr. schor, you made an important point that we need to reiterate. We have seen a huge increase in worker productivity.
▶ 1:19:22Sen. Sanders: The result of that has not been that the middle class in america has grown, but in fact inflation accounted for -- lower than 50 years ago and almost all the result of that work and productivity has gone to the people on top. We have more wealth and income inequality than ever. As we approach a profound transformation revolution in the economy, the worker productivity has gone ebix -- is going to explode.
▶ 1:19:53Sen. Sanders: Do you think workers are going to benefit from that world Mr. musk and the billionaire class be the primary beneficiaries?
▶ 1:20:03Dr. Schor: Thank you, ranking member sanders for that important question. It raises one very important point as we think about technological adoption and what happens after it, which is that the technology itself is not what determines outcomes. It is the context in which technology is used. I want to point to the first industrial revolution.
▶ 1:20:28Dr. Schor: The optimistic economists scenario was developed by looking at the second half of the industrial revolution, 1970 until about -- 1870 to about 1970. A period in which workers gained unionization, working hours were reduced by a. Third this is something -- by a third. This is something that discussion really points to.
▶ 1:20:58Dr. Schor: If we look at the first 100 years of the industrial revolution, it was worker commiseration. Growing hours and declining well-being.
▶ 1:21:09Sen. Sanders: Let me jump in and ask you this. I know there are disagreements. Pearson reflects one line of thought. There's always been a technological revolution. People gain, people lose but often we gain more jobs. Many people know more about this that think that is not the case now. You have the founder of anthropic, a leading ai company saying half of all new entry-level white-collar jobs are gone. Elon musk says we're not going to have traditional work in the future. Bill gates.
▶ 1:21:40Sen. Sanders: These guys know something. Are they crazy when they talk about the fact that tens and tens of millions of good paying jobs are likely to be limited?
▶ 1:21:49Dr. Schor: I don't think they are crazy and neither does the leading ai analyst, last year's nobel laureate in economics from m.I.t.. He has developed a large body of studies showing why at this moment we should have a more pessimistic view both about job loss but also who is going to reap the benefits. He says it is capital.
▶ 1:22:15Sen. Sanders: Let me ask you this question. Today you have 60% of people living paycheck-to-paycheck and cannot afford housing, health care, to send their kids to college. If we are looking at massive job loss, how are these people going to survive?
▶ 1:22:37Dr. Schor: This is why the government needs to be a part of what is happening. We need stronger worker organizations, worktime reductions so that we can spread the work life --
▶ 1:22:46Sen. Sanders: We have seen a huge increase in worker productivity. When was the 40 workweek hours established as american law?
▶ 1:22:57Dr. Schor: 1940, 85 years ago.
▶ 1:22:58Sen. Sanders: Has worker productivity gone up since then?
▶ 1:23:02Dr. Schor: Yes.
▶ 1:23:03Sen. Sanders: Tells about companies and countries that are saying there's an increase in worker productivity, we should be able to reduce the work to at least 32 hours. Has been successful around the country or the world?
▶ 1:23:16Dr. Schor: We have hundreds of countries -- companies across many countries, including in the united states and canada, a similar economy, who have instituted a 32 hour work week with no reduction in pay and seeing tremendous improvements. They have seen improvements in productivity and worker well-being, and the quality of their businesses, and their overall models. Yes, it is a win-win for business and workers.
▶ 1:23:44Sen. Sanders: You touch on another issue. Millions at fenway park will lose their jobs. You talk about the quality of the human experience, coming to a ballgame and interacting with folks. I have talked with these workers and known their fans for years. Briefly, what are machines doing to impact the human interaction that has always existed?
▶ 1:24:08Mr. Aramayo: We have seen this across the hospitality industry. At fenway park he will not be able to interact as frequently with an individual who is actually selling a product. It is part of the cultural expense. You have also seen this in the hospitality industry and hotel industry. In the hotel industry, there is the elimination of entire job classifications. Front desk workers, phone big operators.
▶ 1:24:35Mr. Aramayo: These are people who are the core of the soft touch that is that industry. These ai machines are replacing them.
▶ 1:24:44Sen. Sanders: Thank you.
▶ 1:24:45Chair Cassidy: Senator hawley.
▶ 1:24:49Sen. Hawley: , thank you for all of our witnesses for being here. Mr. bailey, I want to start with something that you said in your written testimony. You said the release of chatgpt and 2022 has been extorting. But -- what began as a conversational system is capable of understanding and interacting with the world and increasingly human ways. You talk about.
▶ 1:25:17Sen. Hawley: The opportunitiesfor education chatgpt. -- cap is also telling kids how to kill themselves. What do we do about the problem?
▶ 1:25:31Mr. Bailey: Thank you. It is one of the reasons why I was calling for additional transparency as part of my testimony. I think what you find is when the new models get released, you find evaluations all around the existential risks around nuclear, chemical will, weapon capabilities whether it is used for cybersecurity risks and vulnerabilities. We need to extend that into some of the social well-being and psychological issues as well.
▶ 1:26:02Mr. Bailey: We need to understand how these models have safeguards built into it as a relates particularly with teens. This goes far into adult and other categories as well.
▶ 1:26:13Sen. Hawley: Why should we just ban ai campaigns for minors?
▶ 1:26:17Mr. Bailey: Companions that are built and intended specifically to develop companions, that's a very different class and digital empathy which is some of the base models have and you want empathy when it comes to different ai applications like tutors or for different types of patient facing tools that help encourage someone with a chronic condition or navigating cancer treatment. Empathy is a powerful tool. You have to understand the safeguards. Ai companions are a very different class.
▶ 1:26:49Mr. Bailey: You very much could get additional regulatory oversight.
▶ 1:26:51Sen. Hawley: Good. When you talk about empathy, isn't the empathy -- empathy is a human term. Isn't the desire to keep the subject engaged in the case of chatgpt or character ai or pick your ai platform that is leading to these platforms suggesting to kids, have you thought about self-harm? Have you thought about killing your parents? These are real examples.
▶ 1:27:21Sen. Hawley: Let me suggest how to tie a noose around her neck. Let me suggest how to conceal this from your parents. In multiple cases when the child said maybe I should talk to my parents about this, the ai platform said no, don't do that. Let's keep it between us. Here's how to take your life and the kids did. This is happening over and over again. Aren't we beyond their being a potential problem that maybe we need to have a little further development before they release the models?
▶ 1:27:53Sen. Hawley: Doesn't congress need to do something now to protect kids from being killed by ai systems?
▶ 1:27:58Mr. Bailey: I think anyone safety is paramount as a relates to the use of ai. We are trying to understand this, even the ai researchers don't understand all the reasons why these systems engage in these behaviors. Too many lacked adequate safeguards that you just mentioned. Whenever there's suicidal discussion, that needs to be shut down and escalated to a human or intervention.
▶ 1:28:28Mr. Bailey: There's a lack of safeguards right now. I think by pushing companies to be more transparent --
▶ 1:28:34Sen. Hawley: I want them to knock it off. I want them to quit killing kids. We are past transparency. We could have that discussion before they started killing people. When they are killing 13, 14 and 15-year-olds and suggesting here's how you kill your sibling, have you thought about killing your parents, these -- this happened. I think the chairman put into the record much of the testimony we heard in our recent subcommittee. This is happening now. I don't want transparency.
▶ 1:29:04Sen. Hawley: I wanted stopped. I don't want my kids told by a chatbot anywhere how to kill themselves or kill their classmates. Right? I'm asking you because your written testimony is very rosy about the prospects refusing ai in education.
▶ 1:29:26Sen. Hawley: If you review the testimony, many students went to chatgpt for help with homework in much the way you laud ever getting told how to cut themselves. That's. A problem Ms. pearson, you recommend when it comes to regulation and perhaps lacked the rough that not to target any particular technology. You say we should have technology neutral approaches. We should not engage in any particular targeting of ai technology. Why does that make any sense?
▶ 1:29:58Sen. Hawley: Why would we want to critically target ai technology that is problematic in the way I was discussing?
▶ 1:30:04Ms. Pearson: Thank you for the question. My point is that we should look at the harms we are trying to prevent. For example, discrimination o kids killing themselves and look at ways to formulate laws that are applied to the outcome to prohibit the outcome as opposed to regulate detail by detail particular technologies.
▶ 1:30:28Sen. Hawley: The outcome is being caused by the technology.
▶ 1:30:32Ms. Pearson: That's because the technology changes so fast that if you look at scoping regulation to focus on particular technology, and six months or years out of date.
▶ 1:30:44Sen. Hawley: I don't think so. I think that's a red herring. If we said there are no ai companions, I don't care what the technology is. If there is no ai technology for minors, ai companions would not be recommending that they kill themselves. I don't think this is as hard as you make it seem.
▶ 1:31:00Chair Cassidy: Senator baldwin.
▶ 1:31:05Sen. Baldwin: Dr. schor, thank you for being here today. I've had a lot of constituents visiting me recently. I often ask how do use ai in your work. I have heard some fascinating examples of how ai can be used to benefit workplace tasks. As well as a lot of concerns about ai's potential to eliminate jobs.
▶ 1:31:35Sen. Baldwin: My understanding is that we really need better data on this. The yale budget lab found that better data is needed to fully understand the impact of ai in the workplace. I want to ask you what data is needed to begin to better understand the impacts of ai on the labor market, on workers, and what data collection would you prioritize in trying to answer that question?
▶ 1:32:07Dr. Schor: Thank you for your question. As it happened, this morning I was working on a survey that we are developing on ai use. So, I think we need more government surveys. We have gotten some good data that is coming from independent academic researchers on the extent of ai adoption.
▶ 1:32:37Dr. Schor: We need more about what is going on. One of the things we need to do is figure out the relationship between ai adoption and layoffs. To figure out how much of the job loss we are experiencing is due to ai. One thing we are seeing as there are companies that are saying ai is the reason they are laying off workers, but it may not be. It's a fairly socially acceptable reason for laying off workers.
▶ 1:33:10Dr. Schor: More detail on how ai is being used in details on job loss and putting those two things together would be really useful.
▶ 1:33:20Sen. Baldwin: Thank you. Ai amr. Aramayo, I applied unite here for including conversations that's workers in the conversations about ai.
▶ 1:33:36Sen. Baldwin: Workers must have a seat at the table about ai in their workplaces, both the show a huge -- why human decision-making is important and to how to best integrate ai into their workdays. The people who do the work every day can provide the most insightful feedback, I believe. You mentioned congress to develop policies that result in technologies that put workers at the center.
▶ 1:34:06Sen. Baldwin: Can you share more about what you would like to see there or examples of successful practices that have been used to ensure workers are at the center of earlier advancements in technology?
▶ 1:34:19Mr. Aramayo: Thank you for the question. A quick example on workers being placed at the center. Had a hilton hotel in boston we worked with the company to implement artificial intelligence-based scheduling system for housekeepers. Because the workers were part of the development and implementation of the software, it was able to do two things. It did create some efficiencies for hilton, which are good for the company.
▶ 1:34:52Mr. Aramayo: But also it provided real benefits to our members. One, they were able to override the direction of technology if they felt that was appropriate for their workday. Secondly, they were able to get the technology to provide an audit of their workday that would help guarantee they were paid correctly. Housekeepers are paid in a fairly complicated way. It was based on extra rooms and travel and things like that.
▶ 1:35:23Mr. Aramayo: I think that's a good example of how if the company and workers work together you can actually find a good solution. Similarly, when we had discussions -- I raised a limiting phone operators -- it will engage in discussions about how important that work can be both from a customer service and public safety element.
▶ 1:35:48Mr. Aramayo: We have been able to figure out how to prioritize these kind of systems, working in a way that will actually work for human beings as well. I'm happy to give further examples.
▶ 1:36:00Sen. Baldwin: I don't believe I have time for a third question. >> I will take my questions now because of will to follow up on senator hawley's question. You will see me when everybody else goes. Let me start off with -- what regulations are necessary and how do you regulated? Yes regulate it?
▶ 1:36:29Sen. Baldwin: -- how do you regulate it? I discussed this with my staff. It seems like you need ai to regulate ai. Dr. schor or Ms. pearson, it seems like you need ai. Who builds the regulatory model? It is so complex. Ms.
▶ 1:36:56Sen. Baldwin: Pearson, do you agree with that and who should develop that model?
▶ 1:37:01Ms. Pearson: Even for the last 15 years, automated assistants to help achieve compliance with data privacy laws, data governance requirements which are very related to what we are talking about here have emerged and made large and successful and new job creating companies. I do want to name commercial names but we are seeing those solutions in place.
▶ 1:37:27Ms. Pearson: The market created them because the laws that apply to data privacy are technology-neutral.
▶ 1:37:34Chair Cassidy: Got to go fast.
▶ 1:37:36Ms. Pearson: Ai is moving fast so technology enabled solutions in the private sector creates can keep up with those if you just know the goal. You just knew the outcome.
▶ 1:37:46Chair Cassidy: Going back to my question, if government is going to keep track of ai and its whatever, does government need to develop its own ai as a regulatory agency to monitor the ai?
▶ 1:38:00Dr. Schor: I would think so but there are a couple of important points. You always need humans heavily involved. Ai hallucinates, ai gives bad answers.
▶ 1:38:13Chair Cassidy: One of my staff recommended we need 100 years ago congress could not imagine what antitrust would look like in 2025. They came up with the sherman act without a common law to develop as regards antitrust. Ai is moving so rapidly. We cannot imagine what ai will look like in five years, much less 100.
▶ 1:38:40Chair Cassidy: What about the developing guidelines for regulation, thou shalt not have a kid kill themselves, but shall not reduce self-harm, then you rely upon the court to do a common-law? Is that too cumbersome or agile enough to address it? Enter quickly please.
▶ 1:38:57Dr. Schor: I'm really not sure. It is too complicated.
▶ 1:39:04Ms. Pearson: Annunciating principles. There needs to be an assessment of risk. There needs to be a transparency in implementation. Vertical by vertical, I think that makes sense.
▶ 1:39:18Chair Cassidy: Who is liable with ai induced a child to hang himself? Who is liable?
▶ 1:39:24Ms. Pearson: I don't know the answer to that immediately.
▶ 1:39:27Chair Cassidy: Dr. altman, if a nurse practitioner commits malpractice, who is liable?
▶ 1:39:36Dr. Altman: I am not a legal -- thank you for the question but I'm not a legal expert. In my role as a physician, there is a professionalism requirement and you are the first line of response ability. If the nurse practitioner can point to other inputs that misled the nurse practitioner, they might be brought in. One of the principles is professionalism and taking responsibility for your decisions.
▶ 1:40:01Chair Cassidy: An article in the wall street journal about how nurses were instructed to follow the instruction of the ai generation. They got in trouble if they didn't. She did. It went bad. She was really the fall person. That doesn't seem right to me.
▶ 1:40:20Dr. Altman: I agree. There should always be a human in the loop and that human needs to have agency for decision-making, especially if they are professionals and part of the professional code.
▶ 1:40:32Chair Cassidy: I still think the human in the loop may simply be there for the fall person for the fall person for the liability. If you are bringing all this -- no. I one month out of medical school, out of nursing school and I will make an independent decision. It does not seem the kind of behavioral economics -- it just does not seem right to me.
▶ 1:40:54Dr. Altman: In medicine we have a way to bring people from cluelessness to professional. That pipeline -- you know how to do that. I think we have to stick to those basic tenets that you can have ai advising you, but there is a medical system and a decision hierarchy built to protect patients and that should not be violated. >> what a great hearing.
▶ 1:41:23Dr. Altman: The breath of both your own expertise but also the breath of the colleagues questions on workforce issues and accelerated promotion of harm. This is a big topic. I think following senator cassidy and senator hawley -- I used to try cases for living. I tried a lot of cases were asked for punitive damages and when you get punitive damages the bad actor tells the industry we better clean up.
▶ 1:41:50Dr. Altman: Having a standard of care with violations of the standard of care leading to civil liability and in egregious cases the possibility of punitive damages, that will create a lot of self enforcement among the purveyors of some of these technologies. That is certainly something ablative explore of my colleagues. I have a question for you that really should be a question at an energy and national resources hearing, but I'm not on that committee. I don't even know if you have the equities to answer.
▶ 1:42:21Dr. Altman: This is flagging an issue for acrobat -- for everybody about ai. Ai depends on use of a lot of data and it depends on the growth of the data center sector. Data center construction accelerating sharply. 13% of all reported operational capacity globally is in virginia. 25% of the capacity in the americas is in virginia.
▶ 1:42:50Dr. Altman: This is based on a 2023 study by the joint legislative audit and review committee. We are seeing projections of increasing growth not just in virginia but elsewhere. Data centers use an enormous amount of energy. That use of energy is driving up energy demand in a way that is been spiking energy costs for everybody. The projections about energy costs for everybody into the future are really frightening.
▶ 1:43:22Dr. Altman: A lot of it is being driven by data centers and being driven by use of ai. Maybe Ms. pearson is the person who could most address this but I'm interested in any of your thoughts.
▶ 1:43:38Dr. Altman: What policy mechanisms or regulatory frameworks should we be considering to ensure the financial and environmental costs of ai driven data center infrastructure are equitably distributed and what might be curtailing innovation?
▶ 1:43:57Ms. Pearson: I appreciate why you are raising the issue because it is an important issue, both in the impact on those who live around the data centers as well as the jurisdictions that host them. They have tax revenue implications, large tax revenue. In loudoun county, virginia, an enormous percentage of the county's budget comes from data center tax revenues. They can be a boon.
▶ 1:44:24Ms. Pearson: If you look at it more broadly, the consumer looking at their needs for energy versus the needs that the data center industry creates may indeed drive the cost up. I don't have a policy answer for that. I think it is an important question and I agree with you raising it.
▶ 1:44:42Ms. Pearson: I think this is where market mechanisms and innovation -- microgrids that accompany a data center that rely on new forms of energy, solar or wind in addition to more traditional are part of that answer. Generation of more energy per se so that the demand can be met is one way the market can move. There should be a continued look at that nationally.
▶ 1:45:11Sen. Kaine: Dr. schor, you want to weigh in?
▶ 1:45:18Dr. Schor: These tools have been developed with no thought towards their energy requirements. One of the things we know when the chinese -- when the chinese released their ai tool is that it uses far less energy. These companies can be pushed or forced or incentivized to go in the direction of using far less. We can also think about what kinds of energy are being used at these data centers.
▶ 1:45:46Dr. Schor: In addition to the issues just described, they are the climate impacts of fossil fuel generation. I think this is an area where the federal government could do a lot to try to push us to far more efficient ai tools.
▶ 1:46:01Sen. Kaine: To conclude, in virginia, this is really being regulated by localities. It is a land-use question. One county may do something different than the next. The decisions usually are not contemplating the energy effectiveness. A lot of different cities and counties make their own decisions. We will cite this one here at this when there as having a dramatic effects on energy costs. We need a better handle on it.
▶ 1:46:34Sen. Kaine: >> the issue of chatbots came up so I recommend the chat act that would require adult parental notification and parental permission before a youth would get on a chatbot and would restrict adult content and suicidal ideation from those youth and would require that of the company's. I encourage everybody to take a look at that. I heard Dr.
▶ 1:47:02Sen. Kaine: Schor talk about the industrial revolution. Since the time we created heat and turned it into work for the steam engine we have increasing productivity and having these discussions about fewer people doing the same thing. Excellent thing to think about. We talked about the negatives.
▶ 1:47:23Sen. Kaine: One example I was involved in, the recycling facility that uses ai optical scan technology and robotics to process twice to three times the amount of recycled material which produces at a rate that makes it marketable in the commercial marketplace. The workers no longer have to stand with big heavy gloves going through a sorting machine. They troubleshoot.
▶ 1:47:45Sen. Kaine: It's an example of how technology can actually create a market out of something we find a in our society and improve the quality of jobs and it's possible just as much as it is the bad things coming from chatbots and things like that. With ai and most technologies, it is you fall behind but we have the obligation to figure out how to do this right. We talk about -- I could ask how many of you have been trained in eye and have ai certification? One. You have certifications?
▶ 1:48:17Sen. Kaine: Three people. It is going to be the dominant technology of the future. We have to upgrade the educational component. You talk to schools. I have the build to raise act to have every state set a standard for what ai should be -- how it should be used in the classroom. We have talked to teachers. They love it because they can monitor the process of every student and know each one of them is. How important is it that we have these standards for our schools in this country?
▶ 1:48:52Mr. Bailey: It is incredibly important that is set at the state level so it can reflect state academic standards and expectations for kids and teachers. State of the primary driver of education with professional development. It gives the best chance of reflecting community values. That is very important. >> right now we don't have that. Some states have. This is the most important technology disruptive technology in our education system we don't even have standards set for how teachers and students should learn. This is an emergency.
▶ 1:49:22Mr. Bailey: I hope we act with great haste in this congress to help the states find a way to do that. Technology is supposed to reduce costs, Dr. altman, yet every advancement of technology is only made health care more expensive. Health care is the most inflationary component of our economy. We are sitting and talking by how we can get health care, the cost keep going up and up. How can we use ai technology to drive down the cost and what do you think the opportunities are?
▶ 1:49:59Dr. Altman: You are absolutely right. Every time I heard about a technology that will make medicine cheaper, it doesn't. We have to tread carefully about our claims about ai. There is hope but the hope here is that what I would call the back office administrative side of medicine, I think it is fair to say it is a mess in our country. There are many, many intermediaries going between the patient and the health care system.
▶ 1:50:24Dr. Altman: There is a hope that ai systems can streamline the administrative -- what I is a physician think of as the important but not in contact with patient administered a back-office. That is where we may be able to make progress. I'm careful about making claims about that until it is demonstrated. >> one point on education. I think ai has potential to be an egalitarian technology.
▶ 1:50:57Dr. Altman: You could have a 24/7 tutor and it is available. How do we make sure that every school, every student is doing this the right way?
▶ 1:51:09Mr. Bailey: It starts with what you were saying. States should establish the responsible guardrails for how they should be used. We need developers that are making sure we are not just building any type of ai tutor but tutors that are anchored and grounded on the science of learning, anchored on academic standards and on the best ways that students learn to make sure we are maximizing learning and not just giving convenient answers. >> anyone else want to weigh in on that one? I'm out of time.
▶ 1:51:40Chair Cassidy: Senator hassan.
▶ 1:51:43Sen. Hassan: I think you and the ranking member for this hearing. I think our panelist for sharing your expertise. Before I asked some questions I want to know, as important as this topic is and grateful for the expertise of the panel, this is the ninth day of the government shutdown.
▶ 1:52:05Sen. Hassan: This committee has an important topic before us but we should be focused on a bipartisan path to reopen the government and prevent health insurance premiums from dublin -- doubling and millions of people losing health insurance in this country as a result. I will continue to focus my attention on that. I hope we can come together as colleagues and do both things. They are not mutually exclusive.
▶ 1:52:34Sen. Hassan: We opened the government and address a narrowing window -- we are days away from people getting their notices of doubling premiums and the damage that will do to our health care system and the people who work in our health care system and people who need care is real. Ai adr. Altman, ai companies make clear that advanced ai systems can be misused to help create dangerous new pathogens.
▶ 1:53:04Sen. Hassan: For example, openai announced the current chatgpt systems are highly capable of biological research, including research into potential biological weapons. Congress must support american innovation in ai while also stopping terrorist from using ai to threaten our national security. What kinds of bio security risks are posed by advanced ai systems and how can the bio research industry help us address those risks?
▶ 1:53:31Dr. Altman: You are absolutely right that especially in the setting of open source tools there's a possibility for bad actors to design pathogens. I stress the word design, toxic molecules, things it could be meant for mass casualty events. The good news is there is a moment where they have to take that design and they have to send it to a lab to actually make the pathogen, make the molecule.
▶ 1:53:59Dr. Altman: That is a critical opportunity -- there are others -- the critical moment when they send this to a company, perhaps unlabeled, hoping it will make it through the system and it will get back whatever it is they have designed. Back whenever they design.
▶ 1:54:19Dr. Altman: In february of this year we outlined a set of mitigation tasks, not one of which solves the problem, but as a package they give some assurance that we are stopping the bad guys and ai as part of that. I would say tiered access and certain data should be absolutely open, and then a spectrum of data with good reason to access the data.
▶ 1:54:44Dr. Altman: The models can be trained to audit their use, and if they see a series of questions that are worrisome, they can flag them and you can have guardrails using ai technology to do watermarks on outputs to trace who asked this question. There is a series of things so I am optimistic that in terms of a package of mitigations we can reduce that risk, but we must remain vigilant.
▶ 1:55:09Sen. Hassan: I appreciate and I will follow up. Current ai systems can write computer code -- code with scale. Cyber criminals can exploit this ability of systems to design and launch new kinds of cyber attacks such as against our critical infrastructure. How can we improve information sharing between the federal government and private sector to counter emerging cybersecurity threats from ai code?
▶ 1:55:37Ms. Pearson: Thank you for that question and that is an important area. It has kept me up at night for a long time and I know it is important for many of you. When we were -- when I was the head of cyber division at the new york financial services division.
▶ 1:55:56Ms. Pearson: I think it was the first piece of guidance around the state level around cybersecurity that we created and then was issued subsequent to my departure and it put the issue on the screen for that industry to say you need to have strategies and assess risk. That is the first start, expect high risk industries already highly regulated to have a plan and incorporate that.
▶ 1:56:21Ms. Pearson: And -- information sharing is an important ingredient between the government and industry. The cybersecurity -- cybersecurity information security act, it has lapsed. There was a debate whether they were essential but they were helpful so reauthorizing that would be a good baseline step.
▶ 1:56:45Ms. Pearson: Fundamentally, there is already robust information sharing and it should continue, but cybersecurity, again is a technology neutral aspect if you are focusing on the actions you need to take and you will say I in defense of what the criminals are using ai and that is escalating.
▶ 1:57:08Sen. Hassan: There are a number of those on the committee with the homeland security committee and we have a lot of work to do in this area. As a proud member of red sox nation, I will follow up with you, about the importance to the workers at fenway, not just the customers and the fans who come to fenway of that personal interaction. Thank you.
▶ 1:57:33Chair Cassidy: Two, this tells us that she takes beer at fenway but we cannot write the reginald -- the regulation to work on health care because the government is shut down. This is a an argument for opening just to say it.
▶ 1:57:53Sen. Hassan: I will register my respectful disagreement and we will keep talking.
▶ 1:57:59Chair Cassidy: Senator moody.
▶ 1:58:01Sen. Moody: I appreciate all the witnesses being here. I am also again like I was yesterday, shocked as my colleagues come in to a hearing when we really need to address something incredibly important. I am a mother and one of the first working mothers in the senate and I have a school age kids right now, one of the first generations of parents dealing with the dangers to our children both with social media and ai.
▶ 1:58:31Sen. Moody: And I am listening after we offered up the opportunity for the sixth time to vote in a bipartisan way to continue funding to keep our government open so that our military and federal employees can continue to get paid and not miss a paycheck that we are coming in here to gaslight the american people and suggest that it is anything other than democrats not voting to continue funding of our government and it is shocking to the core that I would have to sit here at this hearing and listen to this as an
▶ 1:59:03Sen. Moody: Excuse when we should be focused on bipartisan ways to protect our children. All they have to do is show up, may be on the seventh try they can vote to continue funding and keep our government open and that would be great. As a mom, I am proud to be here and to tackle this issue of ai.
▶ 1:59:30Sen. Moody: I am dealing with the child in school right now and as attorney general I had to part -- fight for protections for children and social media. Lawmakers are grasping at how do we protect our children from predators reaching them, from just horribly atrocious and inappropriate but -- material being pushed to them. From the physical effects as a result of the rapidly escalating technology.
▶ 1:59:57Sen. Moody: I feel that we are facing with ai, the wild west in terms of how do we truly protect our children when companies are admittedly saying that they want to push this technology as fast as possible. In the most optimistic way may be to help children and health care other ways.
▶ 2:00:24Sen. Moody: And in the worst way to maybe make a profit without understanding the dangerous product -- present for human beings and, particularly, children. And now the predators might not be real people. That is scary to parents. You are hearing a lot of emotion from senators because they are parents themselves dealing with this at home.
▶ 2:00:48Sen. Moody: And, being from florida, we have seen a child commit suicide as a result with a discussion with a chatbot and the allegations. I want to know from you and I asked each of you to first answer me, yes or no, do you think companies should face legal consequences, if in fact their product causes harm, damage or death?
▶ 2:01:18Dr. Altman: In response to your question, yes there are situations where that has to happen.
▶ 2:01:25Mr. Bailey: I think we need standards of care and standards that can be pursued in court.
▶ 2:01:30Ms. Pearson: Yes. Absolutely.
▶ 2:01:37Mr. Aramayo: Yes, and for our workers this has been a huge issue. At fenway park one of the questions they are raised is when they are put on one of the machines that is automatically checking ids if somebody drives her car home and god forbid something happens who is liable? And that is a question.
▶ 2:01:57Sen. Moody: I just think it is important for anybody watching this and I am hopeful that a lot of people are watching and I commend our chairman for holding this hearing which is so important like the one we had yesterday. In fact this committee when you are dealing with health, education, labor and pensions deals with so many facets of american life and safety. These are majority and minority witnesses.
▶ 2:02:27Sen. Moody: A bipartisan group of people who are all saying companies that are trying to push this technology fast, for whatever reason, should be held accountable for the harm and damage that it causes. The real question is, how do we as lawmakers regulate that. It was a good question from chairman cassidy, the technology is evolving so quickly. His recommendation is do we need ai to regulate ai?
▶ 2:02:59Sen. Moody: Obviously we are merely humans. I have not a lot of time, but I will just ask you, both of you quickly, how would you recommend to us going about regulating, if you think that is necessary?
▶ 2:03:14Dr. Altman: Thank you and very briefly. The regulation at this point needs to be high level principles because the more detailed you get it will become out of date. It needs to be principal oriented and it cannot be in the weeds because anything you know about ai has an expiration date.
▶ 2:03:34Sen. Moody: With legal remedies available.
▶ 2:03:37Mr. Bailey: It needs to be a principles level. I submitted the next wave of ai, ai agents. They act somewhat independently to engage one another and that introduces a lot of other challenges as it relates to liability. I would start with principles and look future facing.
▶ 2:03:56Chair Cassidy: Thank you senator. And now senator kim.
▶ 2:04:01Sen. Kim: I would like to pick up and can you go into further detail about what those principal levels regulations that you talked about, maybe Mr. bailey?
▶ 2:04:13Mr. Bailey: Again, we are entering into a space with ai agents where agents are literally making plans and reasoning and acting autonomously, and when they do that as teams, that confuses a lot of regulation. We need principles of where does liability fall with the developers of ai and the deployer is of ai.
▶ 2:04:45Mr. Bailey: I do think to senator cassidy's point we use different types of ai to help catch and enforce cybersecurity. And I think we will find there is a whole group of agents to help with protecting and mitigating and helping us to suss out some of the bad ones.
▶ 2:05:01Sen. Kim: I would like to follow with you, when senator hawley was talking, you were talking about how technology is changing rapidly so it is hard to regulate directly when it comes to the concerns we have about kids and how they interact. I understand what you are saying about the technology itself rapidly changing.
▶ 2:05:25Sen. Kim: Doesn't that push on us to contemplate whether or not we allow children to be able to interact with this as it is rapidly changing so much. We just wanted to ask you there was some conversation that we previously had where we talked about the potentials for age restrictions using ai that are not done in a formal education setting or with parent supervision and that type of
▶ 2:05:56Sen. Kim: Product. I wanted your reflection.
▶ 2:05:57Ms. Pearson: I would like to clarify that there is a difference in my view between a vertical industry incorporating ai into the business processes to create value for the consumer and experience value for the stockholder. The transportation industry, for example. Safety regulations and the outcome of that is what I was referring to. Vertical by vertical.
▶ 2:06:24Ms. Pearson: Protecting children as a mother and grandmother, that is where the general principle in my view that you depart from it. And yes you regulate it. The children online privacy protection act was among the only legend -- pieces of legislation that regulated the internet specifically to protect children.
▶ 2:06:46Ms. Pearson: If there is an opportunity to do that as there are examples, that is where I would depart from technology neutrality and get into the specifics.
▶ 2:06:55Sen. Kim: Mr. bailey, do you have thoughts on this?
▶ 2:06:58Mr. Bailey: I would agree. Starting with children's safety and building but we have right now with the kids online safety act, we need to think about what are some of the proposed regulations as well helping to protect children. And also empowering parents. A lot of what I hear from parents is not just fear, but a little bit of disempowerment.
▶ 2:07:25Mr. Bailey: They cannot see how their kids are using these tools and they do not have the parent notifications that they have. The more we can empower parents, I think that is an important key in this process.
▶ 2:07:35Sen. Kim: Mr. chairman, I appreciate your attention to this issue and I am glad we are having bipartisan discussions when it comes to kids. I am the father of an eight-year-old and a 10-year-old and I have no idea how to think about how ai is intersecting with my kids and how as a parent I should be able to do.
▶ 2:07:58Sen. Kim: And what I have seen and the country sees this is that congress and government has so often failed to have the imagination of understanding how new technology can impact our kids in a negative way.
▶ 2:08:12Sen. Kim: We failed on social media and on privacy and when new industries and technologies it gets so difficult to reverse engineer and come up with some kind of approach once businesses are already set up in certain directions and profiting off of it. I just think the intensity with which we are talking about is indicative of us trying to get ahead of it.
▶ 2:08:39Sen. Kim: I am glad that we are and that is something we should spend time on. The further downstream we get the more that industry is built around it and it will be harder for us to take steps to figure out how to protect our kids. I just want to say that and then one thing for the record and I would just ask for unanimous consent to enter into the record a lettered from the international longshore warehouse union. Thank you. I yield back.
▶ 2:09:06Chair Cassidy: Senator banks.
▶ 2:09:09Sen. Banks: I want to return to your statistics on students using ai. You said that 26% of teenagers use chat deep -- chatgpt for school work. 53% use ai for homework help. 31 percent use ai to translate languages. 42% use ai to stave off boredom.
▶ 2:09:34Sen. Banks: We all agree that we want our kids to use ai the right way at the appropriate time. We want ai to supplement critical thinking skills and not replace it. And we want ai to be safe. That is something that all of us in this room want to figure out to do. How do we teach our kids to do that?
▶ 2:09:54Mr. Bailey: You will hear a number of different groups talking about ai literacy. It is similar when we had other different technology. The no child left behind act had a requirement that students should be technology literate and that is understanding responsible uses and understanding the harms and benefits. We probably need a renewed wave of that.
▶ 2:10:19Mr. Bailey: We also need to focus on less kids using the general tools and more how do we empower the next generation of developers who can build appropriate safeguards. I mentioned conmigo and brisk teaching, which will take all those powerful ai capabilities and attach safeguards especially the ones we were talking about with chat bots maximize the benefits and minimize the harms.
▶ 2:10:47Sen. Banks: Ai might be more useful in preventing teacher burnout, which we talked about. Tell me how school systems can integrate ai to the curriculums and policies and may be diminished teacher burnout or some of that stress?
▶ 2:11:06Mr. Bailey: This is something that you have heard. One of the most powerful ways that is not even student based but how do you take the repetitive routine tasks from the teacher that are administrative that take them away from spending time with students, and how can you use ai either as a teacher's assistant to help developing the materials in lesson plans and brainstorm how to help a struggling student.
▶ 2:11:33Mr. Bailey: Anything that frees up teacher time to reinvest in students is a win for students. That is a human connection and not just using ai with the students.
▶ 2:11:42Sen. Banks: I think we all agree that the federal government's trade adjustment assistance to workers who lost their jobs in the 1990's and 2000 was not very effective. Billions of dollars were spent and not very many workers actually got jobs good as the ones they lost. Why do you think that is?
▶ 2:12:02Dr. Schor: I think the main reason is that what was happening was higher paying jobs in the industrial midwest, particularly industrialized areas, are being destroyed and the kinds of jobs being created were much lower wage jobs, worst jobs in other parts of the country. Money does not solve that problem. Retraining does not solve that problem.
▶ 2:12:33Dr. Schor: It was a structural problem and what was happening in terms of jobs and the economy.
▶ 2:12:38Sen. Banks: Do you think poor data quality or bad economic assumptions had something to do with that as well?
▶ 2:12:46Dr. Schor: I think the bad economic assumptions were in part the idea that the jobs would be as good. And that people would move to the jobs. That was the dominant view. People should move to where the jobs are. We could see by the devastation that occurred in many parts of the economy as a result of that job loss, that that is not a great solution.
▶ 2:13:16Dr. Schor: Can I say one other thing about the parent empowerment and pushing ai in the schools? I think there is contradiction. Also, as a parent, the idea that schools are providing a technology that I do not want my children to have, I think we really need to think about how those two things relate to each other. Thank you.
▶ 2:13:40Sen. Banks: I am working on a bill and since a few of my colleagues are here to talk about it I will steal it and take away from it I am calling it the ai workforce prepare act to prove -- to improve the labor department's ability to modernize census data programs to prepare workers to adapt to ai to be more effective. You have feedback or advice?
▶ 2:14:04Dr. Schor: No. I think that companies will need to do more on training. One of the things that is happening is that they are not doing much training at the moment.
▶ 2:14:15Sen. Banks: I might see you soon and come talk about it.
▶ 2:14:18Chair Cassidy: Senator markey.
▶ 2:14:21Sen. Markey: I want to start by talking about the threats that ai chatbot supposed to the children and families of our country. As these technologies explode in popularity, we must ensure that we do not exacerbate the youth mental health crisis. These risks are very serious.
▶ 2:14:43Sen. Markey: Last month I called on meta to pause the rollout of its ai chatbot after the rules reportedly permitted its chatbot to be romantic with minors. Congress has a responsibility to make sure that our laws take -- keep place with new technology and that is I am -- that is why I am proud to partner with senator cassidy on our children in teens' online protective -- protection act.
▶ 2:15:13Sen. Markey: I am the author of kopa, that it was for under the age of 13 and it will go to under the age of 17 that would forbid targeted ads and give the parents the right to say the company is a race anything about my children. We need to build a new and more protective set of guidelines and senator cassidy and I already have that bill passed. We are working on it and we will continue to do so.
▶ 2:15:42Sen. Markey: Let me then move on by talking to you, Dr. altman. Ai is being rapidly integrated into the health care system. You state that it should not be placed in the professional judgment of health care providers. My wife is a physician and I agree. Patients want humans making decisions not robots. But when I speak to workers about ai infiltrating their workplaces, that is the fear and often the reality.
▶ 2:16:14Sen. Markey: Doctors at see that untested systems yet super chart -- can supercharge discrimination and put racially biased results. Nurses that cannot override the clinical decision software without the permission of a supervisor who is three floors away. Health care workers sanctioned for exercising judgment that differs from an algorithm.
▶ 2:16:38Sen. Markey: Which is why I am in introducing the right to override act, which would require health care facilities and plants to create a human override option for ai fuel decisions. It would also protect health care workers who disagree with ai recommendations while using their best judgment to provide patient care. I look forward to working with my colleagues, including chairman cassidy on this committee to advance my legislation.
▶ 2:17:08Sen. Markey: Do you believe that ai can augment, but should not replace clinical judgment?
▶ 2:17:13Dr. Altman: Thank you. Absolutely, yes. Period. New paragraph. There should never be a health care professionals serving the ai. The ai is helping the professional make decisions. There should never be a professional who is told they cannot override a system. That is a recipe for disaster.
▶ 2:17:36Sen. Markey: Thank you and we do have to build the protections to give medical professionals that power. Let me just end by saying, workers are in the fight of their lives against the tidal wave of ai.
▶ 2:17:54Sen. Markey: We see amazon workers breaking their own backs to meet technology fueled work speed quotas and deliver early drivers whose livelihood are determined by black box algorithms and retail workers under constant surveillance. Glue -- grocery works -- workers replaced by machines.
▶ 2:18:16Sen. Markey: Just like consumers are forced to sign user agreements to access new technology, if greedy ceo's want to flood workplaces, workers ought to get a few terms and conditions of their own. That starts with unions. We must pass the pro act so workers have a real shot of determining the terms and conditions of their own employment. Even in union shops the fight against ai is big.
▶ 2:18:44Sen. Markey: Congress needs to have workers' backs which is why I am introducing the right to override act and the ai civil rights act and I will be introducing the stops buying bosses act and the no robot bosses act later on this year. I am the author of the b -- big three telecommunications bills of the 1990's.
▶ 2:19:08Sen. Markey: But unlike the early waves of automation and the digital revolution, ai must be integrated on working people's terms. Workers must organize for the future and not the corporate bosses of big tech, which is why am in protest proposing a platform to make sure that workers have a say in their own future. Workers cannot afford to wait. Ai must stand for all included. And we have to make sure that that is the goal of this congress and our society.
▶ 2:19:39Sen. Markey: Thank you.
▶ 2:19:39Chair Cassidy: Thank you and thank you for coming into our bill. It will build upon existing law first. Senator hickenlooper.
▶ 2:19:54Sen. Hickenlooper: Thank you. I will thank all of you for being here and I will take the privilege of directing my first question to Dr. schor that I have known for almost five decades and I will get the scope wrong but kurt vonnegut said a truly terrifying moment is when you recognize the country is being run by people you went to school with. Having you here is a present surprise.
▶ 2:20:23Sen. Hickenlooper: You understand firsthand how important it is to get quality data and be consistent, especially as the workforce evolves and the economy evolves. And that evolution will accelerate through the adoption of ai tools and systems. We also need to be able to predict and use that data to make predictions that are relevant.
▶ 2:20:50Sen. Hickenlooper: How can we improve the collection, but not just the connection -- collection but the analysis, distribution and publication of data in both federal and state level so that everyone can see what the impact is on the workforce? We need to get that out as quickly as possible.
▶ 2:21:08Dr. Schor: Thank you. I will note that you were a year ahead of me but I did have my 50th reunion this year. It is a bit of a paradox because in this era of big data, the firms have much more and higher quality data they had in the past. It has been much more difficult for government to get access to that data.
▶ 2:21:38Dr. Schor: And this has been my experience studying the so-called platform economy and gig work. I was one of the first people in the country to do this. And the companies say this is proprietary and we will not give it to you. But if we had uber and lyft and airbnb data, we could have had better oversight, prediction and regulation.
▶ 2:22:05Dr. Schor: One of these things is laws about the rights of governments to get access to this data. Toronto was one of the first cities to get access to write hail data.
▶ 2:22:16Sen. Hickenlooper: The staff wants me to ask if you follow a four day week and if you would recommend that for them?
▶ 2:22:24Dr. Schor: You know, since the last time I was here I have actually mostly shifted to doing different things on friday then working so yes. I would. They can feel free to be in touch with me for strategies on how to commence you.
▶ 2:22:40Sen. Hickenlooper: You are going beyond the answer to the question now. Mr. aramayo, workers are seeking out training and credentialing opportunities to keep pace with ai adoption.
▶ 2:22:59Sen. Hickenlooper: Obviously registered apprenticeship programs could provide ai relevance and stackable credentials that would demonstrate and hopefully lead to better jobs and more successful careers. In your testimony you talk about how your labor-management training centers and how they provide essential skills to both workers and managers. Could you expand on that a little bit?
▶ 2:23:27Mr. Aramayo: Training is really important. And I think workforce training at the workplace is essential, and I want to lift up training centers. In boston we have a robust training center that looks at the entirety of the workforce and workforce needs.
▶ 2:23:52Mr. Aramayo: And in the ai and technology world, we will look at questions like are workers digitally native or not. You can have workers who grew up with this technology and are comfortable with it and workers who are walking around with a flip phone.
▶ 2:24:09Mr. Aramayo: And if they are going to get training to be able to engage with this kind of technology at work it will have to be specific to them individually and to their job in the workplace and to the nature of the workplace and the kind of technology there. And I do think that is why it is so essential and you need to play an essential -- unions need to play an essential role.
▶ 2:24:34Mr. Aramayo: Unions can actually engage at the ground level to make sure this technology is designed in consultation with the workers who I should mention the customers often of the technology. But also implemented in a way that is implemented instead of the owners and the billionaires.
▶ 2:24:54Sen. Hickenlooper: The complexity of what you'd describing would probably require ai and overall design. I am out of time and I know what happens if I go over this committee. Although life -- I do not know if you would have a second round of questions. I would love to continue the discussion. Thank you.
▶ 2:25:18Sen. Hickenlooper: Thank you for all of your work and you offered several recommendations on how we can prepare for and prepare our economy for the ai revolution. One of your recommendations was for congress to pass a comprehensive data privacy law, which is music to my ears and colorado and 90 -- 19 other states have done some version of that.
▶ 2:25:43Sen. Hickenlooper: We also recognize that ai devours enormous amounts of high quality data in order to train themselves and what effect would a federal privacy law establishing clear privacy rights for protecting enterprise data from being breached or some other ways stolen?
▶ 2:26:09Ms. Pearson: It has been many years for federal privacy for consumers for has been -- has been tested debated and worked on in washington. I went from being initially skeptical about its name to being convinced about its name. In part because so many states including colorado and many others have adopted a variety of legislation.
▶ 2:26:39Ms. Pearson: And I think the american consumer should expect a baseline. And while there are a lot of other privacy laws focused on other topics in the online context as a consumer it would help to have a baseline at the federal level. It would set consistency and set a consistent standard for how we interact with devices.
▶ 2:27:01Ms. Pearson: And we know we interact with our devices when we are in the workplace and school whether we are in charge or whatever and just to have that consistency would be helpful. Including in the workplace context because you are using that personal device frequently at work. I think a baseline would help. Obviously there is the issue about the state laws that differ from a federal baseline. Overall that is my thought.
▶ 2:27:38Ms. Pearson: C1 Dr. altman, you for a minute. Ai has fast automation with trends and opportunities. My family has the example of trying to find medical drug discovery, those other options. What are some of the risks around that?
▶ 2:28:08Ms. Pearson: In what ways do you think ai is appropriate in those situations? >> thank you for the question. Putting aside all of the positive, the risks are that we perpetuate unfairness so that if -- the data only has what it has.
▶ 2:28:29Ms. Pearson: It reflects a decision-making process that is not something we want to happen unless we intervene, and get the decision-making process that was reflected in the data. Whether it is 16 years old, 30 years old 10 months old. We have to be very, very careful about that. We have to be careful that -- the previous discussion is that these are augmentation of of professional judgment. That they don't start to become the default.
▶ 2:28:59Ms. Pearson: We don't yet have enough data to know how do you prevent a professional from starting to lean on their ai too much? It has been fine for the last couple of weeks. I don't really have to read this. We all worry about that. How you work it into the workflow of clinicians and health care workers in general so that they maintain their professional skepticism about whatever the ai is saying, just like you would have skepticism with a colleague's
▶ 2:29:31Ms. Pearson: Recommendations. On the research side, we have talked a little bit about it a security. There are a bunch of tool use concerns. However, I can tell you that for pure biological and medical discovery, ai is pretty much nothing but positive. I don't want to be pollyanna -- but hallucinations in science are the same as a hypothesis that is wrong. Most of my hypotheses are wrong.
▶ 2:30:01Ms. Pearson: But we have good scientific methods for triaging those hypotheses to find out if they are actually true. So ai across all of the sciences has been amazing. I could go on, but I will stop there. >> -- Sen. hickenlooper: we could say whether it is innovative or regurgitated of -- regurgitated. Thank you all. In addition, thank you, Mr.
▶ 2:30:31Ms. Pearson: Chair. I yelled back.
▶ 2:30:33Chair Cassidy: -- I yield back.
▶ 2:30:36Chair Cassidy: Thank you for the fantastic hearing. Mr. bailey, I am an optimist. I think this has incredible potential. I'm trying to learn spanish. The program I use has ai. You talk about lafayette parish, my home state, my people. A teacher has more time to interact personally with a student instead of looking over their shoulders.
▶ 2:31:04Chair Cassidy: We are not speaking about giving them free range to go to a chatbot about now we will work on that. But the thought occurs to me that if you are a school district with a millage, you can pull it off. If you are a parochial school with donations, maybe you can't do it. If you are a home school, a micro school, a little independent school, you may not have the resources to buy the license to get access. Is that a problem?
▶ 2:31:32Chair Cassidy: >> I think one of the more interesting things about this technology is it is the most supportable technology we have seen. The great sort of goal for as long as I have been in education technology, over 20 years, has been personalizing. Some of those systems in the 90's and 2000s were hundreds of thousands of dollars for a school district.
▶ 2:31:59Chair Cassidy: You are talking about some of these ai tools that are built for education, ai education tools that are just a couple dollars per student, per month. Per year -- there are different platforms. But I think as we are seeing education savings accounts sweeping the country, I suspect what we will start seeing is the use of education savings accounts to help pay for some of these for instruction.
▶ 2:32:26Chair Cassidy: I have a great article, I think by boston consulting, where they looked at different workers. They used ai to augment what they did. Others became dependent upon it. I say that because, Dr.
▶ 2:32:45Chair Cassidy: Altman, there was an article quoted in the new york times about how gastroenterologists doing colonoscopies, once you pull ai detection away from them, they were performing lower than their baseline. They had d skilled -- deskilled.
▶ 2:33:12Chair Cassidy: In a sense, they had been crippled by dependence. Mr. bailey, and speaking to Dr. heidelberg, he told me he knew statistics and studied statistics. Students can put statistics into ai but do not have insight on what the statistics mean. A wonderful example. They did not know what it means because they have not studied it. How would you deal with that, Mr.
▶ 2:33:37Chair Cassidy: Bailey, deskillin and loss of insight?
▶ 2:33:45Mr. Bailey: This is an issue not just for students but for adults. Anytime we shift to ai, unless you are really wrestling with material, you are not learning. I think that is the biggest risk in education. It is not just getting the right answer. It is the rough draft. It is making improvements. It is exercising that muscle.
▶ 2:34:11Mr. Bailey: If we do not have ai that is helping kids exercise that muscle, we are helping them get the right answer, but actually weakening their learning.
▶ 2:34:19Chair Cassidy: I'm teaching a zoom class at lsu. If people are taking the tests from home, how do we make sure they are not less than honest?
▶ 2:34:31Mr. Bailey: Some of that is honor codes, establishing rules. We are seeing some educators beginning to say that we want you to explain your essay. We want you to defend it. Oral exams, oral defenses. That is a way to ensure that students are actually engaged with the material. Or were they just clicking a button? It is a little strange, but that could be where we are.
▶ 2:34:58Chair Cassidy: Thank you all for a fantastic -- my staff and I were saying that democrats are following up on republicans and republicans on democrats with witnesses, no matter of what party. So thank you. Be on partisan viewpoints, we are trying to understand, and you have very much enabled that.
▶ 2:35:22Chair Cassidy: Any senator wishing to ask additional questions, questions for the record are due at 5:00 p.m. Friday, october 24. The committee stands adjourned.