▶ 0:17:27Statement. Good afternoon and thank you all for being here for today's important hearing on the future of artificial intelligence. From the tools powering your smartphone to the algorithms predicting weather, recommending medicines, or helping farmers improve crop yields. Ai is already shaping the world around us, just as we once competed for dominance in space or nuclear technology. The united states is now in a race for leadership in ai. American companies are at the frontier for this race.
▶ 0:17:54These companies are pushing the boundaries of what advanced language models can do, and countless startups and research labs are finding new applications for ai in every corner of the economy. The stakes are high. If the united states leads, we get to shape the standards, the ethics and the economic benefits of this powerful technology. If we fail, we cede such influence to adversaries who do not share our values. So the risks are high.
▶ 0:18:22I will have an impact on all americans across all industries. Ai is driving new efficiencies and creating breakthroughs to improve lives and healthcare. Ai is helping to detect cancer early and accelerating drug development and transportation is making cars safer and logistics smarter. And agriculture is reducing waste and helping farmers feed more people with fewer resources. These advances aren't abstract.
▶ 0:18:46They're happening now and are creating better services, lower costs and new opportunities for american workers and american families. But the technological future of ai remains uncertain. Some experts warn we're just a few years away from the emergence of artificial general intelligence, or the singularity. Others argue the technology has inherent limitations, and we are decades away from the singularity, if it's even possible.
▶ 0:19:10We don't know for certain what future of ai will look like, but what I do know is the future is too important to leave up to chance. We need to do our best to understand what kinds of impact ai can have on our economy, our society, and develop potential solutions now before it's too late. This subcommittee takes seriously its responsibility to examine these issues, and I'm looking forward to hearing today from everyone on both the current state of ai and the possible futures which lie ahead.
▶ 0:19:37It's essential the united states lead not just in building these technologies, but ensuring they are developed responsibly, deployed safely, and used in ways which advance american values. When we get this right, we'll ensure artificial intelligence fulfills its extraordinary promise. I look forward to today's discussion, and to working with my colleagues on this committee to ensure america leads in shaping the future of ai. I now recognize ranking member brown for her opening statement. >> thank you, chairwoman mace.
▶ 0:20:07Artificial intelligence is here and it is already reshaping our economy, workforce and daily life as we work to ensure that america leads in ai innovation. We must also lead in responsible and trustworthy use of this technology. Ai holds the promise to strengthen our economy and make government more efficient. However, when common sense safeguards are absent, technology can deepen inequalities, leave workers behind, or allow bad actors to take advantage of gaps in policy.
▶ 0:20:36Even as we look toward the future, we cannot ignore the ways ai is already changing the workplace. And while some of these changes are promising, we must also work to prepare the american people for change. Workers in my cleveland district and across the country are worried about what automation and emerging technologies mean for their job and their security. Black workers, in particular, remain disproportionately concentrated in positions most at risk of automation, according to research by mckinsey and company.
▶ 0:21:06If we fail to provide retraining, education and pathways into the jobs of the future, we risk leaving entire communities behind. A diverse workforce isn't just good for our economy, but a necessity for our national competitiveness. If we don't ensure that employees most at risk of being replaced by.
▶ 0:21:27I have other pathways for employment, adoption of ai will not only drive greater economic disparity, it will also miss opportunities to diversify and elevate the workforce. A diverse and adequate workforce not only builds up our communities, it also enhances our ai ambitions. We know that foreign adversaries, particularly the chinese communist party, are aggressively pursuing technological dominance.
▶ 0:21:55They are not only racing to outpace us in artificial intelligence and cybersecurity, but also actively targeting our institutions, businesses and citizens. At the same time, everyday americans face scams, fraud and data breaches that threaten their livelihoods and erode trust in government and the private sector.
▶ 0:22:17The future of ai will be shaped by our commitment to getting it right today, and our ability to learn serious lessons and mitigate mun several areas defending against hostile foreign governments, holding accountable scammers who prey on vulnerable communities, and investing in our workforce to ensure resiliency. We must ensure that innovation does not come at the expense of fairness, security or opportunity.
▶ 0:22:46Thank you to all the witnesses that are here today to discuss this critically important topic. And with that, I yield back. >> thank you. Congresswoman brown, I'm pleased to introduce our witnesses for today's hearing. Our first witness today is miss kinsey fabrizio, president of the consumer technology association. Our second witness is Mr. samuel hammond, chief economist at the foundation for american innovation.
▶ 0:23:09And our third witness today is doctor nicole turner lee, senior fellow of governance studies and director of the center for technology innovation at the brookings institute. Welcome, everyone, and we are pleased to have you this afternoon. And pursuant to committee rule nine, the witnesses will please stand and raise your right hand. Do you solemnly swear or affirm that the testimony that you are about to give is the truth, the whole truth, and nothing but the truth, so help you god? Let the record show that the witnesses all answered in the affirmative.
▶ 0:23:39We appreciate all of you being here today. You may sit back down and we look forward to your testimony. Let me remind the witnesses that we have read your written statements, and they will appear in full in the hearing record. Please limit your oral statements to five minutes. As a reminder, please press the button on the microphone in front of you so that it is on and the members can hear you. When you begin to speak, the light in front of you will turn green. After four minutes, the light will turn yellow when the red light comes on.
▶ 0:24:05Your five minutes has expired and we would ask that you please wrap it up so we will first recognize miss fabrizio to please begin her opening statement. Thank you. Good afternoon. >> chairwoman mace, ranking member brown and members of the subcommittee, thank you for holding this hearing and for the opportunity to testify. Cta represents over 1200 companies in more than 80%. Are startups, small and midsize businesses. Our members power the american economy and support more than 18 million jobs.
▶ 0:24:35But before I talk about ai's broader impact, I want to share how it's reshaping my own life. I am a wife and I am a mom of two wonderful kids, and I also run ces, the world's most powerful technology event. Each week, I use ai to organize my life at home, pulling together school pickups, drop offs, sports schedules, and at work. These tools help me make smarter decisions, research competitors, and come up with new ideas for products and services.
▶ 0:25:04In many ways, ai is a personal assistant that gives me back time. Time I can spend with my family while staying focused on leading a complex organization, and it makes a big difference in my daily life. I am really excited to share the even greater impact it has at scale across society.
▶ 0:25:22While so much of the public debate around ai focuses on how this technology might evolve in the future, ai is here now and it's integrated into our lives and delivering benefits for millions of americans.
▶ 0:25:38We see these technologies in action at ces, where innovators come together from ai powered health insights, from abbott and withings to john deere's autonomous tractor to oshkosh and waymo's collision avoidance and autonomous technologies and even demons digital twin platform for manufacturing, these products are already in the market and making amazing changes. Still, we're just scratching the surface of what I can do today.
▶ 0:26:06We see ai and digital twins that can simulate everything from factories to city planning. Agentic ai, which are autonomous systems that can manage everyday tasks. Vertical ai models, which are specialized in areas like healthcare and mobility or agriculture. Industrial ai, which is augmenting the workforce and improving safety. And physical ai, which includes more lifelike and useful robots.
▶ 0:26:37American companies are leading the ai race, but their success isn't guaranteed. In china, the government has made ai central to its national strategy and invested heavily in areas like semiconductors, robots and data centers. To counter this strategy, we need policies that help american companies outinnovate the competition. If america falters in ai, we risk ceding entire industries, supply chains, and influence over global standards.
▶ 0:27:05That's why cta has urged congress to adopt a ten year pause on enforcement of state and local ai laws. In 2025 alone, legislators across all 50 states introduced more than a thousand, often conflicting, ai related bills. For a startup or a small business, navigating this patchwork is crippling. A pause gives congress the time it needs to develop a preemptive federal framework for ai.
▶ 0:27:34The administration's recently released ai action plan is a powerful and positive blueprint, ensuring american ai innovators have the guardrails they need to build, grow, and compete.
▶ 0:27:49We also need a comprehensive federal privacy law to power up innovation with more clarity and protect consumers, and lower compliance costs for industries that rely on responsible data use and give americans confidence in these life changing technologies. Congress must also recognize where our laws and frameworks are working. The law is clear. Simply reading or processing content does not constitute infringement.
▶ 0:28:14This clarity is a huge competitive advantage for america, and it's the foundation that allows us companies, from the smallest startup to the largest global brand, to win the ai race. America has led every major technological wave from electricity to the internet. And if we get ai policy right, this technology will be the next great american growth engine.
▶ 0:28:39If we get it wrong through fragmented or restrictive regulation like the eu's ai act, we risk exporting those jobs and that leadership overseas. Cta believes the path is clear. Foster innovation, protect consumers and ensure america sets the rules of the road for ai. I look forward to working with this committee on a bipartisan basis to shape our ai future. Thank you and I look forward to your questions. >> great. I now recognize, Mr.
▶ 0:29:10Hammond. To begin, please begin his opening statement. >> chairman. Ranking member brown, members of the committee, thank you for the opportunity to testify today. My name is samuel hammond. I'm the chief economist for the foundation for american innovation. We're a group of technologists and policy experts focused on developing technology, talent and ideas to support a freer and more abundant future. The capabilities of frontier ai systems are improving at a stunning rate. Five years ago, large language models could barely generate coherent english text.
▶ 0:29:35Today, they can hold forth on any topic reason through phd level math problems, encode entire applications from scratch. Recent ai progress, including the rise of reasoning models and ai agents, has been largely driven by breakthroughs in reinforcement learning applied to llms. Language models gain the raw intelligence by predicting sequences of text, but with reinforcement learning can be trained to follow instructions, use tools, and pursue complex goals. Scalable reinforcement learning for language models to give them reasoning.
▶ 0:30:02Goal directed behavior was only unveiled a year ago, but is already driving rapid improvements in domains like math and programing. The scope and significance of this breakthrough is still not fully appreciated, though in principle, these techniques can be used to create superhuman ai agents in any domain where success can be objectively benchmarked. Math and software engineering are just the low hanging fruit.
▶ 0:30:22The ai research organization meter carefully measures progress in ai autonomy, and has found that the length of tasks that agents can perform doubles roughly every 4 to 7 months, a trend that has held for the past six years. Will the earliest chatbots could only perform tasks measured in seconds or minutes. Openai's latest model, gpt five, can coherently execute tasks that take human engineers two hours and 17 minutes on average.
▶ 0:30:43If this trend continues, we are only two doublings away roughly 8 to 14 months from ai agents that can autonomously, autonomously perform tasks that take humans a full eight hour workday. Progress in non-verifiable and open ended domains is also accelerating, recently, leading to some of the first major examples of ais that have made novel scientific and mathematical discoveries. It is now plausible that we will have the first superintelligent ai scientists and mathematicians by the year's end. Pretending a dramatic speed up in the pace of r&d going forward.
▶ 0:31:11This includes ai's optimized for ai research itself, creating the glimmers of a self-improving feedback loop whereby the ai is rapidly helped build their own successors. What happens when ais get better ai r&d than the best human researchers in the world? At a minimum, we should expect a discontinuous leap in the power and efficiency of the frontier models. But where this process tops out is still a matter of significant uncertainty. It is possible that even with fully automated ai, r&d, progress will remain bottlenecked by the availability of compute data and energy.
▶ 0:31:41It's also possible that we're only 1 or 2 major breakthroughs away from systems that can learn continuously in a unbounded fashion. Regardless, the jump in capabilities unlocked by recursively self-improving ai is likely to be profound. Even within the bounds of existing infrastructure, and is coming sooner than many realize. It's worth emphasizing that creating agi and superintelligent ai is capable of outperforming humans in every domain is the explicit goal of every leading us ai company.
▶ 0:32:06While some dismiss this as science fiction or marketing hype, I assure you the leaders of these companies are deadly serious. As for timing, anthropic co-founder jack clark testified recently that he expects transformative ai to arrive as soon as the end of 2026 or early 2027. Even if these forecasts are on trend, ai capabilities will remain uneven for at least several more years. For a brief, paradoxical moment, we will have superintelligent ais that can prove new math theorems, but still struggle to do many things that humans find trivial.
▶ 0:32:33This is especially true in areas like robotics, which, despite remarkable progress, are still many years away from outperforming humans in every physical domain. Given the paucity of high quality training data. So as we run headlong into this new world, I see four major takeaways for national policy makers. First, monitoring frontier ai capabilities in real time should be a national security imperative of the us government. Early and differential access to developments of the frontier can provide policymakers, national security advisers, with the foresight into the capabilities that are coming down the pike, giving us time to prepare and adapt.
▶ 0:33:02Second, as ai systems become human level and beyond, geopolitical power will be increasingly proxied by the global distribution of computing resources. America's existing lead is downstream of our massive advantages in ai, hardware and data centers, but this is tenuous at best. With china outbuilding us on new energy, we must double down on semiconductor export controls or risk being leapfrogged.
▶ 0:33:23Third, we must quickly advance the frontier of ai control interpretability, review our laws and regulations for the compatibility with powerful ai, and invest in much more robust cyber and infrastructure security. All priorities outlined in president trump's ai action plan. Fourth, and finally, we must open our minds to radically new forms of institutions and structures of government from the printing press. The industrial revolution. Every major technological transition has driven equally transformative changes to our system of government.
▶ 0:33:49I believe the ai revolution will be no different, but it raises unique challenges given ai's use cases for surveillance and censorship, as seen in china's model of a digital panopticon. Reconciling the advent of powerful ai systems with america's tradition of individual liberty and limited government is thus the challenge of our time. Thank you and I look forward to your questions. >> thank you. And now I'd like to recognize doctor turner li for five minutes. >> thank you. Thank you, chairwoman mace, ranking member brown and distinguished members of the committee for this invitation to testify.
▶ 0:34:18My research focuses on policies that govern ai digital divide as well as innovation. Artificial intelligence is not the future. It is here today. Several workplaces require use of ai by workers, and one report shows that 92% of companies have plans to increase their investment in the technology. In just about every sector, companies are figuring out the role of ai, enhancing productivity, as well as the most appropriate investments in talent.
▶ 0:34:42Beyond enterprise use cases, ai is transforming the delivery of critical services and domains such as government services, healthcare and education. And with so much on the horizon, it's imperative that we strike the correct balance of innovation and regulation. We must safeguard consumers, institutions and critical infrastructure from ai risk, including workplace displacement bias and discrimination and the irreparable harm of machines.
▶ 0:35:06My testimony today, I just want to offer three points that I think congress should take, which will be critical of the future of ai, critical to the future of ai, the need for responsible and ethical frameworks at ai, designing governance, the importance of a ready and agile talent pipeline and workforce, and the importance of monitoring the unknowns in ai to ensure its safety and security while addressing the clear and present challenges. Ai depends on responsible and ethically designed models and national governance. The stakes are too high.
▶ 0:35:33The rapid advancements of generative ai in video, text, and voice extraction have contributed to consumer fraud. Our nation's seniors are increasingly being targeted, in some cases falling for financial ai, voice cloning, and deepfake scans that ask them to send their money to relatives. Being responsible requires national governance, and we have laws, existing laws for highly regulated industries protects consumers and digital spaces, including the fair housing act, the equal credit opportunity act, among others.
▶ 0:36:00But without congressional resolution and other critical policy areas like data privacy, everyday americans will be exploited by malicious uses of ai systems. We need clear measures that ensure human oversight, disclosures and independent audits over automated on autonomous decisions. We've started this process with the ai action plan this past summer, but our focus on promoting U.S. leadership against china as the prime goal will only allow these consumer protection goals to fall to the wayside. Yes.
▶ 0:36:26Recently, states in the absence of federal legislation have actually moved forward with their own legislation. And since january, over 100 measures across 38 states have been enacted into law. Multiple state attorney generals have also issued guidance on how to apply these laws to ai.
▶ 0:36:42The rejected, proposed ten year moratorium of states would have threatened states rights, and the public interest and leadership needs to continue to protect the independence of federal agencies so they can serve as a bulwark against deceptive and unfair consumer facing ai applications. Cuts to these agencies weaken our ability to hold bad actors accountable, but they also undermine consumer trust. Second, the future of ai depends on a ready and agile talent pipeline and workforce.
▶ 0:37:09Maintaining our edge means that we have robust talent that incorporates diverse viewpoints into design, development, and deployment of ai. Immigrants are central to the story of innovation. 77% of the top ai companies were founded or co-founded by first generation immigrants. Policies that restrict immigration may threaten our innovation capabilities, and defunding research at world class universities and scientific institutions may also do the same.
▶ 0:37:32Let me just share a few statistics of what's happening as a result of other countries exploiting these opportunities, spain opened its doors to students subject to us restrictions. European universities offering scientific asylum to scientists. China is using the reverse brain drain to aggressively recruit our top technological talent.
▶ 0:37:52Cultivating talent at that level has impacts, as well as ensuring that researchers here in the states have what they need to get the research done for the next big idea, and we cannot neglect our domestic talent pipeline. Recent declines in math and reading scores are warning signs that much more needs to be done to cultivate homegrown talent, not only through national apprenticeships, but also realigning our core objectives in schools so that we meet the demands of the future workforce.
▶ 0:38:17Casualties will abound in the workplace if we have a less agile, ready workforce capable of shifting gears. And finally, I'll just say this though the future of ai is largely unknown. We need to solve the first point I made in the second point, to ensure that we actually get to a place where we understand the power of artificial intelligence or artificial general intelligence agentic ai, as well as what is largely unknown at this point.
▶ 0:38:43However, people like me are far more skeptical of agi and generally the existential threats for the foreseeable future, simply because we just don't know enough. And it's important for us to have those guardrails in place so that we ensure the rights and safety of all americans. Central to that development, and we not abandon consumer protection in the rush to just innovate. So I will close here and implore congress to continue to think about policies that allow us to grow a healthy ecosystem where consumers are centered, and our economy and our trust in these products are also prioritized.
▶ 0:39:15Thank you again to the members of the committee. I look forward to working with you and taking your questions. >> and thank you, ma'am, and I will now recognize myself for five minutes and for questioning. Uh, miss fabrizio, thank you for being here today. What has surprised you the most about ai? >> um, at ces, I think the most surprising thing are all the different ways that ai is solving the world's biggest challenge. The healthcare applications are the most exciting to me personally.
▶ 0:39:40I saw a digital twin of a heart at ces, which is used to train surgeons so that they can understand how to safely do heart surgery. That's a huge impact and very amazing. >> um, and then you talked about, uh, the ten year moratorium and for states, why is it so important, you know, the states, the states rights thing has to be balanced. But also we don't want to stifle innovation. We know china, russia, iran. They're not they don't have any guardrails. They don't care. Um, talk about that a little bit.
▶ 0:40:10>> yeah. Well, you said it. China doesn't have that. And it's it's impossible for our member companies. Like I said, we have 80% small businesses and they cannot compete and understand when there are a thousand different potential laws that they have to comply to. It just stifles innovation completely. And for us to win the ai race, we need to remove that barrier. >> and it is a federal issue because it's commerce across state lines. And having all those, uh, that regulatory environment that's a patchwork. And every state does make it very difficult to operate. And then, Mr.
▶ 0:40:40Hammond, one thing that struck me in your testimony, you're talking about compute, uh, energy. Um, talk to me a little bit more about that. Let's go into detail, because I agree with you. Uh, it's a huge problem. How do we solve it? >> uh, it's a great question. So, you know, there's only a handful of inputs that go into training and competing at the frontier of these models. There's the data, the human talent, the compute and the energy. Um, with china, we're basically at parity with with talent, with data that may have advantages because they don't have privacy laws. They can build.
▶ 0:41:10>> on a bunch of our data. >> of course, and they also steal data and ip. Yeah. And so really it comes down to hardware and energy. China has added over 400gw to the grid last year. They're about to do the same thing this year. So our only real advantage. >> how much have we added to our grid? >> uh, approximately zero. I mean, we've removed coal and added renewables and that has canceled out. Um, and so what that means is in lieu. But for these export controls that are barring china from our most advanced hardware, they would surely leapfrog us within a matter of years. >> mhm. And you think nuclear is the way.
▶ 0:41:41>> uh, to support nuclear? I think the earliest that we'll see new reactors come online is in the 2030s. >> uh, you know, it's frustrating because we see small nuclear reactors or smrs in japan and in france, and we don't have them here like we have them in our nuclear subs. Like, I just can we just this is this is I'm it's a joke, but it's like, why can't we just plug one into an outlet? I mean, I just we have the technology here. Why aren't we using it particularly with our data centers, and allow them to grow, um, as data center technology, as data centers grow as well.
▶ 0:42:11But, um, I do want to talk to you about, um, the future of ai. Elon musk has said that as early as 2026, um, we would have singularity, basically. So define singularity. And how quickly do you think we're going to get there? >> so I think there's two two ways to think of singularity. One is as a metaphor for our ability to predict the future. Um, so whatever the technology is, singularity is a point in time where it becomes impossible to look beyond that point.
▶ 0:42:37And then there's the technological singularity, which is when we have eyes that can build their own successors and then potentially go off to infinity. And we don't really know what comes out of that process. Uh, I think 2026 is a pretty aggressive, uh, expectation. But I think something like that crossing that threshold will happen this decade. >> and I and I've talked to some folks in the tech space that say a thousand days or 2000 days, what what is your what do you think it'll take to get there where ai is creating its own ai?
▶ 0:43:04>> so, you know, I got into this topic as a young kid reading ray kurzweil, uh, 1999 book where he predicted we'd have human level ai by 2029. That was a 30 year forecast. And actually, uh, current trend lines suggest that he was dead on 30 years ago. Um, and so I tend to lean towards that as a date, but I think this won't look like some threshold that we pass, and it looks completely different. I think we're on this sort of continuous exponential.
▶ 0:43:30And the thing about exponentials is they look flat, looking backwards and vertical going forwards. >> what's the biggest concern with when we do get there? When we do hit that milestone. >> uh, that we lose control in some manner, whether literally over the ai's themselves or that the proliferation because, well, it does cost billions of dollars to train these models at first. Subsequent generations, the cost comes down orders of magnitude, that there will be a mass proliferation of powerful capabilities, that that our institutions are just not capable of adapting to.
▶ 0:43:58>> that could hack every system, every grid, everywhere, all at the same time, all around the world, essentially. Potentially. >> yeah. Essentially. I see it. You know, I think there's going to be attacks on critical infrastructure. But there'll also be, you know, the the high school student that hacks their their school's it system. And the system admin happens to be the gym teacher. Right. >> and our personal medical records or, you know, whatever the case may be. Yeah, that's one of my greatest concerns on the cyber side. I've run out of time. I could have, uh, I could talk all day. Maybe we'll have some more time later.
▶ 0:44:27So I will yield back to the ranking member and recognize her for five minutes. >> thank you. Um, thank you, madam chair. Artificial intelligence has the potential to be in every aspect of the workforce, in every corner of our daily life. This means that the federal government and private sector must collaborate to adapt to an ai present and future.
▶ 0:44:46Algorithms and automation are not inherently harmful, but the way in which they are developed and deployed has the potential to have profound consequences for american workers, especially those from diverse backgrounds. Black and brown communities have long carried the weight of the wage gap in this country, and we cannot allow ai to deepen those inequities, whether through bias algorithms in hiring or automation that displaces jobs, we cannot afford to be caught flat footed
▶ 0:45:17Or let ai run unchecked. So, doctor turner, lee, can you speak of the work that ai equity lab and brookings is doing to ensure that ai does not worsen historical inequities and the importance of acting proactively? >> thank you, ranking member. So what I actually figured out as a sociologist is that we're not a lot of sociologists sitting at the table or people of various backgrounds when it came to thinking about the outcomes of these models.
▶ 0:45:43And so, running on the back of a napkin, I thought about an experiment to actually bring interdisciplinary experts, people from various backgrounds, various industries together to think about areas in which we're going to have the most high risk and consequential outcomes, particularly with marginalized communities.
▶ 0:46:00To your point, not only do some ai models come with a series of bias in the training data, where it's actually picking up information that may be false, inaccurate, or unrepresentative, the outcomes of that data can contribute to a widening wealth gap. When that algorithm suggests that I'm not creditworthy, I cannot get a home loan. In essence, my quality of life is actually impacted.
▶ 0:46:23So I think one of the best ways to start with this is to widen the seats at the table and to ensure that we have scientists alongside social scientists, alongside industry sector, alongside people with various backgrounds who have concerns for the lived experiences of populations that you spoke about, especially those that are black and brown communities. >> thank you. And as more companies adopt ai technology, we must accept that the future of work is changing.
▶ 0:46:49Doctor turner lee, what steps can the federal government take to ensure that american workforce is prepared to succeed in the ai future? And what legislative steps can congress take to ensure that there are adequate guardrails overseeing ai adoption throughout the economy and society, while also encouraging ai innovation? >> well, first and foremost, I'd like to commend the work that you've done in a bipartisan manner on your act, which is around training the federal workforce to be exposed to.
▶ 0:47:18I think that's the first step. We have to be very transparent again, that the train has left the station. Ai is not only dictating how they do work, but it's also managing their productivity in how they're processing that work as well. So it may help them with research, but it's also managing the time that it takes to do the research. Being transparent with the federal government, I think, is one way to actually help disclose the use of ai. They're also making available all types of data. We have some concerns, uh, given the recent scrutiny of a lot of information from federal data sets, etc.
▶ 0:47:47That there just won't be the quality of data and the integrity of data that we need. So just making sure the federal government stays on point with that and does the appropriate audits of the data that is available, I would just suggest in terms of guardrails, uh, there's enough ai for everybody to eat.
▶ 0:48:03It's a statement that I've been making lately, and that means that wherever you are in the workforce, you are going in some way touch this improving upon our literacy, our upskilling, our ability to mentor people who may be from different generations, where I was not necessarily something that they ever thought of. I used to watch the jetsons. I never thought that I would come to fruition. I think our ranking member, those are the steps that I think the us government can put in place. That really, in the long run, promote transparency, disclosure and effective use.
▶ 0:48:34>> thank you very much. Um, I'll close with this. Our future is one where ai technology will no doubt impact our everyday lives, which is why we must carefully consider ai development. Now we need federal legislation that protects americans rights and freedoms while preventing bias and injustice. I look forward to continuing to explore this topic with experts, so that we can ensure safe and responsible innovation. And with that, madam chair, I yield back. >> thank you. I'll now recognize Mr.
▶ 0:49:04Mcguire for five minutes. >> thank you, madam chair, and thank you to our witnesses for being here today. Uh, for the first few questions, if you would. Let's answer as quickly as possible. Um. Let's see. Um, Mr. hammond, you said you think singularity, based on your experience would be 20, 26. And just real quick, doctor lee, when do you think we'll have that? Um, real simple. >> I'm a little less optimistic. I think it's going to take longer, because I still think it's a little bit more hypothetical in its framing. >> all right, miss fabrizio. >> I feel the same way.
▶ 0:49:30I think right now, human in the loop is still important. And we're seeing ai augment what humans can do. >> all right. This is just a yes or no. Um, do you do you believe china is using ai to manipulate their people, or do they have plans for that? Yes or no? Miss fabrizio. >> I think china is using ai in ways that we are not, and that is why it's important for us to continue to focus on winning the ai race. With the issues I laid out earlier. >> Mr. hammond. >> uh, yes. Absolutely. In some cases, using us technologies provided by us companies.
▶ 0:49:59>> um, I do agree that china, the chinese government, has a highly surveilled state and they are using ai, uh, not to the protection of their citizens. >> but again, just yes or no, because we're running out of time. Yes or no, the stakes are very high for ai development in the us. >> yes. >> extremely. >> yes. >> so very important that we have the best workforce possible to win this battle. Yes or no? >> yes. Yes. >> uh, yes. In the short run, I wonder what the workforce will look like when ai's can do everything. >> so, yeah, I'm with you.
▶ 0:50:27So we're developing this workforce, and I gotta tell you, as a navy seal veteran, if somebody say, my life on the battlefield, I don't care if they're pink or blue, male or female, democrat, republican, independent. We're all human beings. But I believe that the decisions that we make should be color blind because we need the best for us. And I don't care if you're pink or blue. Do you agree that it should be colorblind? We should have the most qualified people to win this race. Mr. forbes, yes or no? >> we need the best and brightest colorblind. >> yes. Not necessarily nationality blind.
▶ 0:50:56One of the challenges here is some of the most sensitive technologies are being developed by foreign nationals, including chinese nationals. >> but again, if they're the best of the best. >> or the best of the best, but we have to compartmentalize. In some cases. >> I think we should have the best of the best. But I also think that we need to have the doors open for people who are the best of the best in all of our communities to actually participate. >> okay, miss hammond, I've got a lot to learn, but I'm listening and I'm learning. And I like what you talked about the digital twin and practicing the heart surgeries.
▶ 0:51:25There's so much more I want to learn. Uh, but in your testimony, you said even if we were to reach a.I. Singularity, we might not have enough energy to keep going. So this question is for all of the witnesses, and hopefully it's an easy one. Let's keep it real simple, because we don't have time. Um, what? Unleashing american energy. Give us a better chance of winning the ai race. I'm talking the ability to drill nuclear. All the above coal plants. >> we need to modernize our energy grid. It cannot handle what is in store with ai. >> maybe the single most important factor. Yes. >> single. Most.
▶ 0:51:53>> I do agree that we need to do more to upgrade the energy grid, particularly if we're actually building data centers. But I want us to be cautious about the environmental consequences of actually moving too fast in communities where we already know we have a disadvantage. >> and personally, I wouldn't care about that. I want to win. All right, miss fabrizio, I have to say I agree with your testimony. Um, overly strict regulations can stifle innovation. The first thing I think about the so-called green new deal, better known as the green new scam. The biden administration spent hundreds of billions of dollars on these green energy projects like solar panels and windmills.
▶ 0:52:23Yesterday, I asked an ai chat box, how many acres of solar panels would you need to power ai in the us by 2030? Uh, anyone take a guess how much that would be? Uh, a thousand acres? Actually, let me see here. It's way more than a thousand acres. It's 500,000 acres. That's half the size of delaware. We should be investing in fossil fuels and nuclear. Small modular nuclear reactors. As we discussed earlier, we only need 500 acres to do the same job nationally.
▶ 0:52:54By 2030, we should be using coal, natural gas, traditional nuclear power until smrs are ready, not solar panels. All right. So let me ask this question. Is china building thousands of solar farms to power their ai? Yes or no? >> china is looking at energy in different ways than we are, but there are solutions that we can look at to to modernize our energy grid. And I will help. I will help develop solutions and help us be smarter about the future. >> Mr. hammond? >> uh, both. They added over 30 new coal plants while also adding 300gw of renewable.
▶ 0:53:24>> so today's technology, what's more effective? Solar or fossil fuels? Just keep it simple because I'm running out of time. >> I that I would have to get back to you on that. >> uh, I support all the above, but in the short run, these data centers are only going to go up with natural gas. >> um, I can't answer the particulars. I can get back to you that for my team. But I want to go back, congressman, to your question. >> I'm very limited on time. No problem. All right. So let's see, uh, let's see Mr.
▶ 0:53:52Hammond and miss fabrizio, what are some of the ways ai superintelligence might actually help us solve the energy problem? >> well, when it comes to research and data and looking for solutions, ai is faster and can help predict different models and find different solutions where we may not be able to find them on our own. >> I believe we're a year or two away from having fully autonomous ai labs that could discover new materials, new energy sources, all the above. >> we're out of time, I yield back. Thank you. >> thank you. I'll now recognize Mr. subramaniam for five minutes. >> uh, thank you, madam chair.
▶ 0:54:21Thank you to the witnesses for coming today. Um, I wanted to talk about a couple things. We've had several hearings on capitol hill about ai, uh, in recent months and since I've been here. Certainly. And, um, one thing that hasn't been talked about as much, uh, is, is job displacement as a result of ai. And I think, you know, I served in the obama white house as a technology policy advisor, and we were talking about this, but it was a little bit theoretical.
▶ 0:54:49I mean, there was job displacement happening because of emerging technologies, certainly making jobs easier, but automating some jobs and some tasks. But now we're seeing it at a different level, and we're seeing companies now basically lay off entire departments and replace them with ai. And they are saying this publicly and they're saying this, um, you know, very um, they're not hiding it anymore. Right? And it's their prerogative.
▶ 0:55:14I mean, we're not here to tell companies how to run their business, but it's creating a problem, uh, that congress has to figure out how to address, which is jobs, because one thing we want is jobs available for people. And we've been telling people, we've been telling kids for the past 10 or 15 years, go into stem, right? Learn to code like that's going to be your meal ticket. You can have a job for 30 years being a coder, an engineer, and now we have ai that can do their job.
▶ 0:55:44And I talk to a lot of kids who got their it degrees and cybersecurity or different types of technology, and now they're having trouble finding a job in this market. And so I'd love to hear maybe 30s each witness, um, your thoughts on what congress can do about it, whether we can do anything about it at all. Um, whether it's a fixable problem right now. Uh, miss fabrizio. >> thank you for the question.
▶ 0:56:10I think the important thing to continue to focus on is investment in stem education for ai, and investment in reskilling and upskilling the existing workforce. Apprenticeships will help there. Um, the white house ai and education pledge, uh, is looking at these areas and cta was happy to sign on to that. I'll also say that while the workforce will shift and workers will be given new tools if they use ai properly and they'll be able to take on more capacity, be more efficient and work smarter. >> Mr. hammond.
▶ 0:56:41>> uh, I'm relatively optimistic about the jobs picture over the over the short run, um, in particular, because I think the people who are going to be most displaced are often white collar workers who are more adaptable. However, uh, just look at wages for electricians. They've they've spiked dramatically. We we have a shortfall in hvac and cooling, construction, all these things that are going in as inputs for data centers, that could be a major source of job growth. And more generally, I think we need to deregulate aspects of the labor market to make transitions easier. >> what do you mean by deregulate aspects of the labor market?
▶ 0:57:11>> things like occupational licensing. Uh, what kind of accreditations do you need? >> okay, miss turner, doctor turley. >> yeah, I would say this on the education side, and I do agree with you, congressman, that we spend so much time in stem and computer science. And, you know, since those efforts were actually, uh, made, we've evolved and sort of retracted on those investments. I think we still need to use ai to augment education.
▶ 0:57:30We often put ai in the classroom instead of talking about the education of ai for students, which I think is somewhat of a challenge for many teachers and educators, I think it's important to ensure that there's equitable distribution of resources that actually train students on ai literacy, so that they're actually prepared to do more of a k through 20 shift as opposed to teaching it in early education and then teaching it in college. And different levels would provide more consistency and more opportunities.
▶ 0:57:57And I think that companies need to do a better job of qualifying what jobs are going to be lost. I think it's still an unknown number of where jobs or companies are going to be affected by ai, based on the decision of what departments they choose to absorb it. And then I also want to just respond to the data set aside. I think we need to be careful in thinking that data centers will generate post-construction jobs, and really focus on how many jobs will actually be created as a result of the data center ecosystem versus what's going to go. What are going to be the job creation numbers going into its construction?
▶ 0:58:28>> do you think I will create more jobs than it's displacing, yes or no, miss fabrizio? >> yes. I think whole new industries will be developed because of ai, and a tremendous amount of new jobs in the workforce will be created. >> Mr. hammond. >> when we have fully ai automated software engineers, it's less that we lose that job category and more that we all become software engineers. And I think that will be a general pattern where we're all empowered as individuals to take on these new roles. >> doctor turley.
▶ 0:58:55>> well, I think if we believe that agi is coming quite quickly, that they will take on more of the jobs that people because of the superintelligence. I honestly think, congressman, that ai will change the nature of jobs. And that's a conversation that we need to be having as opposed to job loss. >> I have more questions, but my time is up. We might have a second round, but I'll yield back. Thank you. >> all right. I'm now going to recognize Mr. burleson for five minutes. >> thank you, madam chair. Thank you for conducting this hearing.
▶ 0:59:25This is one of my favorite topics. Um I want to I want to begin by quoting irving john good, who was a british cryptologist or mathematician, uh, famous for working with alan turing. He's quoted as saying, let an ultra intelligent machine be defined as a machine that can far surpass all the intellectual activities of any man, however clever.
▶ 0:59:46Since the design of machines is one of these intellectual activities, and ultra intelligent machine could design even better machines, there would then be unquestionably an intelligence explosion, and the intelligence of man would be left far behind. Thus, the first ultra intelligent machine is the last invention that man may ever need to make. Um, Mr.
▶ 1:00:08Hammond, can you elaborate on some of the some of the success recently of some of the ai self-improvement that's occurring where it's it's constantly improving itself? >> yeah, absolutely. I think it's a coming in in gradations. So we already have eyes that are good at coding. And a lot of the job of an ai engineer is coding. And so, you know, there's a joke now that these ai labs, they're no longer coding, they're just kicking the ai to to fix the bugs.
▶ 1:00:37Uh, beyond that, we also have ais are now writing their own algorithms. So earlier this year, google released alpha evolve, which was an evolutionary algorithm that discovered new bounds on, uh, on a mathematical theorem that hadn't been beaten in 47 years. >> yeah. And with that, madam chair, I've got an article about alpha evolve ai powered coding agent for designing advanced algorithms. If I could submit that for the record. >> and so ordered. >> thank you. Um.
▶ 1:01:04When miss doctor. Doctor lee. When when it comes to jobs, I'm of the opinion and I think you kind of touched on it that things will change. It doesn't mean there won't be any. There will no longer be the need for us to have work to do. Um, I think of it often this way. My great great grandparents, your great great grandparents, probably everyone in this room's great great grandparents were all farmers.
▶ 1:01:34Because that's what it took. It took everyone working the fields in order to produce enough food to feed people. Today, very few people farm. It's because of machinery and no one would go back in time. Would agree that we should go back in time and say, don't let them have the tractor or the harvester, right? If if anything, it's taken the power or the productivity of one person and magnified it. Manyfold.
▶ 1:02:03And that's the way I think we should think about ai. In fact, an ai will be a magnifier for productivity for any individual, no matter what they do, and thus you need the individual. You need that individual. That's the core, the beginning. Would you agree with that, doctor lee? >> oh, I definitely agree. I published a book last year called digitally invisible, where I went to talk to farmers in places in southern maryland and across the country. And guess what?
▶ 1:02:27They said a tractor is only as good as the broadband, that it has to actually be more productive in the work that they do. But most importantly, with the compute power of ai, it'll only be as good as the facilities that we offer them to be able to be connected to these new resources. And it'll just change and transform how they do their work in ways where they don't have to go out there and measure how much rain came. They'll actually know from the comfort of their office that's sitting on their land.
▶ 1:02:51So I do agree that we need to have conversations about job loss, as well as the transformational capacities in which jobs will change the way in which we work. I don't agree that we'll have robots bossing us around yet. >> I don't either, I don't either. Um, miss fabrizio, the european union has. They reportedly their their regulations have kind of created a chilling effect. Can you elaborate on that? >> yes. It's it's been hard for companies to to innovate.
▶ 1:03:19And that's why you see fewer unicorns and fewer tech innovations out of europe. I think the us has is doing it right. And we have many companies here. We have the most robust startup ecosystem in the world here. And we see that firsthand at rs 1400 startups and many of them creating new ai innovations and launching them at rs. That's why it's really important to make sure that we have a framework that supports them.
▶ 1:03:46>> and I think we're at a place now where we just realize if we're going to be, if we're going to stay on top competitively, we have to be the location for these data centers. We have to have these ai, you know, housed in the united states. In order to do that, we need we need electricity. And, Mr. hammond, would you agree that right now we are at a we're at an electricity crunch. We don't have the what's what's what's needed for that demand. We've got to we've got to change that. >> absolutely.
▶ 1:04:14Not just the data centers, but all all forms of things are being electrified from from vehicles on down. >> thank you. My time is expired. >> okay. I'll now recognize Mr. crane for his five minutes. >> thank you, Mr. chairwoman. Uh, thank you guys for coming today. I want to start with, uh, miss fabrizio. I met with organizations tied to blue collar jobs all the time, and they are constantly in my office begging for more help.
▶ 1:04:41And people that need training in specialized fields like plumbing, electrical work and carpentry. Meanwhile, ceos are warning that I will take 30% of the workforce by 2045. My questions are how do we ensure we are sending more people to get to go to trade schools for these blue collar jobs, and not just allowing them to get laid off by ai? >> we need to continue to invest in reskilling and upskilling, and we need to make sure that we have the resources available at rs.
▶ 1:05:10We're doing ai trainings for the first time at ces 2026. For people who are in the industry and want to learn more, this is an important solution to this challenge. >> do you agree with those assessments? Um, that I will take 30% of the workforce by 2045. >> is going to change the workforce, but it's going to make it better. It's going to create new jobs, and it's going to augment the existing jobs. It's going to make people smarter. It's going to make them more efficient and give them tools that they never had before to do their work.
▶ 1:05:41>> yeah. Thank you. And I'd just like to say make a statement for the young generation out there watching what's going on with ai and who may be in college or maybe in high school. You know, I think one of the fields that's the least susceptible to ai taking over and eliminating your career opportunities are the trades. Um, what advancements right now are we seeing with technology that are creating new job opportunities for americans? >> we're seeing it in many areas in healthcare.
▶ 1:06:09You know, we're seeing, um, individuals that are learning how to create autonomous, um, healthcare monitoring systems for first line intake that will help nurses and doctors get better information so that they can be physically with patients that they need to be with. While a autonomous system is collecting information on patients. And that would be hugely helpful given the health care crisis we have. >> thank you. I want to shift to, uh, Mr.
▶ 1:06:35Hammond, last week, we lost a true american patriot, charlie kirk, following his assassination. There were reports of chinese and russian bots encouraging violence through spreading misinformation, attempting to create division. Did you see any of those reports, sir. >> I did not. >> okay, how should the american people be weary of the increase of inflammatory speech following the assassination of charlie?
▶ 1:07:02Um, and other mass violent events from chinese and russian bots? >> it is a really big open challenge. Uh, these social media platforms, um, have their work cut out for them. Uh, we do not yet have a reliable means of identifying what is a bot, what isn't. Especially as these systems become more and more human like in the way they speak. And so I think it's something we need to put much more resources into. >> okay.
▶ 1:07:25What what do you think federal agencies should be doing to prevent the spreading of misinformation by these bots to sow discord in our communities, in our country? >> I mean, at a minimum, we should stop selling china and russia the technology they used to run those bots. Um, you know, we these h20 chips which just got approved or liable to be approved for export to china if they're if they all go through, it's going to roughly double their data center capacity for running advanced ai models.
▶ 1:07:53We know from the past that they've used these chips to power their surveillance drones, to power their gait recognition technology. Um, so we've given them the ammunition that they're using on us. >> okay. Next question for you, sir. When I was growing up in school, it was often considered cheating to use a calculator on a test. Right now, we have ceos of fortune 500 companies basically telling their employees that they need to be using ai a few times a day. Um, or they'll be falling behind or become obsolete.
▶ 1:08:23My question is, how do we balance the expanded use of ai and not demonize the use of ai while preparing our students for the future? >> I think education is a good example of how ai is going to force a massive rethink and reckoning in how we do a lot of things, including how we design curriculum for k-12. And, you know, there's going to be resistance. Um, but there's already new models that are emerging.
▶ 1:08:45So there's alpha school in austin, which is trialing, running, uh, you know, ai assisted tutoring in the mornings and project based learning in the afternoons and seeing tremendous results. And so I think we just need much more innovation in how we how we do education. >> what advice for this committee and for congress do you have in regards to any regulations that you think are responsible regarding ai in the future? >> uh, my my three big, uh, bullet points are one, we need to monitor the frontier.
▶ 1:09:13So we need to know what's coming and be able to prepare and adapt, because it's going to be a very fast moving period of human history. Um, so, you know, we don't want the government to be the last one to know. Uh, number two is we need to be investing in research and development, especially around issues like control and interpretability. How do we interpret how these models work? How do we understand their behavior? How do we control their behavior? Still the companies are under investing in that. And then third, we need to protect our our competitive advantage which is ai hardware. So as I mentioned earlier, um, our one big advantage is chips and hardware.
▶ 1:09:44China is trying to catch up, but they are cut off right now. And if we open up those chips to china, they're going to jump ahead. >> thank you I yield back. >> all right. Thank you. We're going to do a second round if that's okay with the witnesses I request unanimous consent that the subcommittee have a second round of questioning of the witnesses. So without objection it is so ordered. I want to pick up where Mr. crane left off about the chips. Uh, your referencing the h20 chip, right? Nvidia, basically.
▶ 1:10:11Um, I know china just said that they were. They were encouraging, uh, folks not to buy from nvidia, but it was a different chip. Right. Do you know anything about that? >> uh, yes. China's trying to indigenize their own chips made by huawei. Um, deep sea, uh, was reported to have had some, uh, botched efforts using the huawei chips. And so the chinese companies are all hungry for american chips. >> and then how do we I mean, how do we prevent china from getting the chips? >> uh, first orders.
▶ 1:10:38We should not approve the sale of these chips. Um, and to the extent that we do try to minimize the damage, senator jim banks has a bill called the gain act, which would give us companies a first right of refusal to to buy the chips that are destined for china. I think that's the least we could do. >> is there a house version of that bill? >> I don't believe so. Not yet. Okay. Um, and then, uh, you know, even with these controls in place, china has very sophisticated smuggling operations.
▶ 1:11:04Um, so not a month after the blackwell series of chips were were announced, there was a report that china had already smuggled in $1 billion worth of them. Um, and so we need to do much more to crack down on. >> how did they do that? How were they able to get away with that? >> uh, through third party intermediaries in malaysia, singapore, taiwan, they just, you know, they buy them through these third parties that aren't on the entity list. In some cases, they're even incorporated in the united states. And it's just a matter of getting them over the border.
▶ 1:11:32>> and are we cracking down on the intermediaries here in the us now? Have you heard any anything about that? >> uh, there's been some doj style investigations. There's been actions taken in malaysia and singapore have enlisted the governments in those in those countries in some cases that the challenge is really one of scale. So once these chips are at the door, how do we know where they're ending up? Um, there's a bill called the chip security act. There's a house version of that that, um, you know, would require these chips to have basic location verification. >> and who's doing that bill?
▶ 1:12:00>> uh, it's it's before, uh, hvac right now. >> okay. Yeah. Um, and then, uh, so, um, miss fabrizio, you said earlier in one of the questioning about china using ai differently. Can you give some examples? >> well, china is using ai for surveillance. They're using it for military purposes. They're, um, you know, they also don't have privacy in china. And so they use the data differently than than we do here.
▶ 1:12:28Um, that's why I think it's really important that we continue to move forward. Um, we, we look for a national ai framework that addresses some of these big issues. It's risk based, it's tech neutral, and it allows innovation to continue to flourish here so we can continue to beat china. >> um, is china our greatest threat? Yes. Do you agree, Mr. hammond? China being the greatest threat. >> certainly on the geopolitical stage. >> doctor lee, would you agree? >> yes.
▶ 1:12:56>> um, what are the consequences? Uh, and this is for all the panelists. If the us fails to outpace china in the race for domination in ai, miss fabrizio. >> well, it's important that we move forward in the best way that we can and that we focus on innovation. We have the best startup ecosystem here. We have the best tech companies here, but we need to not get in their way. So that means a pause on state legislation.
▶ 1:13:21It means a federal framework where companies can innovate and know the guardrails and be able to move forward. >> Mr. hammond. >> uh, I worry that we need to buy time with china, because over the last 40 years, this is part of the deindustrialization story. We've we've shifted all our industries into services, entertainment, law, finance. All things are about to be deflated by ai, whereas so so there's a world where we build the agi, we build the general intelligence. But china is the one that puts in factories and has the growth benefits.
▶ 1:13:52>> there are people because I know one personally. Um, someone I'm suing, literally, uh, he's using chatgpt as his attorney. I just can't think of any. I mean, okay, yeah, but we're going to beat the llm in court. But, uh, doctor lee. >> um, I would agree with my colleagues in terms of china being a threat to our dominance in the ai space. I also would put out there, though, that we have to find alternative markets if we're actually going to grow the economy and scale of us companies.
▶ 1:14:19So I'm thinking about my experiences with 5g, where we actually opened up other markets. We missed the opportunity to do work with the global majority, the african union. So just thinking, rethinking our industrial policy as we think about china as a threat as to if we want people to have american products embedded in their technology, where are we selling it to and making sure we're agile? I do also want to respond that I do think states have to play a role in experimentation.
▶ 1:14:45I think the federal government would be too premature to come up with a national policy that limits states rights, because what we are seeing in terms of experimentation of states is that they're looking at more consumer protection. Ag's are trying to figure out ways to keep our grandmothers safe from ai. They're not necessarily trying to compete against china. They're just competing against the various misnomers and lack of information that people. >> already have that like, I got a nigerian scam the other day in an email that, you know, I forwarded or took a screenshot and forward it to my family, I'm like, don't click on this.
▶ 1:15:15>> we do the same thing because we're not coming to congress. We're going to our state ags. So we can't debo like what they're able to do to make sure people are safe. >> yes. Okay. I will now recognize Mr. subramaniam for five minutes. >> uh, thank you, madam chair. I just want to finish the job displacement conversation. And, uh, Mr. hammond, you said something interesting about, uh, the us has shifted its economy toward industries that are susceptible to ai. You mentioned law. I'm a lawyer, too, and so I'm sensitive to that. Uh, but can you expound a little bit on that as well as. What should we be doing?
▶ 1:15:45I know Mr. crane made some good points as well about, you know, different industries that might be more important moving forward. But I'd love to hear more about that. >> it's it's a story of relative scarcity. So if intelligence becomes abundant, if service labor, cognitive labor becomes abundant, then what remains scarce? And it will be the heavy industry, the factories, the actual and not just the factories, but also the know how, the tacit knowledge that's embedded within the workforces that china has that we don't.
▶ 1:16:11And so, um, you know, one of the reasons that we need to buy time is in part to try to rebuild some of those new sectors, in part to, uh, to ensure that we don't just give away our economy. >> and what would buy time, like, what can we do to buy time? >> well, number one, denying china access to the most advanced ai. >> oh, I see okay. Got it. You mean like export controls and that sort of thing? And, um, I'd ask, I guess the other witnesses as well. Um, what your thoughts are on this? Like, what jobs are we going to lose in the future as well?
▶ 1:16:39What jobs are we going to be gained? As fabrizio mentioned, this is going to create more jobs than we're losing. Um, do we even know right now what those jobs would be? I know it's, you know, there's this idea of, okay, well, ai is going to displace this it department, but it's going to replace it with ai buddies that, you know, help fix the ai or make it better. But it you know, I look at what's actually happening and that's not quite the reality right now, but perhaps it will be in the future.
▶ 1:17:07But right now what I'm seeing is, um, you know, 40 people get laid off in an it department. They're replaced with ai and like five people. Um, and so I'm just curious what you think that future would look like, what jobs will be gained? >> well, every major shift we've seen, new industries and new jobs created. Think about the internet, things, you know, so many new industries were developed. And we see that at ces.
▶ 1:17:30Um, you know, in terms of jobs, there's ai data scientists, ai ethicists, there's new ways of building and manufacturing technology that didn't exist before. And we will continue to see more and more of that as new industries use ai and as new whole new industries that we don't even know what they are yet emerge. Um, but agriculture is one great example where we've seen new roles develop because of ai. >> uh, miss doctor turley. >> yes.
▶ 1:17:57And I would just say in sort of looping back to the previous question as well, I think we have to be distinguishing between the loss of jobs in the blue collar sector and the white collar sector. Right. So I think on the white collar sector, it's been very clear that ai is going to improve the efficiencies of lawyers or paralegals who do the type of research that ai can do expeditiously. On the blue collar side, it becomes a little bit more tricky because what we're talking about are trade jobs that are going to be lost.
▶ 1:18:23I'm not sure yet if a robotic plumber can come to my house and fix my toilet, but I'm not. We still have to see that we're going to be running these parallel workforce opportunities, and it's important for that plumber to have the skills necessary to be able to innovate and to grow into the new economy, where maybe they're not actually doing the physical labor, but they're managing schedule, or they're managing invoices, or they're trying to do trade service calls. I want to put that out there.
▶ 1:18:47I think the question that we all should have is we're going to lose collective bargaining jobs when ai comes in, because it is going to replace frontline workers who essentially have had that job security as well as the union to be able to do the work that they do. So I think we have to just look at the scan and do a better scan of what that means. When we say job loss, congressman, you're right on point and sort of divide that out based on not only productivity but where we're going to see the most vulnerability. >> and I'd ask all the witnesses, what should congress do, if anything, about job displacement? Mr. fabrizio.
▶ 1:19:18>> I think investing in in workforce development and training and apprenticeships, I mean, you're right, you might not have a robotic plumber, but you might have a plumber that's able to look at a digital twin of your home and identify a better solution faster that they wouldn't have had the ability to do before. So continuing to invest in upskilling is extremely important. >> Mr. hammond. >> I would say two things. One, that the the us government has a poor track record of running employment and training and retraining programs. They tend not to work very well. And I believe gao last reported that there's 46 of them already.
▶ 1:19:48Um, and so I would look to opportunities to shift training on the job as much as possible and to the extent possible, reducing barriers to entering new jobs so that you don't need to get that certificate or don't need to get that extra piece of education just to enter the workforce. >> and I would say, I would. >> say the same thing upskilling, obviously, cross skilling as the nature of jobs actually are transformed. I love the example of my colleague with regards to digital twins and some of the trade areas.
▶ 1:20:14And I also would say national ai literacy, so that people also understand that ai is not just the job, it's the behavior in which you approach the task that's before you. Uh, that is one of the most transformative aspects of this technology compared to any internet technologies that we've had. You can actually see ai through an appliance and figure out ways in which you're interacting with it. As a student, you can see it on your phone. So really thinking about what does ai literacy looks like so that people understand how that actually fits into the job, the labor force.
▶ 1:20:46>> thank you I yield back. >> thank you. Mr. burleson. Would you like to be recognized for five minutes? Sure. Thank you. >> um, miss fabrizio, at one point you in your I think in your testimony you talked about that, um, without federal privacy law, businesses and consumers are navigating a confusing state by state patchwork.
▶ 1:21:06Um, and I, I couldn't agree more, but can you elaborate on what, by us not having some kind of standardized, um, laws in place and really kind of open the door for one state to, to to squelch this industry? >> well, yes. And and more so it's very harmful to businesses, especially small businesses or startups with a unique idea to try to figure out how to scale.
▶ 1:21:35It's it's very stifling to try to comply with many different state laws and to to grow your market and to grow your product. And so that can be discouraging and that can discourage innovation. And that's not what we want to do here. We want to encourage innovation. And a federal privacy framework would help that give some consistency and some clear guardrails and rules of the road so that our innovators can innovate. >> right.
▶ 1:22:00Do you think we need in general, just more broadly speaking, privacy laws related to our own personal data? >> well, in terms of looking at a framework for privacy or for ai, it should be risk based and we should approach it that way. But I do think versus a patchwork approach, that again, companies just can't adhere to it. They can't small companies can't afford, um, big privacy law firms to support them. That gives the big companies a competitive advantage, boxes out small companies.
▶ 1:22:31>> so that with that being said, what would you like to see as a framework for, um, setting wide sweeping federal law? >> yeah. So we would be looking for something that is tech neutral, that's preemptive, that it's one framework and that it, um, it's risk based and that it identifies those core categories. And it also, you know, removes liability for companies that are compliant and that I think would really unlock and propel innovation and move us forward. >> interesting.
▶ 1:23:00I mean, can you were speaking kind of esoterically or like a little bit philosophically there, but specific. Can you, can you think of specific rules or regulations that you would that you would want to see implemented? >> I would love to get back to you with specifics and and have a further conversation about it. >> I respect that. Thank you, Mr. hammond. Um, what a lot of this debate is about what the values that we're going to teach this ai, which I find interesting, that we can't even agree on what free speech means.
▶ 1:23:30Uh, humanity. How can we trust ai to, um, determine, you know, how can we we attribute or provide some kind of values to ai? >> uh, this is an unsolved problem as it stands. So, you know, right now, you know, chatbots will adopt whatever value you put in. If you tell it to talk like a pirate, it will talk like a pirate. Uh, the question is, when these systems become more autonomous and acting on their own, how do they not just follow our values but follow them reliably?
▶ 1:24:00Um, and I think that's an area still of needed for further research. >> yeah. Thank you. Because I I'm touching back on the question I asked miss fabrizio. Um, which is that I think that the vulnerability for each one of us is, is that we're entering into an era where, um, an ai can, can devour as much information about you as possible and, and, and weaponize that against you.
▶ 1:24:27And so that is something I think is a concern. And the question is, how do we protect, um, people's people's rights? Um, for example, every day I get annoying phone calls for people from, like, people that want to know if I want to sell my rental homes every day. I had never signed up for anything. Somehow they know which which homes I have and they just always want to call me.
▶ 1:24:55I can only imagine how bad that's going to get when how much worse it's going to get. So which is why I think we should be considering as a on a federal level, having a tighter grip or control on the on the data about individuals particularly important data, whether it's the electronic medical record, we may need to do something like what we did with hipaa or in extend that beyond just your medical privacy. Doctor lee, you're nodding your head. >> I'm not in my head.
▶ 1:25:25We are way overdue in the united states for a national privacy standard. That would actually dictate what goes into the machines, what comes out of it in terms of what we consent to, not consent to implicitly as well as, you know, the ability of third parties to get ahold of that, you get calls about your rental property. I'm offered $100,000 every day through some type of voice clone.
▶ 1:25:46So I think that alongside deepfakes, which actually only exploits the opportunities you're talking about, we really do need federal privacy legislation to slow this down. >> thank you. I yield back. >> all right. I'll now recognize Mr. crane for five minutes. >> thank you. Um, I want to get back on this chip conversation that we were having. Doctor hammond. Um, you were talking about the need to buy time. I believe we were talking about the h20 chip. Is that correct? >> yes, sir. >> are those manufactured in taiwan or the us? >> primarily taiwan?
▶ 1:26:17>> do you do you have fears that, um, china has stated repeatedly that their plans are to take taiwan and by doing so would have access and control of all of these chip companies that the us has invested in. >> I'm very concerned about potential invasion or blockade against taiwan. I would say that if that were to occur, that tsmc would not be long for the world on.
▶ 1:26:42>> on that note, just to unpack it for the american people that might be watching, um, what percentage of chips that we use as american consumers? And an I come from taiwan. >> when it comes to the most advanced ai logic chips, like the the products, nvidia produces over 90%, 90%. And with the tsmc factories going up in arizona and elsewhere, we're going to try to grow grow our share. But right now. >> isn't it like 60% of semiconductors that we use come from taiwan as well? >> yeah. Across the board.
▶ 1:27:13>> what do you think would look like if that scenario were to unfold within the next couple of years? What do you think affects that would have on the us economy? >> I think we'd have a global depression for number one. But and then number two, you know, every all our products from our car door to our toaster have have semiconductors in them. >> when I've heard experts talk about, you know, standing up these manufacturing capabilities that, you know, building these advanced chips, it's not easy. It often takes decades. Would you agree with that? >> uh, it can from from scratch.
▶ 1:27:45But the, the fab going up in arizona is actually ahead of schedule and showing very good results after only a couple of years. >> so you think within a couple of years we can be producing these most advanced chips right here in the united states. >> with if similar style like the chips act 2.0 were to come across. We could do this again. Yeah. >> Mr.
▶ 1:28:03Fabrizio, you talked about other eras within history, like the invention of the internet, and you compared that to ai to kind of make your point that, you know, anytime there's been serious innovation within industry, it often leads to new jobs, other fields that we don't even yet know about. Do you think that that's coherent and fair?
▶ 1:28:25Looking at what we're facing here with ai, do you think it will be similar, or do you think ai will be a lot more disruptive for jobs and the economy than anything we've ever seen? >> I think it'll be it'll be a good shift that we see. There will be disruption, but I think it will be positive. I think it will allow us to solve some of the biggest challenges that we have. I think in healthcare we have a health care worker shortage, we have more sick patients. I think food insecurity, think farming, agriculture, energy, mobility, smart cities.
▶ 1:28:54I see all of these solutions at ces. I see amazing, groundbreaking technologies. And I would invite you all to come to ces and see them too, and see how the future will be amazing once we continue through this shift. If we allow the us to be the leader and continue to be the leader. >> do you agree with that, Mr. hammond? >> yes, I'd add a component to this vital component is cybersecurity infrastructure security. Right.
▶ 1:29:22So it's one thing we need these data centers in this country, uh, first and foremost for sovereignty, right. If we're going to be running a third of the economy through these data centers, if they're going to contribute to gdp, they should be within our borders, within our jurisdiction. But then that also creates a single point of failure, right? If the power goes out today, we're still able to have a conversation. But in the future, if the world is running on ai, that's that's a huge critical piece of infrastructure that could be taken out.
▶ 1:29:47>> yeah, I think we did a hearing on this same topic probably a couple of months back, and one of the ai titans, you know, we brought in an article and he said within the next 1 to 5 years, like 50% of all entry level white collar jobs are going to be gone. So you think those folks are all going to be able to find new jobs? >> well, I don't think those jobs will be gone. I think there there will be different jobs. And those people hopefully will will be focusing on upskilling. They'll be using ai.
▶ 1:30:15We've all heard that, uh, I won't replace people, but people who use ai will replace people who don't. And so that is how I look at it. And I do think that is a shift that we will experience. >> okay. So you guys don't think this is going to be like one of those times where we told coal miners, hey, when you're when your plant goes under and we get rid of it, we're going to teach them how to code. You guys remember that a couple of years ago. What do you think those coal miners are thinking right now?
▶ 1:30:42How many of them do you think learn how to code? And obviously this. >> is mine. >> it's not your response. I'm not saying it's your responsibility. I'm just saying this is one of the biggest concerns that I have. When I look at what's going on with ai. >> I agree. I think upskilling and investing in the future workforce is extremely important, and it's something that we have to do and we have to bring people along, but that involves education. It starts with stem education early in ai and continued investment there and being future looking. >> I get it.
▶ 1:31:11But a lot of these, a lot of these folks that already have these jobs, they're they already got their education that in many cases are going to be worthless. Would you agree, doctor turner? >> oh, of course, I think what you're actually pointing out, congressman, is that we need to do more research on this, and we need a little bit more data before we jump in here and start asserting what we think the workforce will look like. >> thank you, I yield back. >> okay. In closing, as we wrap this up, I want to thank our panelists once again for being here this afternoon, providing their testimony.
▶ 1:31:41I would like to yield to the ranking member for any closing remarks. >> uh, thank you, madam chair. Uh, I think some really interesting points were made, um, by members of the committee today, and I just I want people to understand that we should not downplay the job losses, the job displacement that is already happening. This is not theoretical.
▶ 1:32:04Uh, I had a job fair in my district a couple months ago, and I was ready to to meet, um, many folks who had lost their jobs because they're federal workers and have a lot of federal workers in my district. They've been laid off. And certainly there were those, but they were even more it students and graduates who were there because they had lost the ability to get a job because companies were not hiring them anymore. They wanted ten years of experience.
▶ 1:32:31Well, if you're a it student who just graduated from four year college, you were told your whole life to go into stem and you got the best stem education from the best stem education schools. And and now there's no job. And they're they were asking me why what happened. And um, and then I look at what's happening at the companies and they're laying off entire departments, um, and replacing them with ai and putting out press releases, in some instances bragging about it to their shareholders.
▶ 1:32:59Again, I'm not here to run their business. I think it's reality. If you could replace, uh, 100 people that you spend $1 million paying with a tool that cost $50,000, like it's it's part of your mandate to do that. But I think we have to understand that this is an imminent problem. Um, that, um, certainly we've invested a lot already in stem education and, and job training programs. But I actually kind of agree with you, Mr.
▶ 1:33:28Hammond, that I don't we have a lot of them already. I don't know if that's the only solution here. Um, I don't have any answers for you today. That's why I was asking these questions. Um, but I do want to work with my colleagues on both sides of the aisle to figure out what comes next, because I have a lot of students with really, um, good skills. They don't need to be upskilled anymore. They have really good tech skills. They just can't find jobs. Right. And so I want to see if there's something we can do about that. I yield back. >> thank you.
▶ 1:33:58And with that, without objection, all members will have five legislative days within which to submit materials and to submit additional written questions for the witnesses, which will be forwarded to the witnesses for their response. And if there's no further business without objection, the committee stands adjourned. Thank you. Yes.