Bresnahan Says US Risks AI Race to China on Grid, Labor Shortages

AI Regulation in America's Technological Future - AI Policy Summit Keynote Presentation — Ai4

Ai4

Massive AI data center expansion faces severe power grid and skilled labor deficits, threatening American computational supremacy over China unless Congress accelerates permitting and baseload energy buildouts.

AI-assisted summary of AI Regulation in America's Technological Future - AI Policy Summit Keynote Presentation — original source. Check the original for full context.

Source published: 2025-08-28

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 28, 2025, this keynote presentation features Congressman Rob Bresnahan, Jr. of Pennsylvania's 8th Congressional District. According to the video description, the session explores the evolving legislative and regulatory landscape for artificial intelligence in the United States, examining America's global leadership, governance challenges, and opportunities for bipartisan progress toward an ethical and competitive AI future.

## TL;DR Congressman Rob Bresnahan, Jr. examines the rapid expansion of artificial intelligence through the lens of physical infrastructure, energy demands, and skilled labor shortages facing northeastern Pennsylvania. Drawing on his background as an electrical contractor, he argues that supporting multi-gigawatt data center buildouts requires urgent permitting reform, workforce investment in technical trades, and the strategic expansion of baseload nuclear and natural gas power. Bresnahan calls for proactive, bipartisan congressional action informed by technical experts to establish regulatory clarity, maintain American competitiveness against China, and translate technological expansion into local economic prosperity.

## Atmosphere & style The presentation carries an energetic, pragmatic, and blue-collar tone, grounding high-level Capitol Hill policy debates in the tangible mechanics of substations, wire-stripping, and construction permitting. Bresnahan speaks with approachable authenticity and humor, candidly sharing personal stories about his upcoming wedding vows and floor-negotiation workflows to demystify how lawmakers grapple with emerging tools. The delivery feels collaborative and urgent, framing the AI race not as a distant tech novelty, but as an immediate infrastructure challenge that requires bipartisan cooperation and continuous dialogue with industry practitioners.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) Artificial intelligence policy now intersects virtually every major legislative priority in Congress, including transportation, energy, commerce, and national defense. - [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) Northeastern Pennsylvania is seeing an unprecedented data center boom, driven by nine planned campuses and a $20 billion development commitment from Blackstone. - [02:51](https://www.youtube.com/watch?v=KNajDlBMHTM&t=171s) Projected AI data center power demand severely strains historical grid capacity, with nine regional campuses requiring 11 GW compared to PPL's existing 7 GW built over a century. - [02:51](https://www.youtube.com/watch?v=KNajDlBMHTM&t=171s) Constructing AI infrastructure faces an acute labor bottleneck, currently facing an immediate shortage of roughly 8,000 skilled trade workers in northeastern Pennsylvania alone. - [02:51](https://www.youtube.com/watch?v=KNajDlBMHTM&t=171s) The nation's electrical transmission and distribution backbone heavily relies on mid-century infrastructure that has already surpassed its intended 50-year design life. - [05:22](https://www.youtube.com/watch?v=KNajDlBMHTM&t=322s) Lengthy bureaucratic permitting delays severely escalate project costs, making federal and state permitting reform a rare area of urgent bipartisan collaboration. - [07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) Congress must act quickly to establish clear federal AI guidelines and avoid the multi-year regulatory paralysis and fragmented state patchwork seen in digital assets. - [07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) Unresolved legal and liability questions regarding enterprise prompt data ownership, algorithmic fault, and legal discovery require close consultation with technical and industry experts. - [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) The global race for AI dominance is fundamentally an energy race tied to national security, as China builds around 28 nuclear reactors to overtake US capacity by 2030. - [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) Hyperscalers are increasingly bypassing long turbine lead times by co-locating data centers directly behind the meter at baseload facilities, such as Amazon's Salem Township nuclear project. - [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) Regional advantages like localized natural gas supplies and geographic proximity to 60% of the US population make northeastern Pennsylvania prime real estate for AI infrastructure. - [17:05](https://www.youtube.com/watch?v=KNajDlBMHTM&t=1025s) AI and data center infrastructure are permanent economic fixtures that represent a generational opportunity to redefine societal productivity and global competitiveness.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) Introduction and the AI Data Center Boom in Northeastern Pennsylvania

Congressman Rob Bresnahan, Jr., representing Pennsylvania’s 8th Congressional District in northeastern Pennsylvania, opens his keynote by framing artificial intelligence not as an isolated technological novelty, but as a ubiquitous reality confronting the halls of Congress. Only eight months and seven days into his first term as a representative, Bresnahan observes that virtually every major legislative policy discussion across Capitol Hill—spanning commerce, labor, energy, and national defense—inevitably circles back to AI. Serving on the Transportation and Infrastructure Committee (where he serves as Vice Chairman of the Subcommittee on Highways and Transit), as well as the Committees on Agriculture and Small Business, he highlights how the Transportation and Infrastructure Committee has suddenly found itself at the epicenter of resolving permitting bottlenecks required to support massive AI data infrastructure.

In Bresnahan's home region of northeastern Pennsylvania, this technological wave has arrived with astonishing velocity. The district currently has nine projected data center campuses in development. At an Energy Summit in Pittsburgh just weeks earlier, Blackstone announced a staggering $20 billion commitment to data center development in the district. Turning these massive financial commitments into operating facilities, however, requires getting shovels into the ground—a domain Bresnahan understands intimately from his previous career as a third-generation electrical and power distribution contractor. Reflecting on his family’s business, founded by his grandfather in 1973 with little more than "a station wagon and a ladder," Bresnahan recalls being handed a razor blade as a young boy in 2009, parked in a workshop corner, and told to strip wire away from his body while his mother checked at the end of each day to ensure he still had all his fingers. That operational background gives him a pragmatic view of technology; he has witnessed firsthand how AI has already been woven into physical infrastructure through intelligent transportation systems—optimizing highway tolling, adaptive lighting, closed-circuit television monitoring, and automated roadway de-icing systems—as well as legacy modernization projects, such as a colleague's initiative converting 40 years of paper blueprints into CAD drawings to catalog complete component inventories.

- Rob Bresnahan, Jr. represents northeastern Pennsylvania (PA-08) and has served in Congress for 8 months and 7 days. - Serves on the House Transportation & Infrastructure Committee (Vice Chair of Highways and Transit), Agriculture Committee, and Small Business Committee. - Northeastern Pennsylvania is slated for 9 major new data center developments, headlined by a $20 billion investment commitment from Blackstone.

---

[02:51](https://www.youtube.com/watch?v=KNajDlBMHTM&t=171s) The Power Grid Bottleneck and the Looming Skilled Trades Shortage

The conversation surrounding AI scale inevitably runs directly into the hard physics of power generation and electrical distribution. In Pennsylvania, Pennsylvania Power and Light (PPL), the Commonwealth’s primary electric utility, spent over an entire century developing roughly 7 gigawatts (GW) of generation and grid capacity. By comparison, the nine proposed data center campuses slated for northeastern Pennsylvania alone will require an estimated 11 gigawatts of new power. This explosive demand creates an immediate friction between commercial technological ambitions and consumer welfare: policymakers must fiercely protect residential constituents to guarantee electric utility bills do not spiral out of control, while holding corporate developers accountable to ensure they act as constructive, civic-minded neighbors.

``` PA Power Grid Capacity vs. Projected AI Data Center Demand ======================================================================== PPL Grid Capacity (Built over 100 years): [#######] 7 GW Northeastern PA Data Centers (New Demand): [###########] 11 GW ======================================================================== ```

Even if the primary energy sources—principally natural gas and nuclear power—are readily available in the Commonwealth, an equally acute bottleneck looms: human capital. The massive electrical facilities, transformer substations, and multi-million-square-foot compounds require a small army of electricians, carpenters, plumbers, and operating engineers. Right now, planned projects in Bresnahan’s district face an immediate shortfall of approximately 8,000 skilled trade workers. As Bresnahan points out, the workforce needed to erect these projects over the coming decade cannot simply be hired off the street; many are currently sitting in seventh and eighth-grade classrooms. Without rapid, aggressive investment in career and technical education (CTE), high school trade pathways, and joint apprenticeship training committees (JATCs), labor shortages will throttle AI infrastructure development. Furthermore, the physical grid relies heavily on transmission and distribution systems installed during the mid-20th-century New Deal and post-war expansions—systems engineered for a 50-year lifespan that are already operating on borrowed time.

- PPL's historical grid capacity: ~7 GW built over 100 years. - Projected regional AI data center power demand: 11 GW. - Local labor deficit: 8,000 skilled trade workers needed immediately across planned facility construction sites. - Grid vulnerability: Major electrical transmission backbones date back to the 1950s New Deal era and have exceeded their planned 50-year design life.

---

[05:22](https://www.youtube.com/watch?v=KNajDlBMHTM&t=322s) Permitting Delays, Infrastructure Costs, and Bipartisan Solutions

Bridging the divide between conceptual infrastructure and physical facilities requires tackling America's broken permitting system. Bresnahan notes that despite partisan rancor dominating national headlines, infrastructure bottlenecks represent one of the few areas yielding genuine bipartisan collaboration. Serving as a member of both the Building America Caucus and the Problem Solvers Caucus (a coalition equally split between 22 Republicans and 22 Democrats), he sees lawmakers on both sides of the aisle uniting around the shared urgency of cutting red tape to deploy critical projects before capital evaporates.

To illustrate the human and financial cost of bureaucratic inertia, Bresnahan shares the plight of an aging, 1,000-foot bridge in Nanticoke, Pennsylvania, carrying between 6,000 and 10,000 vehicles per day. The bridge was recently designated structurally deficient. Yet, even under an ideal scenario where federal lawmakers, state senators, and state representatives execute every procedural step flawlessly, the Pennsylvania Department of Transportation (PennDOT) will not be legally or administratively positioned to put the project out to bid for another seven years. Today, replacing that bridge carries an estimated price tag of $55 million. Factoring in seven years of compounding material cost escalations, regulatory delays, and labor premiums, the project will likely balloon to $85 million. Bresnahan argues that burning tens of millions of taxpayer dollars on procedural delays rather than physical construction is unconscionable, underscoring why both parties are eager to reform federal and state permitting timelines to get critical projects moving.

- Nanticoke, PA Bridge: 1,000-foot span carrying 6,000–10,000 vehicles daily. - Permitting timeline: Requires an estimated 7-year administrative lead time before PennDOT can even solicit construction bids. - Escalation cost impact: Project costs are projected to rise from $55 million today to $85 million by bid date.

---

[07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) Navigating AI Policy, Education, and Legal Exposure in Congress

As the seventh-youngest member of the House of Representatives, Bresnahan emphasizes that Congress faces a once-in-a-generation window to establish a sensible regulatory framework for AI, cautioning against the agonizing delays that previously paralyzed tech policy. He cites crypto regulation—such as the recent passage of comprehensive digital asset market structure legislation (the FIT21 / Financial Innovation and Technology for the 21st Century Act)—which languished for years despite sustained pleas from industry stakeholders for clear regulatory boundaries. With AI, lawmakers cannot afford years of indecision while fragmented state-level regulations create a fractured patchwork across the country.

Acknowledging candidly that members of Congress are generalists rather than technical specialists, Bresnahan stresses that Capitol Hill must listen closely to technical, medical, and legal experts. Recent roundtables with educators, physicians, and attorneys revealed profound unanswered questions regarding civil liability and evidentiary discovery: - Are enterprise prompts and uploaded data considered proprietary assets or open-source material? - Who bears tort liability when an algorithmic diagnosis or automated process causes catastrophic harm? - Are search queries and internal chats conducted through LLMs discoverable during civil and criminal litigation?

On the human side, Bresnahan reflects on how rapidly the technology has infiltrated daily life compared to his own college days at the University of Scranton. Drawing a laugh from the audience, he shares a personal anecdote about his upcoming wedding: > "I did. I'm getting married on Saturday and I've been asked to... thank you. Hopefully, we've been asked to write our own vows. And I'll tell you—this, we're all friends in this room—don't tell Chelsea what I want to do here. But the reality is you need to be able to marry critical thinking to what information is actually coming out of different programs and different language models. Like, I can say Chelsea is a wonderful person because she suffers through me in a congressional campaign and she still decides to marry me. And you know, that language model might spit out something, but it still comes down to critical thinking. Like, I have to decide as a human if that is who I am, if that's my voice, and then be able to differentiate what the reality is and what's actually tangible."

Bresnahan explains that this tension between raw algorithmic output and critical human judgment played out directly on the House floor during rapid-fire negotiations on H.R. 1. When the Senate stripped and restructured major energy provisions before sending the bill back to the House, lawmakers had mere hours—not weeks—to digest hundreds of dense pages of legislative text. Bresnahan and his legislative team utilized large language models to ingest spreadsheet data and statutory text, rapidly tracking phase-out timelines for renewable energy tax credits (including wind and solar provisions) and modeling how proposed permitting exemptions would directly alter the viability of planned 9-million-square-foot data center developments in his home district.

---

[13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) National Security, Nuclear Energy Competition, and Regional Advantage

When engaging members of Congress who might otherwise be indifferent to digital technology, Bresnahan explains that the most potent catalyst for legislative action is national security—specifically the strategic race against China. The global AI dominance contest is fundamentally an energy race. China currently has roughly 28 nuclear reactors under construction or planned and is on course to surpass the United States in aggregate nuclear energy generation capacity by 2030. If the United States fails to expand its baseload energy supply, it will inevitably surrender technological leadership.

``` Global AI Race: The Nuclear Energy Dimension ======================================================================== China Nuclear Buildout: ~28 reactors active / in pipeline; projected to eclipse US output by 2030. US Strategic Challenge: Must deploy baseload (nuclear + natural gas) or forfeit technological and computational primacy. ======================================================================== ```

This strategic dynamic explains why northeastern Pennsylvania has emerged as prime real estate for hyperscalers. Rather than waiting five to ten years and paying a quarter of a billion dollars for utility-scale gas turbine deliveries, tech giants are co-locating directly alongside baseload power sources. Amazon Web Services, for example, secured nearly 1,000 acres in Salem Township, Pennsylvania, to tap zero-carbon energy "behind the meter" directly from the Susquehanna Steam Electric Station nuclear plant.

Bresnahan makes an explicit pitch for his region, pointing out that because interstate pipeline constraints frequently prevent Pennsylvania from exporting its abundant natural gas reserves, localized on-site generation and turbine power create an ideal environment for data center campuses. Positioned within a single commercial driving day of 60% of the U.S. population—and with 50 million people living within a 200-mile radius—northeastern Pennsylvania offers prime zoned industrial acreage. Closing his address, Bresnahan reiterates that AI will define the next century's geopolitical winners and losers. He provides his direct email address (`robb@b4pa.com`), reminding the summit attendees that representatives have "two ears and one mouth" and must rely on continuous guidance from engineers, operators, and builders to craft regulatory policies that foster national competitiveness while delivering tangible economic prosperity to local working-class communities.

- China's nuclear pipeline: Rapidly deploying approximately 28 reactors, tracking to overtake the U.S. in nuclear generation by 2030. - Amazon's Salem Township investment: Acquired roughly 1,000 acres for direct, behind-the-meter nuclear power access at the Susquehanna nuclear plant. - Geographic advantage of NEPA: Within a single commercial driver's license (CDL) drive to 60% of the U.S. population, with 50 million residents within 200 miles.

[17:05](https://www.youtube.com/watch?v=KNajDlBMHTM&t=1025s) Shaping the Next Century and Moving AI Forward Together

Concluding his keynote address, Congressman Rob Bresnahan emphasizes the transformative potential of artificial intelligence to expand thinking and drive unprecedented gains in overall societal productivity. He makes it clear that neither AI nor the critical data center infrastructure supporting it are fleeting trends—they are here to stay. Because of this permanence, society faces a pivotal, once-in-a-lifetime generational moment that holds the power to fundamentally reshape the trajectory of the next century.

Expressing his excitement and gratitude for the opportunity to speak with the summit attendees, Bresnahan notes that he will be walking around the conference and actively seeking conversations with participants. He urges attendees to connect and collaborate with him on practical ways to work together, keep the momentum going, and steer the future of technology and policy forward.

- **Societal Impact**: Artificial intelligence and data centers represent permanent fixtures of modern infrastructure with the capacity to elevate national productivity. - **Generational Moment**: Policymakers and industry leaders share a rare window of opportunity to guide the technological trajectory of the coming century. - **Collaboration**: Open invitation to industry stakeholders at the conference to partner directly on legislative and governance strategies.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) a station wagon and a ladder — Rob Bresnahan, Jr.

> [07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) I'm getting married on Saturday and I've been asked to... thank you. Hopefully, we've been asked to write our own vows. — Rob Bresnahan, Jr.

> [07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) don't tell Chelsea what I want to do here. But the reality is you need to be able to marry critical thinking to what information is actually coming out of different programs and different language models. — Rob Bresnahan, Jr.

> [07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) Like, I can say Chelsea is a wonderful person because she suffers through me in a congressional campaign and she still decides to marry me. — Rob Bresnahan, Jr.

> [07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) And you know, that language model might spit out something, but it still comes down to critical thinking. — Rob Bresnahan, Jr.

> [07:13](https://www.youtube.com/watch?v=KNajDlBMHTM&t=433s) Like, I have to decide as a human if that is who I am, if that's my voice, and then be able to differentiate what the reality is and what's actually tangible. — Rob Bresnahan, Jr.

> [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) behind the meter — Rob Bresnahan, Jr.

> [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) two ears and one mouth — Rob Bresnahan, Jr.

## Mentioned Rob Bresnahan, Jr., Blackstone, Pennsylvania Power and Light (PPL), Pennsylvania Department of Transportation (PennDOT), Building America Caucus, Problem Solvers Caucus, Financial Innovation and Technology for the 21st Century Act (FIT21), University of Scranton, House Transportation and Infrastructure Committee, House Committee on Agriculture, House Committee on Small Business, Amazon Web Services (AWS), Susquehanna Steam Electric Station, H.R. 1, Nanticoke, Pennsylvania, Salem Township, Pennsylvania, China

## Video description Rob Bresnahan, Jr., Congressman, 8th District of Pennsylvania

Abstracts

In this session, we will explore the evolving legislative and regulatory landscape for artificial intelligence in the United States. The keynote will set the tone for the summit by reflecting on America’s global leadership in AI innovation, addressing the challenges of governance, and highlighting opportunities for bipartisan progress in shaping a safe, competitive, and ethical AI future.

[Source](https://www.youtube.com/watch?v=KNajDlBMHTM)

中文简报

布雷斯纳汉称电网与劳工短缺或令美国在对华AI竞赛中失利大规模AI数据中心扩张面临严重的电网与技术劳动力短缺，除非国会加快审批流程并扩建基荷能源，否则美国对中国的算力优势将受到威胁。联邦众议员小罗布·布雷斯纳汉（Rob Bresnahan, Jr.）从宾夕法尼亚州东北部面临的物理基础设施、能源需求和技术工人短缺等角度，审视了人工智能的快速扩张。结合自身作为电气承包商的背景，他认为，支持吉瓦级数据中心的建设，亟需进行许可审批改革、加大对技术工种的劳动力投资，并战略性扩建核能与天然气等基荷电力。布雷斯纳汉呼吁国会在技术专家的建议下采取积极的两党行动，以明确监管规定，维持美国对中国的竞争力，并将技术扩张转化为当地的经济繁荣。宾夕法尼亚州东北部正迎来前所未有的数据中心建设热潮，规划了9个园区，其中包括黑石集团（Blackstone）高达200亿美元的投资承诺。 (0:04) 区域内9个规划园区的电力需求预计达11 GW，远超当地电力公司PPL用一个世纪建成的7 GW电网容量，且建设面临约8,000名熟练技工的严重短缺。 (2:51) 冗长的官僚审批延迟导致基础设施成本急剧上升，使联邦和州的审批制度改革成为两党紧迫合作的罕见共识领域。 (5:22) 国会必须迅速确立明确的联邦AI监管框架，避免重蹈加密资产领域多年监管停滞与各州法规割裂的覆辙，并妥善解决提示词数据所有权及法律责任等未决问题。 (7:13) 全球AI竞争本质上是能源竞赛；得益于核电并网与区位优势，亚马逊AWS等科技巨头正加速在宾州直接利用核电部署数据中心，以确保基荷能源供应。 (13:37)

Reading summary

U.S. Representative Rob Bresnahan, Jr. argues that artificial intelligence leadership is bounded by the physical realities of power generation, labor capacity, and permitting rather than software innovation alone. Balancing rapid computational expansion against residential consumer protections and infrastructure bottlenecks represents a decisive economic and national security imperative. Grid Capacity Limits Confronting Regional Data Centers (2:51) In northeastern Pennsylvania, nine projected data center campuses—anchored by commitments such as a $20 billion investment from Blackstone—face severe electrical infrastructure bottlenecks. Pennsylvania Power and Light (PPL) required roughly a century to establish 7 gigawatts (GW) of generation and grid capacity, whereas the regional data centers under development will require an estimated 11 GW of new power. Bresnahan cautions that policymakers must actively safeguard residential ratepayer bills from escalating as commercial demand surges. This localized demand shock is further exacerbated by the physical grid itself: critical transmission and distribution corridors engineered during mid-20th-century New Deal expansions have already exceeded their 50-year operational life spans. Skilled Labor Shortages and Bureaucratic Permitting Costs (5:22) Deploying physical AI infrastructure is severely constrained by human capital deficits. Bresnahan reports an immediate shortfall of approximately 8,000 skilled trade workers—including electricians, plumbers, carpenters, and operating engineers—across planned facilities in Pennsylvania's 8th Congressional District, requiring expanded high school trade pipelines and technical apprenticeships. Compounding labor constraints, systemic administrative permitting hurdles delay project delivery and drive severe budget inflation. In Nanticoke, Pennsylvania, replacing a structurally deficient, 1,000-foot bridge handling 6,000 to 10,000 daily vehicles requires an estimated seven-year administrative lead time before state agencies can solicit construction bids, escalating projected costs from $55 million to $85 million. Bresnahan notes this capital waste has mobilized bipartisan coalitions, including the Problem Solvers Caucus, to pursue regulatory permitting reform. Geopolitical Nuclear Race and On-Site Colocation Advantages (13:37) Bresnahan asserts that the global AI rivalry against China is fundamentally a competition over baseload power capacity. With China constructing or planning roughly 28 nuclear reactors and projected to surpass U.S. nuclear energy generation by 2030, domestic developers are shifting to behind-the-meter power solutions to bypass multi-year turbine procurement delays and transmission interconnection queues. For example, Amazon Web Services acquired approximately 1,000 acres in Salem Township, Pennsylvania, to draw zero-carbon baseload electricity directly from the Susquehanna Steam Electric Station nuclear plant. Because interstate pipeline restrictions frequently constrain natural gas exports out of Pennsylvania, localized on-site generation makes the region attractive for hyperscalers, reinforced by geographical access to 60% of the U.S. population within a single commercial driving day. Federal Regulatory Delays and Unresolved Legal Liabilities (7:13) Warning against the prolonged regulatory indecision that stalled cryptocurrency policy before the passage of the FIT21 Act, Bresnahan contends Congress must act swiftly to prevent a fragmented patchwork of conflicting state-level AI mandates. Consultations with legal, medical, and educational professionals highlight unresolved statutory questions concerning tort liability for automated decision errors, whether enterprise prompts constitute trade secrets or public domain data, and the civil discovery limits of user queries. While congressional offices are already deploying large language models operationally—such as modeling tax credit phase-out schedules and permitting exemptions under H.R. 1 within hours—Bresnahan emphasizes that legislative frameworks must balance automated tools with human critical reasoning, requiring ongoing technical guidance from outside engineers and industry practitioners. Bresnahan views artificial intelligence and its physical data infrastructure as permanent foundations of future productivity, arguing that current legislative choices will dictate national competitiveness for decades. However, unresolved tensions remain between accelerating capital deployment to match foreign nuclear-backed initiatives and shielding domestic utility customers from associated infrastructure costs.

精炼阅读摘要

美国联邦众议员Rob Bresnahan认为，全球人工智能的主导权竞争本质上取决于底层电力与基础设施的交付能力；面对数据中心建设潮带来的电网容量挤压与熟练工短缺，国会必须通过两党协作加速许可审批改革，并厘清AI民事责任规则，以维持国家安全战略优势。算力爆发加剧区域电网负荷 (2:51) Bresnahan指出，数据中心的扩张速度正与电力供给的物理限制产生直接摩擦。在其代表的宾夕法尼亚州东北部第八选区，目前有9座数据中心园区正在规划中，其中包括黑石集团（Blackstone）近期宣布的200亿美元投资承诺。然而，当地主要电力公用事业机构宾夕法尼亚电力与照明公司（PPL）耗费一个多世纪才建立起约7吉瓦（GW）的电网容量，而这9个拟建园区的新增电力需求预计将达11吉瓦。此外，区域主要输电干线大多建于20世纪中叶新政时期，已超出原本50年的设计寿命。这种供需矛盾为政策制定带来了新的挑战。Bresnahan强调，在支持基础设施落地的同时，立法者必须保障居民电费不受商业用电需求激增的冲击，并促使大型科技开发商承担社区责任，防止基础设施改造成本单方面外溢至普通纳税人。技工断层与审批延误制约落地 (5:22) 物理设施的建设不仅面临电力约束，还面临严重的熟练劳动力缺口。Bresnahan说明，仅其所在选区规划的数据中心及变电站等配套工程，就面临约8,000名电工、木工、水暖工和操作工程师的即时用工缺口。他指出，此类劳动力无法即时从社会招募，必须通过对职业技术教育（CTE）、高中职教路径以及联合学徒培训委员会（JATC）进行持续投资来补充，否则劳动力断层将直接阻碍AI算力基建的成型。与此同时，行政审批制度的滞后显著增加了基础设施的落地成本。Bresnahan以宾州南蒂科克（Nanticoke）一座日均通行6,000至10,000辆车、已被定为结构缺陷的千英尺桥梁为例：即便各级政府程序配合无误，宾夕法尼亚州交通部（PennDOT）也需要7年行政审批周期才能启动招标，导致预计建造成本从当前的5500万美元膨胀至8500万美元。他表示，此类程序性延误正是促使国会跨党派议员联合推进联邦与州级许可改革的核心原因。国会面临监管滞后与法律界定难题 (7:13) 在政策框架方面，Bresnahan警告国会必须抓住当前窗口确立明晰的AI监管边界，避免重蹈加密货币监管（如FIT21法案经历多年推迟）的覆辙，防止各州自行出台割裂的法规。他坦承国会议员多为通才，需依赖跨领域专家的输入。在近期的行业座谈中，包括企业提示词与上传数据的专有产权属性、算法医疗诊断等自动化流程致害时的民事侵权赔偿主体，以及大模型交互记录在诉讼中的证据开示（discovery）范围等关键法律问题，目前均未获得明确规范。 Bresnahan同时透露，AI已在国会立法日常中发挥实际作用。例如在审议H.R. 1法案的关键博弈阶段，参议院对能源条款进行了大幅删改，议员团队在仅有数小时审阅数百页复杂文本的情况下，借助大语言模型快速提取数据并分析风能、太阳能税收抵免的退坡时间表，以评估法案对本选区900万平方英尺数据中心建设可行性的影响。他同时指出，使用此类工具必须依托人类的批判性思维来核实真实性与立足点。基荷能源竞争与表后核电战略 (13:37) Bresnahan认为，推动国会跨越数字技术分歧的最强动因在于国家安全与对华战略竞争。他提出，AI主导权的竞争核心在于基荷能源供应，中国目前在建或规划中的核反应堆约有28座，并预计在2030年前后在总核能发电量上超过美国；若美国无法迅速扩充基荷电力供给，将可能在长远计算与技术实力上落后。这一战略背景加速了科技公司在区域基荷电厂周边的“表后”（behind-the-meter）零碳供电布局。由于采购公用事业级燃气轮机通常需要等待5至10年且单项开支巨大，大型企业更倾向于直接毗邻现有电源建厂。例如，亚马逊网络服务（AWS）在宾州塞勒姆镇购买了近1,000英亩土地，直接接入萨斯奎哈纳（Susquehanna）核电站供电。由于州际管道运力限制阻碍了宾州天然气的对外输送，就地转化为电能成为优势方案，加上该区域单日车程覆盖全美60%人口（200英里半径内有5000万人），正使其成为支撑算力基础设施的关键枢纽。 Bresnahan总结强调，人工智能与数据中心基础设施不是短期趋势，而是将重塑下一个世纪竞争力的永久性物理存在。当前正处于制定长期规则的关键战略窗口，立法者需要持续与一线工程师、操作人员及建设者对话，以便在解决电网和劳动力供给瓶颈的同时，平衡国家科技战略安全与地方工薪阶层的实际经济利益。

Detailed notes — zh

## Background 该视频由 Ai4 于 2025 年 8 月 28 日在 YouTube 发布，记录了美国宾夕法尼亚州第 8 选区联邦众议员小罗布·布雷斯纳汉（Rob Bresnahan, Jr.）在 AI 政策峰会（AI Policy Summit）上的主题演讲。根据页面简介，本次演讲聚焦美国人工智能立法与监管格局的演变，探讨了美国保持全球创新领导地位的治理挑战，以及通过跨党派合作塑造安全、具竞争力和合乎伦理的 AI 未来的战略机遇。 ## TL;DR 布雷斯纳汉结合其家族电气承包背景与选区发展现状，深入剖析了宾州东北部数据中心建设所面临的 11 吉瓦电力缺口、8,000 名熟练技工短缺以及冗长审批周期的现实瓶颈。他分享了国会团队利用 ChatGPT 快速评估法案税收条款的实践经验，并指出了数据专有权和法律可调取性等隐私监管隐忧。此外，他从中美核能竞争与国家安全战略的高度出发，强调必须吸取加密货币监管迟滞的教训，呼吁两党携手加速许可审批改革，在发展算力基础设施的同时深化产教融合与社区合作。 ## Atmosphere & style 整场演讲呈现出务实、接地气且兼具战略紧迫感的基调。布雷斯纳汉将早年在家族企业剥电线皮的草根经历、用 ChatGPT 构思婚礼誓词的幽默轶事，与国会山高强度的立法抉择和地缘政治博弈巧妙穿插，展现出极具亲和力与穿透力的叙事风格。听众能真切感受到一位年轻立法者对基础设施短板的清醒洞察，以及他积极架起科技界与政策制定者沟通桥梁的开放态度。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) 众议院交通与基础设施委员会等机构正将核心议题转向解决 AI 数据集建设、算力中心落地及基础设施审批瓶颈。 - [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) 黑石集团宣布在宾州东北部投资 200 亿美元规划建设 9 座数据中心，但项目落地面临审批、电力与劳动力等多重挑战。 - [03:00](https://www.youtube.com/watch?v=KNajDlBMHTM&t=180s) 电力负荷缺口巨大，规划中的 9 座数据中心需 11 GW 电力，而当地最大电企历经百年才建成 7 GW 供电容量，且 1950 年代老旧电网普遍超期服役。 - [03:00](https://www.youtube.com/watch?v=KNajDlBMHTM&t=180s) AI 基础设施面临高达 8,000 名熟练技工的短缺危机，亟需工会学徒培训委员会与中等教育深化合作以更新课程体系。 - [05:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=330s) 美国现行冗长的审批机制导致重大基础设施建设常拖延多年（如桥梁审批需 7 年），引发建设成本恶性通胀并浪费纳税人资金。 - [07:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=450s) 国会必须汲取加密货币监管立法迟缓的前车之鉴，抓住代际机遇，及时为 AI 构建协调一致的联邦与州级监管框架。 - [07:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=450s) 无论日常应用还是工业落地，大语言模型生成的内容都必须与人类的批判性思维紧密结合，才能确保真实性与可行性。 - [10:25](https://www.youtube.com/watch?v=KNajDlBMHTM&t=625s) 国会自身已在立法实务中利用大语言模型快速解析繁杂法案条款，但这也引发了关于输入数据的专有权归属、隐私泄露及诉讼调取证据的监管隐忧。 - [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) AI 竞争本质上也是国家安全与能源保障的较量；面对中国加速建设 28 座核反应堆的挑战，美国需充分利用现有核电等基础设施为数据中心供能。 - [15:47](https://www.youtube.com/watch?v=KNajDlBMHTM&t=947s) 人工智能将直接决定下一个世纪全球竞争的赢家与输家，立法者需多倾听产业一线诉求，协同推进审批改革与经济机遇落地。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) 国会山上的 AI 浪潮与宾州东北部的数据中心投资来自宾夕法尼亚州斯克兰顿（Scranton）的联邦众议员罗布·布雷斯纳汉（Rob Bresnahan, Jr.）在演讲开场时介绍，作为上任刚满 8 个月零 7 天的新晋国会议员，他发现如今美国众议院几乎所有的议题探讨，都在以某种方式与人工智能（AI）产生交集。他在众议院交通与基础设施委员会担任公路与过境、航空等相关小组的副主席，同时兼任农业委员会与小企业委员会委员。近期，交通与基础设施委员会正面临一项新任务：解决从基础设施审批难题到 AI 数据集与算力中心落地的全链条瓶颈。 ```mermaid flowchart TD accTitle: AI 数据中心建设与基础设施挑战 subgraph District ["宾夕法尼亚州第8选区 (NEPA)"] A["黑石集团 200 亿美元投资"] --> B["规划建设 9 座大型 AI 数据中心"] end subgraph Challenges ["核心基础设施瓶颈"] B --> C["电力短缺 需额外 11 GW 负荷"] B --> D["劳动力缺口 急需 8,000 名熟练技术工"] B --> E["审批与老旧电网 1950年代电网超期服役"] end ``` 在布雷斯纳汉代表的宾州东北部选区（NEPA），目前已规划了 9 座大型数据中心。数周前在匹兹堡举办的能源峰会上，黑石集团（Blackstone）宣布在该选区投资 200 亿美元建设数据中心设施。这一庞大的投资虽然令人振奋，但如何让“铁锹真正铲进土里”（shovels in the ground）动工建设，成为摆在眼前的最大现实挑战。在步入政坛之前，布雷斯纳汉长期经营一家家族电力与配电承包企业。该企业由其祖父于 1973 年仅凭一辆旅行车和一把梯子白手起家创立；布雷斯纳汉于 2009 年正式接班。他回忆起童年时祖父塞给他一把美工刀、让他躲在角落学着朝外剥电线皮，而母亲每天傍晚最大的期望就是“数数他的十根手指头是不是都还在”。结合自身行业经历，他指出了 AI 已经在传统行业中的具体落地：如智能交通系统（ITS）中运用的自动化道路收费、智能照明控制、监控摄像头以及自动除冰系统；还有企业正利用 AI 工具将积攒了 40 年的传统蓝图工程图纸批量转换并录入 CAD 系统以建立数字资产库。 --- ### [03:00](https://www.youtube.com/watch?v=KNajDlBMHTM&t=180s) 电力需求激增与 8,000 名熟练工人的劳动力缺口 AI 基础设施的建设首先面临着严峻的电力负荷挑战。宾夕法尼亚电力与照明公司（PPL）作为宾州最大的电力供应商，耗费了整整 100 年时间才建成了 7 吉瓦（GW）的供电能力。然而，仅宾州东北部规划中的 9 座数据中心，就将带来高达 11 吉瓦（GW）的新增电力需求。布雷斯纳汉强调，在满足算力电力需求的同时，必须顾及本地居民的利益，防止电价飙升至居民无法承受的水平。 ```mermaid flowchart LR accTitle: 宾州电力容量与数据中心电力需求对比 A["PPL 历经百年建成容量 7 GW"] -.->|缺口高达 11 GW| B["9 座新建数据中心需求 11 GW"] ``` 除了天然气与核能等发电端资源外，电力传输与分配网络也是巨大的短板——美国现存的大部分输配电系统建于 20 世纪 50 年代罗斯福新政与战后扩张时期，设计使用寿命仅为 50 年，如今普遍处于超期服役状态。另一个迫在眉睫的瓶颈是劳动力危机。仅这些在建与规划项目，就面临高达 8,000 名熟练技术工人的短缺： - **技工断层**：电工、水管工、木工及基础建设人员严重不足。布雷斯纳汉形象地指出：“未来负责搭建这些数据中心建筑框架的木工师傅们，现在还在上初中七、八年级呢。” - **教育与职业培训衔接**：呼吁工会学徒培训委员会与高中、职业技术学校紧密合作，更新课程体系，在学生步入劳动力市场时提供精准对接产业实际需求的技能培训。 --- ### [05:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=330s) 跨党派合作与破除 7 年审批沉疴布雷斯纳汉作为跨党派“建设美国党团”（Building America Caucus）和由 22 名共和党人与 22 名民主党人组成的“问题解决者党团”（Problem Solvers Caucus）成员，指出基础设施建设与审批改革正在成为打破华盛顿党派僵局的关键纽带。为了说明美国现行审批制度的冗长与低效，他列举了家乡南蒂科克（Nanticoke）一座跨度约 1,000 英尺、日通行量为 6,000 至 10,000 辆机动车的桥梁案例： - **审批周期冗长**：该桥梁在一个月前被正式评定为“结构性缺陷”，即便联邦议员、州参众议员以及宾夕法尼亚州交通部（PennDOT）每一个环节都做到极致完美，该项目依然需要漫长的 7 年时间才能正式进入公开招标程序。 - **成本恶性通胀**：该桥梁在当下的建造成本约为 5,500 万美元；若拖延 7 年，加上材料价格通胀、劳动力成本攀升以及工程机械操作员短缺等因素，总造价可能飙升至 8,500 万美元。布雷斯纳汉直言，这种审批拖延是对纳税人资金的极大浪费。两党议员虽然在诸多政策上存在分歧，但在“必须大幅缩短重大项目审批与落地周期”这一问题上达成了罕见共识。 --- ### [07:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=450s) 监管框架构建、加密法案前车之鉴与人类的“批判性思维” 作为国会第七年轻的联邦众议员，布雷斯纳汉探讨了如何构建 AI 的国家级监管与法律框架。他回顾了国会此前通过的《21世纪金融创新与技术法案》（FIT21 / Genius Act 等加密货币监管立法），指出加密行业曾苦苦恳求国会介入并划定清晰的合规边界，国会却耗费多年才完成立法；而在 AI 领域，联邦政府不能重蹈覆辙，必须抓住当下“一生一次的代际机遇”，协调好联邦法规与各州法律之间的关系。在与高等教育界、医生和律师讨论民事责任（civil liability）边界的过程中，布雷斯纳汉坦言，国会议员需要就广泛议题进行决策，但绝非技术底层逻辑与垂直行业的专家，立法者极度依赖现场专业人士的深度输入与建言献策。在谈及 AI 在日常生活与工作中的普及度时，布雷斯纳汉幽默地分享了一则个人轶事：他将在本周六举行婚礼，未婚妻切尔西（Chelsea）要求两人亲笔撰写结婚誓词。他打趣道：“大家在屋里都是朋友，可千万别告诉切尔西——我确实试着用 ChatGPT 跑了一下誓词。”他表示，AI 可能会输出“切尔西是个伟大的人，因为她忍受了我经历漫长的国会竞选，却依然决定嫁给我”这类文案，但最终依然需要人类依靠“批判性思维”（critical thinking）去审视： - 这段话是否代表了自己真实的声音？ - 输出的内容是否契合现实且具有真实的情感温度？他强调，无论是在日常生活中撰写誓词，还是在工业场景中将几十年的工程图纸数字化，大语言模型与 AI 工具生成的答案都必须与人类的批判性思维紧密结合，才能真正将技术潜力转化为切实可行的成果。 ### [10:25](https://www.youtube.com/watch?v=KNajDlBMHTM&t=625s) 国会如何使用 AI 工具及隐私监管隐忧国会议员罗布·布雷斯纳汉（Rob Bresnahan, Jr.）分享了他与业界人士关于 AI 人才培养的对话。当被问及企业如何寻找人才来提取和分析数据时，对方表示既依赖高校开设有针对性的交叉学科课程（结合软件工程与数据分析），也需要能够将复杂数据转化为清晰逻辑思考的复合型人才。随后，布雷斯纳汉通过一个具体案例，展示了国会自身在立法实务中对 ChatGPT 等大语言模型的实际应用。在近期 HR1 法案（H.R. 1）投票期间，众议院通过的版本被送往参议院，经过参议院议事专家（Parliamentarian）删除部分条款并修改通过后，又重新送回众议院。国会议员们没有数周的时间来仔细研读这些繁杂的文本，而是只有几个甚至几个小时的紧迫窗口。为了评估法案中关于可再生能源税收抵免（Renewable Energy Tax Credits）的取消或阶段性淘汰时间表对本选区重大投资项目的影响，布雷斯纳汉的团队直接将数据电子表格导入语言模型中进行分析，快速生成图表和总结，从而在极短时间内做出审慎的决策，并向选区投资者和居民澄清误导性信息。然而，这种高效的应用也引发了深层的法规与隐私思考。布雷斯纳汉指出，当议员或企业将机密或敏感数据输入大语言模型时，这些数据是否还具有专有权？是否会变成开源内容？在法律诉讼程序中，这些对话记录和输入内容是否属于可被调取（Discoverable）的证据范畴？这些监管框架与隐私问题目前仍处于探索阶段，也是国会必须面对的现实课题。 - **立法应用**：在 HR1 法案投票的紧迫时间内，利用 ChatGPT 分析风能和太阳能税收抵免政策的取消时间表。 - **产业背景**：当地正在规划占地 900 万平方英尺的巨型数据中心园区，建设和运营将创造数百个高技术工种与长期就业岗位。 - **监管隐忧**：数据输入大语言模型后的产权归属、隐私保护以及在法律程序中的可调取性（Discovery）问题。 ### [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) 国家安全、核能竞争与宾州东北部的崛起人工智能的影响力正在渗透到各个行业，从医疗保健、能源到电子游戏和娱乐，没有任何领域能够独善其身。但在国会内部，最能引起议员们高度警惕和关注的切入点无疑是**国家安全**，特别是美国与中国在能源保障与 AI 基础设施方面的激烈竞争。布雷斯纳汉特别强调了宾夕法尼亚州东北部（Northeastern Pennsylvania）正在成为全国数据中心新枢纽的原因。传统天然气或风力发电往往需要等待 5 到 10 年才能交付价值 2.5 亿美元的离网发电机组，且受到管道输送限制。而该地区拥有萨兰镇（Salem Township）的核能发电设施，能够直接提供“表后发电设备”（Behind-the-meter）的稳定电力。例如，亚马逊（Amazon）已在该地区收购了近 1,000 英亩土地，用于建设由核能直接供电的数据中心。对比之下，中国正在加速推进核能基础设施建设。布雷斯纳汉警告称，中国目前拥有多达 28 座在建或规划中的核反应堆，预计将在 2030 年前在核能发电总量上超越美国。如果美国不能加快许可审批改革（Permitting Reform）并整合自身丰富的自然资源与技术优势，就将在未来的全球科技竞争中失去主导权，被动接受外部制定的规则。 - **区位优势**：宾州东北部位于优越的物流网节点，单日 CDL（商业驾照）货运车程可覆盖全美 60% 的人口，200 英里范围内拥有 5,000 万消费人口。 - **中美核能对比**：中国正推进 28 个核反应堆项目，预计 2030 年在核能发电量上超越美国；美国需利用现有核电基础设施（如萨兰镇核电站）应对 AI 数据中心的巨大用电需求。 ### [15:47](https://www.youtube.com/watch?v=KNajDlBMHTM&t=947s) 拥抱 AI 时代：协作、经济机遇与未来展望在总结演讲时，布雷斯纳汉重申了高等教育与 AI 素养、许可审批改革、国家安全以及数据中心电力保障等核心议题。他强调，人工智能将直接决定下一个世纪全球竞争的“赢家与输家”。为了鼓励行业参与者与政策制定者开展更紧密的沟通，布雷斯纳汉公开了自己的联系邮箱（`robb4pa.com`），并表示国会议员应当秉持“两只耳朵、一张嘴”的原则，多倾听来自一线的声音，深入了解各项法律和监管框架将如何影响甚至决定下一个颠覆性创新成果的成败。他呼吁科技界与产业界积极向立法者阐释 AI 带来的巨大经济机遇——包括创造就业岗位、扩大地方税基以及提升社会整体生产力。通过成为良好的社区邻居与明智的投资者，完全可以将技术进步与社会福祉融为一体。 - **联系方式与开放态度**：公开个人邮箱 `robb4pa.com`，欢迎科技界与企业界就政策监管与项目落地进行直接沟通。 - **核心倡议**：抓住这“一生一次”的世代机遇，通过良性投资、批判性思考与社区合作，共同塑造未来一个世纪的发展轨迹。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) 铁锹真正铲进土里 — 罗布·布雷斯纳汉 > [00:04](https://www.youtube.com/watch?v=KNajDlBMHTM&t=4s) 数数他的十根手指头是不是都还在 — 罗布·布雷斯纳汉 > [03:00](https://www.youtube.com/watch?v=KNajDlBMHTM&t=180s) 未来负责搭建这些数据中心建筑框架的木工师傅们，现在还在上初中七、八年级呢。 — 罗布·布雷斯纳汉 > [05:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=330s) 必须大幅缩短重大项目审批与落地周期 — 罗布·布雷斯纳汉 > [07:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=450s) 一生一次的代际机遇 — 罗布·布雷斯纳汉 > [07:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=450s) 大家在屋里都是朋友，可千万别告诉切尔西——我确实试着用 ChatGPT 跑了一下誓词。 — 罗布·布雷斯纳汉 > [07:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=450s) 切尔西是个伟大的人，因为她忍受了我经历漫长的国会竞选，却依然决定嫁给我 — 罗布·布雷斯纳汉 > [07:30](https://www.youtube.com/watch?v=KNajDlBMHTM&t=450s) 这段话是否代表了自己真实的声音？输出的内容是否契合现实且具有真实的情感温度？ — 罗布·布雷斯纳汉 > [13:37](https://www.youtube.com/watch?v=KNajDlBMHTM&t=817s) 表后发电设备 — 罗布·布雷斯纳汉 > [15:47](https://www.youtube.com/watch?v=KNajDlBMHTM&t=947s) 赢家与输家 — 罗布·布雷斯纳汉 > [15:47](https://www.youtube.com/watch?v=KNajDlBMHTM&t=947s) 两只耳朵、一张嘴 — 罗布·布雷斯纳汉 ## Mentioned 罗布·布雷斯纳汉 (Rob Bresnahan, Jr.), 切尔西 (Chelsea), 黑石集团 (Blackstone), 宾夕法尼亚电力与照明公司 (PPL), 宾夕法尼亚州交通部 (PennDOT), 亚马逊 (Amazon), ChatGPT, 智能交通系统 (ITS), 建设美国党团 (Building America Caucus), 问题解决者党团 (Problem Solvers Caucus), 《21世纪金融创新与技术法案》 (FIT21), HR1 法案 (H.R. 1), 宾夕法尼亚州东北部 (NEPA), 斯克兰顿 (Scranton), 南蒂科克 (Nanticoke), 萨兰镇 (Salem Township) ## Video description Rob Bresnahan, Jr., Congressman, 8th District of Pennsylvania Abstracts In this session, we will explore the evolving legislative and regulatory landscape for artificial intelligence in the United States. The keynote will set the tone for the summit by reflecting on America’s global leadership in AI innovation, addressing the challenges of governance, and highlighting opportunities for bipartisan progress in shaping a safe, competitive, and ethical AI future. [Source](https://www.youtube.com/watch?v=KNajDlBMHTM)

Source quotes

[0:04] 罗布·布雷斯纳汉: 铁锹真正铲进土里 [0:04] 罗布·布雷斯纳汉: 数数他的十根手指头是不是都还在 [3:00] 罗布·布雷斯纳汉: 未来负责搭建这些数据中心建筑框架的木工师傅们，现在还在上初中七、八年级呢。 [5:30] 罗布·布雷斯纳汉: 必须大幅缩短重大项目审批与落地周期 [7:30] 罗布·布雷斯纳汉: 一生一次的代际机遇 [7:30] 罗布·布雷斯纳汉: 大家在屋里都是朋友，可千万别告诉切尔西——我确实试着用 ChatGPT 跑了一下誓词。 [7:30] 罗布·布雷斯纳汉: 切尔西是个伟大的人，因为她忍受了我经历漫长的国会竞选，却依然决定嫁给我 [7:30] 罗布·布雷斯纳汉: 这段话是否代表了自己真实的声音？输出的内容是否契合现实且具有真实的情感温度？ [13:37] 罗布·布雷斯纳汉: 表后发电设备 [15:47] 罗布·布雷斯纳汉: 赢家与输家 [15:47] 罗布·布雷斯纳汉: 两只耳朵、一张嘴

Source quotes

[0:04] Rob Bresnahan, Jr.: a station wagon and a ladder [7:13] Rob Bresnahan, Jr.: I'm getting married on Saturday and I've been asked to... thank you. Hopefully, we've been asked to write our own vows. [7:13] Rob Bresnahan, Jr.: don't tell Chelsea what I want to do here. But the reality is you need to be able to marry critical thinking to what information is actually coming out of different programs and different language models. [7:13] Rob Bresnahan, Jr.: Like, I can say Chelsea is a wonderful person because she suffers through me in a congressional campaign and she still decides to marry me. [7:13] Rob Bresnahan, Jr.: And you know, that language model might spit out something, but it still comes down to critical thinking. [7:13] Rob Bresnahan, Jr.: Like, I have to decide as a human if that is who I am, if that's my voice, and then be able to differentiate what the reality is and what's actually tangible. [13:37] Rob Bresnahan, Jr.: behind the meter [13:37] Rob Bresnahan, Jr.: two ears and one mouth

Original transcript

[0:04] Good afternoon everyone. Thank you so [0:06] much for having me. Um, [0:10] what a beautiful tide to celebrate here. [0:12] I'm uh I'm actually from Scranton, [0:14] Pennsylvania, nor eastern PA. I'm one of [0:16] those new members of Congress. I'm I've [0:18] been on the job here for 8 months and uh [0:22] 7 days. And I'll tell you a big part of [0:26] every conversation that we are having in [0:29] the United States House of [0:30] Representatives somehow someway [0:32] someplace ties back to artificial [0:35] artificial intelligence. Uh I serve on [0:38] the transportation infrastructure [0:39] committee where I serve as the vice [0:41] chairman of highway transit aviation as [0:44] well as FIA. I serve on the committee [0:47] for agriculture and small business. And [0:50] all of a sudden the TNI subcommittee has [0:53] been tasked with trying to solve for the [0:56] permitting challenges all the way back [0:59] the development of AI data sets. And in [1:02] my district, if anyone's familiar with [1:04] this the Kunnel of Pennsylvania, I [1:06] represent the northeastern quarter. Um, [1:10] and right now we have nine projected [1:13] dayup data centers uh in northeastern [1:16] Pennsylvania where Blackstone at the uh [1:18] Energy Summit just in Pittsburgh a few [1:20] weeks ago made a 20 billion investment [1:23] into our district and which is extremely [1:27] exciting on so many different levels. [1:30] Uh, but you have to think about getting [1:32] shovels into the ground. And in my last [1:35] life, I was a power distribution and [1:37] electrical contractor. I'm from a family [1:39] in business. Uh my grandpa grandfather [1:41] started the company in 1973. Uh really [1:44] with the station wagon and a ladder uh [1:46] and grew it to when I came into the [1:48] business in 2009 was always around the [1:51] business. My grandfather gave me a razor [1:52] blade, popped me up in a corner, told me [1:55] to strip the wire away from myself and [1:57] my mother would just make sure I had all [1:58] my fingers when I I'm at the end of the [2:00] day. uh but we have seen the [2:02] implementation of artificial [2:04] intelligence through intelligent [2:06] transportation systems just the way that [2:08] we're looking at polling on roadways and [2:10] manage lightings and uh CCG cameras and [2:13] BD icing systems. study. I'm traveling [2:15] with a good friend here today from from [2:17] Wow, who helps me be out on our special [2:19] projects and uh he's the interp company [2:22] where they are actually taking 40 years [2:25] of blueprints and converting them to [2:29] different versions of CAD and drawing so [2:32] they can culminate their entire [2:33] inventory. So, I'm not going to tell you [2:36] all about the different niches of [2:37] industry that it affects because you're [2:39] going to know that a lot more than me [2:41] because by the end of this, I'm going to [2:43] have a plea for each and every one of [2:45] you in here. Uh, and ultimately, it's [2:47] going to be to help us better [2:49] understand. Um, but some of the [2:51] conversations that we have is when PPNL, [2:55] Pennsylvania Power and Light, is the [2:57] largest uh producer of electricity in [3:00] the Commonwealth of Pennsylvania, [3:01] they've spent the last 100 years to [3:06] create a 7 gawatt availability of power. [3:10] Right now, between the nine data centers [3:12] that are projected in northeastern [3:14] Pennsylvania, uh we are going to need an [3:16] additional 11 gawatts of power. And [3:19] that's what we have been seeing take [3:21] place and unfold where you have to be [3:24] conscientious of our residents to make [3:26] sure that it's not going to raise the [3:29] the electricity pricing to the point of [3:31] it being unaffordable. But then you also [3:34] have to be able to be a good neighbor. [3:36] When you look at my district, uh we have [3:38] the availability of natural gas. We have [3:41] an availability of workforce. The [3:43] problem is the electricians that we need [3:46] right now to help us construct these [3:48] facilities are carpenters, our plumbers, [3:51] um our are our our carpenters that are [3:53] going to end up laying the framework. [3:54] They're in seventh and eighth grade. So [3:57] something that I've spent a big part of [3:58] my previous life doing was working with [4:02] the the various different apprentichip [4:03] training committees to make sure that we [4:06] are figuring out the lane in which we [4:09] need to develop the next generation of [4:11] of skilled trade. Um, and that's going [4:13] to become the pinch point. Right now, [4:15] just in those projects that are being uh [4:18] planned, we're going to need 8,000 [4:20] workers. 8,000. And the reality is we [4:24] don't have them right here and right [4:25] now. So, we need to be working inside of [4:28] our high schools and our career [4:29] technical schools and developing the [4:32] curriculum that meets the actual job the [4:35] time that they're ready to enter the [4:36] workforce. Because there's no question [4:38] in my mind that that's going to become [4:40] the next pinch point. there's going to [4:41] be a labor crisis and a labor shortage. [4:44] Um, and when you're thinking about the [4:45] power generation side of things, you [4:47] have to be able to generate the power. [4:49] You need to be able to transmit the [4:51] power and then ultimately distribute the [4:53] power. And if you think about when the [4:55] majority of our electrical systems were [4:57] built, it was in the 1950s under FDR's [5:00] New Deal with 50 years of usable life. [5:04] And we all know that the actual [5:06] infrastructure behind the actual data [5:09] composition is astronomical. Um so [5:12] finding that mechanism whether that's by [5:15] using natural gas whether that is using [5:17] nuclear um we have to be outfront and I [5:22] believe that the United States House of [5:24] Representatives including myself and [5:25] specifically the transportation [5:26] infrastructure committee realizes this [5:29] and we also realize that it's not [5:31] necessarily a partisan issue. Um, I sit [5:33] on a a caucus called uh Building America [5:36] Caucus where it's bipartisan [5:38] legislators. Uh, and it's actually [5:40] really exciting when Republicans and [5:42] Democrats come together and talk about [5:45] how do we [5:47] make a construction project get a shovel [5:50] in the ground faster. And I'm going to [5:51] give you a quick example. Uh, we have a [5:53] bridge uh near one of our uh down in [5:56] Nanako, Pennsylvania. And this bridge a [5:59] month ago was just deemed structurally [6:01] deficient. I mean, it's an old bridge. [6:03] It sees about 6 to 10,000 cars a day, [6:06] about a thousand foot span. But the [6:08] reality is that bridge will not be put [6:11] out to bid for another 7 years. Even if [6:14] I did every single thing correctly, if [6:17] the state legislators and the state [6:19] senators did exactly what they were [6:20] supposed to do, PennDOT, Pennsylvania [6:23] Department of Transportation, would not [6:24] be able to put that broad project out [6:27] for bid for another seven years. And the [6:30] reality is this bridge today costs $55 [6:33] million. When you take the escalation [6:34] cost of material, the escalation of [6:37] labor, if you can find the labor to [6:40] actually put those shovels into the [6:41] ground, those operating engineers, your [6:44] concrete, your foundation layers, that [6:46] bridge could cost $85 million. And I [6:49] don't think that is a prudent use of of [6:51] taxpayer dollars when you're making [6:52] critical infrastructure investments. Um, [6:56] so that is what's really exciting and [6:58] encouraging that we we both recognize [7:00] Republicans and Democrats and we have [7:01] plenty of other things to disagree on, [7:03] but one thing that does certainly bring [7:05] us together uh is the reality of how [7:08] long it takes to get a project in and [7:11] out of the ground. Um, so when I'm [7:13] talking to colleagues and I'm on the [7:15] problem solvers caucus where it's 22 [7:17] Republicans and 22 Democrats where we [7:19] sit and talk every single day and with [7:21] the explosion of AI and I say AI as a [7:26] really encompassing category, it's the [7:29] the regulatory framework and I'm the [7:32] seventh youngest member of Congress. Um, [7:36] so we're all fully aware of where the I [7:40] don't know if we're fully aware where [7:41] the future of it is going. I think we [7:44] all recognize that we have a once in a [7:46] generational opportunity to get a a hold [7:49] on what this all looks like. But you [7:52] also have to compare the federal [7:53] government versus individual states and [7:56] commonwealths and decide what they're [7:58] doing, what trajectories that they're [8:00] pursuing. And you know, just two weeks [8:03] ago, we passed the Genius Act on crypto [8:05] legislation. I mean, how many years was [8:07] that in the making before we were able [8:09] to get that across the finish line? And [8:11] this was something the crypto industry [8:13] was begging Congress to get involved on [8:16] and provide envelopes and framework as [8:18] to how does this move forward. Um just [8:21] on Wednesday last week, I had an [8:23] opportunity to speak with uh some [8:26] educators, higher education and um some [8:29] physicians and some lawyers about the [8:31] the civil liability and the civil [8:33] exposure. And as much as this I hope [8:35] you'll learn or take away something [8:37] about how we're thinking about the the [8:40] industry, uh we're really going to [8:43] depend on the input from a lot of you [8:46] because members of Congress are expected [8:48] to learn a lot of information about a [8:51] lot of different subjects. And I would [8:53] be totally remissed or downright lying [8:57] to say that I am an expert by any means [9:00] into the synergies and or the different [9:02] horizontal lines in which you're [9:04] operating from. Um but you know we were [9:06] talking to educators and higher [9:08] education is really also at a pinch [9:11] point here. You know I when I went [9:12] through high school and went to the [9:14] University of Scranton chat GBT wasn't a [9:17] reality. Now, I I'll tell you, I'm sure [9:19] if I asked anybody in here who's used AI [9:21] today, I think it would be unanimous. [9:23] Everybody would raise their hand and say [9:24] yes. Now, maybe a little bit different [9:26] if you use chat GBT, you know, per se. I [9:29] did. I'm getting married on Saturday and [9:31] I've been asked to thank you. Hopefully, [9:34] I've been we've been asked to write our [9:36] own vows. Um, and I'll tell you, I don't [9:39] this we're all friends in this room. [9:40] Don't tell Chelsea uh what what I want [9:42] to do here. But the reality is you need [9:44] to be able to marry critical thinking to [9:47] what information is actually coming out [9:51] of different programs in different [9:53] language models. Like I can say Chelsea [9:56] is a wonderful person because she [9:58] suffers through me in a congressional [10:00] campaign and uh she still decides to [10:02] marry me. And you know that language [10:04] model might spit out something but it [10:06] still comes down to critical thinking. [10:08] like I have to decide as a human if that [10:10] is who I am, if that's my voice and then [10:13] be able to differentiate what the [10:16] reality is and and what's actually [10:18] tangible. So when Ricky and I were just [10:21] talking about some of the development on [10:23] the um you know on the blueprinting [10:25] side, I said, "Well, how are you finding [10:27] people to extrapolate this information? [10:29] How are you finding are you working with [10:32] colleges? Are you working with you know [10:34] software engineers?" And you know, he [10:36] said really it's it's both. I mean, [10:38] sure, some of our universities are are [10:40] getting out in front and starting to [10:43] craft the curriculum around, you know, [10:46] that hybrid of software engineer, but [10:49] still he needs someone to take that [10:50] information and be able to formulate [10:54] that into a coherent thought. And I'm [10:55] going to give you an example of how [10:57] Congress is actually using chat language [11:00] models or chat GBT. And this was during [11:03] the uh HR1 vote uh just a few weeks ago. [11:06] And when the House passed the version uh [11:09] we had sent that back over to the [11:11] Senate. The Senate then did what they [11:13] did with it. Then there was a Senate [11:14] parliamentarian and different things [11:15] were stripped out. Then it went back in [11:16] the Senate. Then it passed the Senate. [11:17] Then came back to the House. So we had [11:20] not two weeks to read through these [11:22] different documents. We had hours. And [11:25] something that I fought extremely hard [11:27] for was for the uh the renewable energy [11:30] tax credits because of how much economic [11:31] development and I'm a big believer in [11:33] I'm not an and I'm I'm an and guy. I'm [11:35] not an or guy and I think that big [11:37] island for uh power generation [11:39] specifically when we're talking about [11:41] these data center compounds. Uh I mean [11:43] one of these data center compounds is 9 [11:45] million square feet um which will create [11:47] hundreds and hundreds of jobs and after [11:49] the construction on the skilled trade [11:50] and then full-time jobs. Um, the reality [11:54] is we need to get back out in front of [11:59] where this is all going and we took the [12:03] spreadsheet of the HR1 when the timeline [12:06] and how things were being phased out, [12:09] whether it was the wind credit, it was [12:11] the solar credit, and we were able to [12:13] because a lot of us didn't know what was [12:15] happening. And like I said, we had [12:17] minutes to hours to be able to make a [12:19] conscious conscientious decision with [12:22] our investors, with our investments that [12:23] are taking place inside of our district [12:25] to say, "Hey, could you get that shovel [12:27] into the ground?" If permitting is [12:30] excluded from this parameter, is this a [12:32] project that you can turn into reality? [12:35] And that's as simple as, you know, you [12:37] utilizing something, being able to build [12:39] out graphs and charts to be able to then [12:42] articulate a lot of misinformation or [12:45] possibly quell some concerns about the [12:47] reality of of performing a project. But [12:50] when you're talking about the privacy, [12:51] you're talking about the regulatory [12:52] framework, you know, like is this [12:54] something that, and this is what I'm [12:56] thinking about, and I I hope other [12:58] members of Congress are thinking about [12:59] it, but if I put something into a [13:01] language model, does that information [13:03] proprietary? Is that open source? Who is [13:07] exposed? Is this discoverable during a [13:09] legal proceeding? Is this something when [13:12] you're expected to turn over any and all [13:14] documents? Are are language search [13:16] models, chat GBT search models, you [13:19] know, expected to be able to be used in [13:22] discovery to be either used in your [13:24] defense or against you. And those are [13:27] the realities that we're all facing [13:29] with. And you know, Congress going back [13:33] to the point of being exposed to many [13:35] different subjects, [13:37] it's a lot to digest. And you could talk [13:40] about the healthcare space, you could [13:41] talk about the energy space, you could [13:42] talk about the gaming space, you could [13:44] talk about the entertainment space. [13:45] There's not going to be an industry that [13:49] will be immune from where we are going [13:53] with AI. [13:55] Now, where you'll get a member of [13:56] Congress's attention is when you start [13:58] to talk about national security and the [14:00] reality of what China is doing relevant [14:04] to nuclear energy. Um, and a big reason [14:07] why northeastern Pennsylvania became the [14:09] epicenter of of data centers is because [14:13] we have the ability to generate the [14:15] power um short of waiting 5 to 10 years [14:19] for a turbine to be delivered and [14:21] ordered and at a quarter of a billion [14:22] dollars. But we have uh a nuclear um [14:26] generation facility out in Salem [14:27] Township. That's why Amazon targeted [14:30] that area and acquired over close to a [14:32] thousand acres where they're going to [14:34] have behind the meter energy. I think [14:36] China is up to 28 uh different nuclear [14:39] reactors and they are scheduled to [14:42] surpass the United States on nuclear [14:44] generation by 2030. And I think we have [14:47] a unique opportunity to either set the [14:49] pace as the United States or we are [14:52] going to have the pace set for us. And [14:54] that's something that I'll tell you as a [14:57] member of Congress is certainly [14:58] concerning is because we have the [15:00] skilled resources, we have the natural [15:02] energy and we have the total package. I [15:05] mean where I am and this is my plug. If [15:07] anybody wants to build the data center, [15:08] as long as we can be good neighbors uh [15:10] with the existing residents, uh you make [15:13] the capital improvements into our power [15:15] distribution systems, northeastern [15:16] Pennsylvania is an incredible place to [15:18] call home. I mean, within one day CDL [15:20] drive, you can reach 60% of the United [15:22] States population. Within 200 miles is [15:25] 50 million people. So because of our [15:28] inability to be able to export energy [15:30] because of pipeline restrictions, [15:31] natural gas and the turbine [15:33] self-generation has really started to [15:36] see an increase that we have the ability [15:37] to generate the power on site, power [15:38] these facilities and then ultimately [15:41] create these data centers in very [15:44] wellestablished zoned industrial spaces. [15:47] So I touched on the higher education and [15:49] the AI literacy. I talked to the [15:51] permitting reform, the national security [15:52] in China, uh the data construction, the [15:55] power generation. Really, in closing, I [15:57] mean, AI is going to define the next [15:59] century of winners and losers. And I'm [16:02] going to have you, my email is [16:03] robb4pa.com. [16:05] That's how accessible I am. And if you [16:08] are in Washington, if you do have [16:09] various different projects, we have two [16:12] ears and one mouth as a member of [16:13] Congress. And our job truly here is to [16:16] learn about how different parameters, [16:19] different regulatory framework could [16:21] affect, make or break the next [16:23] incredible innovation. So I don't know [16:26] if that's about my time here. I got two [16:29] more minutes. Okay. So keep it going. um [16:33] when you're talking to and I think you [16:35] also need to do a better job and all of [16:37] us need to do a better job is talking to [16:39] members of Congress and and talking to [16:41] your legislators about what this [16:44] actually means the opportunity to create [16:46] more economic opportunity to create more [16:48] jobs to put people back to work increase [16:51] your tax bases the community benefit [16:53] impact on when these monumental projects [16:56] actually break ground but there's a way [16:59] that we can marry all of this Together [17:01] we can marry good investment by being a [17:03] good neighbor. We can marry critical [17:05] thinking, expand and increase our actual [17:08] productivity as society. I don't feel [17:11] like AI is going anywhere. I don't think [17:13] data centers are going anywhere, but [17:15] there's a great opportunity as a society [17:18] to be able to really change the [17:19] trajectory of the next century. We have [17:21] a once in a-lifetime generation here. [17:23] I'm so excited and blessed to be here [17:25] and and join all of you today. Uh and [17:27] we'll be scurrying around the [17:28] conference. I would love to get to meet [17:30] uh all of you if I could about how we [17:33] can all work on this together and keep [17:35] this band moving forward. So with that, [17:36] thank you all very much.