의회가 AI를 활용하여 미국 민주주의에 대한 신뢰를 회복할 수 있을까? LORELEI KELLY
의회가 AI를 활용하여 미국 민주주의에 대한 신뢰를 회복할 수 있을까? LORELEI KELLY
미 의회, 인공지능을 활용하여 미국 민주주의에 대한 신뢰를 회복할 수 있을까?
로렐라이 켈리 / 2024년 9월 18일 미국 의회는 수년 동안 기술 역량 측면에서 민간 부문과 다른 정부 기관에 뒤처져 있습니다. 이러한 디지털 격차는 기본적인 유권자 소통에서부터 국가 안보에 대한 상황 인식에 이르기까지 모든 것에 영향을 미치며, 미국 정부의 세 가지 권력 분립 균형을 깨뜨려 의회의 비용으로 행정부와 사법부에 권한이 집중되도록 만들었습니다.
그러나 의회는 이러한 역학을 바꿀 수 있는 엄청난 자원을 가지고 있습니다. 의원과 유권자 간의 소통은 민주주의의 핵심 임무 데이터입니다. 시민들의 실제 경험에서 비롯된 이 데이터는 의회에 의해 생산되거나 의회를 위해 생산된 과학적 및 기타 형태의 전문 지식과 결합되어 현대적 대의 민주제의 디지털 기반으로 활용될 수 있습니다.
현대화의 다음 단계는 무엇인가? 이러한 높은 목표를 달성하려면 많은 노력이 필요할 것입니다. 입법부의 기술에 대한 구식 접근은 수년 동안 만들어진 상황입니다. 원격 화상 회의를 도입하는 데는 COVID-19 비상 사태가 필요했으며, 최근까지 필수적인 의회 지원 사무실은 무료 버전의 Zoom을 사용하고 있었습니다. 심지어 지방 수준에서도 2023년이 되어서야 하원 선거구 사무실이 의회 Wi-Fi에 연결되었습니다. 많은 미국인들이 자신들의 목소리가 들리지 않는다고 느끼며, 그들은 완전히 틀리지 않았습니다. 미국의 가장 강력한 민주주의 기관인 의회는 스스로를 들을 수도 없습니다. 의회의 기억력 대부분은 국회 의사당 지하실의 상자에 있습니다. 의회는 매일 자신의 기록에 접근하고 현대 세계에 적응하기 위해 고군분투하고 있습니다.
의회의 효과성 저하는 우연이 아닙니다. 의도적인 선택의 결과입니다. 정보 공유는 당 지도부에 통합되었으며, 양극화로 악화되었습니다. 235년 된 이 기관에서는 구식 관점과 전통이 지속되며, 의회는 고집스럽게 자신에게 투자하기를 거부하여 디지털 인프라 개발을 지연시켰습니다.
다행히도 이러한 문제 상황이 개선되기 시작했습니다. 2019년 이후 의회는 기관 현대화를 통해 자체적으로 업데이트하고 개혁하기 시작했습니다. 양당 간 위원회가 202개의 권고안을 통과시켰으며, 현재 현대화 소위원회와 의회 개혁 초당파 그룹은 이러한 변화를 이행하기 위해 노력하고 있습니다.
현재 의회에는 441명의 의원이 있으며, 이 중 435명은 주에서 선출되고 6명은 미국 영토를 대표하는 비투표 의원입니다. 현재 각 의원은 자신의 데이터를 독립적으로 관리하고 있습니다. 현대화 개혁은 의회 디지털 서비스의 도움과 "의회 개혁 코호트"의 시민 사회 단체와의 협력을 통해 이러한 과제를 해결하기 시작했습니다. 또한 의회는 기술 기반 생태계를 조성하고 업무 흐름에서 AI를 실험하고 지침을 수립하고 의원 주도 AI 태스크포스를 시작하고 직원에게 교육을 제공함으로써 AI 발전에 보조를 맞추기 시작하고 있습니다.
이러한 노력을 기반으로 의회는 이제 미래로 과감한 도약을 할 수 있는 기회를 가지고 있습니다. 의회 지도자들이 선택한다면, 의회는 신흥 공공 AI 운동에서 중요한 역할을 하고 의회의 기능과 미국 민주주의 자체를 개선하기 위한 플랫폼을 구축할 수 있습니다. 다음은 그 방법입니다.
민주주의 모델 구축 대의 민주주의의 중심 기관인 의회는 매일 방대한 양의 데이터를 생성하고 대규모 국가 기록망을 감독합니다. 데이터 처리 및 인공 지능 기술의 융합으로 인해 이 정보를 활용하는 것은 그 어느 때보다 설득력 있는 전망입니다.
아래의 첫 번째 분파 LLM(대규모 언어 모델) 개념도는 공개 및 입법자 모두를 위한 접근 가능하고 책임감 있고 영구적인 정보 시스템을 만드는 데 필요한 기업 수준의 중요한 인프라를 상상합니다. 이 시스템은 또한 더 포괄적인 기록을 만들어 더 명확한 법령 해석을 지원하며, 이는 특히 대법원의 최근 Chevron 존중 결정을 뒤집은 후에 중요합니다. 이 개념의 주목할 만한 특징은 유권자의 목소리를 포함하는 것으로, 이제 비영리 심의 기술 및 음성 데이터 캡처의 발전으로 가능합니다. 공공 서비스 데이터를 최적화하기 위해 AI 도구를 사용하는 것은 21세기의 주요 과제이지만, 의회의 기능과 대중에 대한 서비스를 변화시킬 수 있는 잠재력을 가지고 있습니다.
저자가 제공한 일러스트레이션. 위 다이어그램의 왼쪽에 있는 모양은 의회의 개별 의원이 워싱턴 D.C.에서 일상 업무를 수행할 때 사용하는 다양한 정보 소스를 나타냅니다. 이러한 각각의 소스는 데이터의 끊임없는 흐름을 생성하지만, 문제는 이러한 데이터의 대부분이 구조화되지 않고 입법에 집단적으로 사용하기에 쉽게 접근할 수 있는 방식으로 구성되지 않는다는 것입니다. 예를 들어, 모든 의원은 정치적 소속에 관계없이 재향군인, 사회보장, 연방 보조금과 관련된 문제를 처리합니다. 이러한 사례에서 익명화된 데이터는 모든 의원이 액세스할 수 있는 공유 리소스의 일부가 될 수 있습니다.
다이어그램의 중앙에는 "첫 번째 분파 샌드박스"가 있으며, 이는 미국 민주주의에 중요한 데이터를 관리하고 공유하는 방법을 재설계하고 개선할 수 있는 창의적인 기회로 가득 찬 공간입니다. 첫 번째 분파는 각 주에 있는 연방 예금 도서관과 의회 도서관과 같은 방대한 기록 보관소를 보유하고 있으며, 이미 재향군인 및 COVID-19과 같은 주제별 시민 목소리 기록을 보유하고 있습니다. 도서관은 신뢰할 수 있는 정보원이며, 의회를 지원하는 역할이 최신화되어야 합니다.
인터넷 아카이브의 웹 수확 및 코넬 대학교의 법률 정보 연구소와 같은 디지털 협업은 의회 데이터를 효과적으로 사용하는 훌륭한 예입니다. 유사한 접근법을 의회 의원이 주도하는 공개 모임에서 시민 목소리 데이터를 수집하는 데 적용할 수 있습니다. 이 데이터는 일반 대중을 위한 "의회 기록"과 유사하게 구성되어 접근 가능하도록 할 수 있습니다. 이는 입법 과정에 지역사회의 의견을 통합하여 심의를 개선하고 의회가 제1차 수정안의 청원 의무를 이행할 수 있는 능력을 향상시킬 것입니다.
낮은 신뢰와 극단적인 사회 양극화의 시대에 인민 중심의 첫 번째 분파 LLM은 민주 정부의 정당성을 회복하는 데 도움이 될 수 있습니다. 선거는 민주주의에서 최고의 책임 형태이지만, 유권자 목소리 기록은 의원들이 선거주기 외에 피드백을 받고 조정할 수 있는 방법을 제공할 것입니다. 지역 데이터가 더 가치롭고 생산적이 될수록, 의원들이 유권자와 선거 기부자뿐만 아니라 모든 유권자를 더 잘 대표하도록 유인할 것입니다.
다이어그램의 오른쪽에 있는 모양은 시민 사회가 첫 번째 분파 LLM 개발에 기여하는 방법을 보여줍니다. 의회 의원이 지역사회에서 수집하는 정보는 여기서 지역사회 참여확대를 가져와야 한다.
시민 참여 확대와 의회 효율성 증대지역의 목소리 수렴
민주주의 신뢰 회복
구현을 위한 노력과 과제
결론미 의회는 인공지능을 활용하여 민주주의를 강화하고 시민 참여를 확대할 수 있는 잠재력을 가지고 있습니다. 이 보고서는 의회가 데이터를 효과적으로 활용하고 시민의 목소리를 반영하는 시스템을 구축할 수 있는 구체적인 방안을 제시합니다.
미래를 위한 투자 미 의회가 이러한 변화를 이끌어낸다면, 미국 민주주의는 더욱 투명하고 효율적이며 시민들에게 더 가까워질 수 있을 것입니다. 이는 단순히 기술적인 변화를 넘어, 민주주의의 미래를 위한 중요한 투자입니다.
핵심 메시지:
Lorelei Kelly
Lorelei Kelly is the founder of Georgetown Democracy, Education + Service at the McCourt School Of Public Policy at Georgetown University. An expert on building inclusive and informed democratic systems, she founded the Resilient Democracy Coalition where she leads research on modernizing Congress. ...
Could Congress Leverage AI to Help Restore Faith in US Democracy?Lorelei Kelly / Sep 18, 2024
For years, the United States Congress has lagged behind the private sector and other branches of government in terms of its technological capabilities. This digital gap–which impacts everything from basic constituent communication to situational awareness about national security– has disrupted the balance of power among the three branches of the US government, contributing to the growing concentration of authority in the executive and judicial branches, at Congress's expense. But Congress is also sitting on a tremendous resource that could change this dynamic: communication between members and their constituents is democracy’s mission critical data. This data, drawn from both the lived experiences of citizens and combined with scientific and other forms of expertise produced for and by Congress, could serve as the digital foundation of a modern representative system. The next step in modernization?To achieve such a lofty goal will require a great deal of work. The antiquated approach to technology in the legislative branch is a situation years in the making. It took the COVID-19 emergency just to introduce remote video conferencing, and until recently, essential Congressional support offices were using the free version of Zoom. Even at the local level, it wasn't until 2023 that House district offices connected to congressional Wi-Fi. Many Americans feel unheard, and they aren't entirely wrong—Congress, the nation’s most powerful democratic institution, can’t even hear itself. Much of its memory is in boxes in the basement of the Capitol. On a daily basis, Congress struggles to access its own records and adapt to the modern world. This decline in Congress’s effectiveness isn’t accidental; it’s the result of deliberate choices. Information sharing has consolidated into party leadership, made worse by polarization. Outdated perspectives and traditions persist in this 235-year-old institution, and Congress has stubbornly refused to invest in itself, stalling the development of its digital infrastructure. Fortunately, this troubling situation is starting to improve. Since 2019, Congress has begun updating and reforming itself through institutional modernization. A bipartisan committee passed 202 recommendations, and now a Modernization Subcommittee and a Fix Congress Caucus are working to implement these changes. Currently, Congress has 441 members, including 435 elected from the states and 6 that represent US territories in a nonvoting capacity. Presently, each manages their own data independently. Modernization reforms have started to address this challenge with the help of a Congressional Digital Service and collaboration with civil society groups in the “Fix Congress Cohort.” Additionally, Congress is beginning to keep pace with AI advancements by creating a skills-based ecosystem, experimenting with AI in its workflows, establishing guidelines, initiating a member-led AI task force and providing training to staff. Building on these efforts, Congress now has the chance to take a bold leap into the future. If its leaders so choose, Congress could play a key role in the emerging public AI movement, and build a platform to improve its own functions and American democracy itself. Here’s how it could work. Building a model for democracyAs the central institution of representative democracy, Congress generates vast amounts of data every day and oversees a large national archival network. Putting this information to work has never been a more compelling prospect given the convergence of data processing and artificial intelligence technologies. The First Branch LLM (Large Language Model) concept map below imagines the critical infrastructure needed at the enterprise level to create an accessible, accountable, and permanent information system for both the public and lawmakers. This system would also help create a more inclusive record for clearer statutory interpretation, which is especially important after the Supreme Court's recent decision to overturn Chevron deference. A notable feature of this concept is the inclusion of constituent voices, which is now possible due to advances in nonprofit deliberative technology and voice-to-data capture. Using AI tools to optimize this public-serving data is a major challenge for the 21st century, but it holds the potential to transform how Congress functions and serves the public. The shapes on the left side of the above diagram represent the various information sources for an individual member of Congress as they go about their workday on Capitol Hill. Each of these sources generates a constant stream of data, but the problem is that much of this data is unstructured and isn't organized in a way that makes it easily accessible for collective use in lawmaking. For example, every member, regardless of political affiliation, handles issues related to veterans, social security, and federal grants. The anonymized data from these cases could be part of a shared resource that all members can access. In the center of the diagram is the "First Branch Sandbox," a space filled with creative opportunities to redesign and improve how Congress manages and shares data crucial to American democracy. The First Branch has vast archival resources, such as Federal Depository Libraries in every state and the Library of Congress, which already holds topic-specific civic voice archives like Veterans and COVID-19. Libraries are trusted sources of information, and their role in supporting Congress should be brought up to date. . Digital collaborations like the Internet Archive’s webharvest and Cornell University’s Legal Information Institute are excellent examples of using congressional data effectively. A similar approach could be applied to collecting civic voice data from public gatherings led by members of Congress. This data could then be organized and made accessible, similar to a "Congressional Record" for the general public. This would integrate community input into the legislative process, improving deliberation and enhancing Congress's ability to fulfill its First Amendment Right to Petition duties. In an era of low trust and extreme social polarization, a people-centered First Branch LLM could help restore legitimacy to democratic government. While elections are the ultimate form of accountability in democracy, a Constituent Voice Archive would provide a way for members to receive feedback and make adjustments outside of election cycles. As local data becomes more valuable and productive, it would create incentives for members to better represent all constituents, not just voters and campaign donors. The shapes on the right of the diagram illustrate how civil society contributes to developing the First Branch LLM. The information that members of Congress gather from their local communities—referred to here as the "district water cooler"—is very valuable but often poorly organized for use in policymaking. The diagram shows how members of Congress can meet with their communities to collect authenticated feedback or advisory data. Deliberative platforms are already being used for this purpose, organized through networks like the Council on Technology and Social Cohesion. From imagination to implementationMembers of Congress, in collaboration with nonprofit organizations, have already started experimenting with these methods. For example, Georgetown’s Beeck Center for Social Impact + Innovation piloted a community-informed data gathering method in New Hampshire, and the Popvox Foundation worked with the House Natural Resources Committee on a collaborative editing pilot focused on Environmental Justice legislation. The next step is to create a publicly maintained digital space where this community-generated data can be stored and accessed. A "data fabric" architecture is a promising model for organizing data in a complex system like Congress. The private sector provides many effective examples of data architecture that support geographically diverse and complex organizations. We can learn from these models to create an enterprise system that strengthens American democracy. The First Branch LLM is a conceptual plan for such a system, and while it's achievable, it will be costly to implement. Currently, Congress has a $10 million innovation fund that, despite being understaffed, is making remarkable progress with its extensive mission. Both in Congress and the private sector, financial constraints can drive innovation, but in Congress, this innovation should also focus on improving public access. Public access is crucial for restoring trust and legitimacy in democracy, so modernization is a necessary investment in the success of the American experiment. Democracy is at risk both in the US and globally. When institutions like today’s Congress seem ineffective, the appeal of authoritarianism as a means of restoring order grows. We must do everything we can to prevent this outcome. The First Branch LLM is just one idea, and it should spark a broader national conversation about democracy, technology, and how we do the work to protect a shared and prosperous future. Paul Werbos comments on the above Lorelei's piece addresses digital governance, the main theme of a major UN meeting organized by Mei Lin at the UN summit for the future.
In 45r Minutes, I get to speak at another session. Here is my comment on the digital governance session in general:
===================================================================
=============================
Thank you very much, Mei Lin, for your important informative report on this important UN activity you are leading.
In a way, this report reminds me of the first days when climate threats finally got the attention of high level decision makers. It was very exciting, and it was exciting to see key people in the room whom we all need to connect to better.
(I envy your access to that key woman at Cisco, though she would need to be connected herself to Clint Crosier of AWS and new technologies I will discuss.)
Better, somehow, but the challenge at issue is difficult enough that being in the room is only a first step, crucial as it is. In climate, for example, there are key ISSUES which still need to be connected and integrated (like the great battery technology of Lonnie Johnson which even now is not receiving the level of support, respect and understanding which the world needs him to have). The same problems of fragmentation are growing visibly in the internet area as in the climate area.
TODAY... in my 10AM talk in the space track of this UN gathering, I will do my feeble best in 15 minutes (followed by 15 minutes from Gary Barnhard) to describe how a new sustainable development Goal for space development
might help solve the larger problem, building new substantive connections between the larger policy needs (as in my first slide, in the deck attached) and solid, real, fundamental science and technology, which is more about bringing mathematical models together than having famous bodies in the room. We need to COMBINE that lower-level ground truth with the higher visions, which I agree are also essential.
Many in the space community are as bad as some of the narrower older SDGs, which often seem like collections of hungry people standing around a gravy train looking for anything edible.
I STRONGLY HOPE that the space community does not follow THOSE kinds of bottom-up examples, but instead asks the top-down questions: HOW CAN WE (WHEN WE ARE PART Of THAT COMMUNITY)
remember what we should be trying to accomplish here? Above all, how can we contribute to the success of humanity in rising to ALL SEVEN of the huge new challenges listed in the first slide, which begin with the usual four great fears
from climate and internet/AI issues to nuclear and biohazard issues?
I DO HOPE people will engrave three points in their memory crucial to all of us:
(1) LARGE PARTS of the space community are unique in the will and ability to "get their hands dirty" (like when Lonnie Johnson worked at JPL before building his companies), in a way which DEVELOPS FUNDAMENTAL NEW TECHNOLOGIES URGENTLY NEEDED ACROSS ALL SEVEN CHALLENGES. (Of course NASA's "spinoffs" and NSF's "broader benefits" displayed a similar understanding of what matters.)
(2a and 2b): TWO of the three "great hopes" needed to counterbalance the four great fears HAVE ALREADY RECEIVED serious, focused attention and strategic thinking -- far beyond what I see in any UN conferences except for narrower specialty conferences. The space community is not yet doing or integrating all we need, but they provide a foundations which a new UN SDG **COULD** pull together in an essential, open transparent global cooperative manner.
WE REALLY NEED HIGHLY FOCUSED STRATEGIC THINKING AND TRUE GENERAL INTELLIGENCE TO FOCUS ON TWO KEY GLOBAL VALUE MEASURES, REFLECTING TWO OF THE GREAT HOPES:
(2a) Economically sustainable human settlement of space, both in this solar system and beyond. The first stage is VERY much like the economic chall\enge faced by the US when it developed the transcontinental railroad and expanded from east to west. It was not just one big silver bullet (as hungry vested interests often insist on, with the help of governments corruption, like what weakens present US government funding and leaves open opportunities to others).
It was an integrated new economic system, just as slides 2,3 and 4 in the attached deck summarize. They map out a start. The SDG should go further, update and integrate, in global recurrent (two-way) networks of human and computer communication.
(2b) SEEING THE SKY with orders of magnitude greater depth and resolution and color. Just as climate policy needs more use and understanding of proven new technologies from Lonnie Johnson (whom people like Harris and Vance might really love, if they and their family knew him better), this SDG needs to build on great proven underused breakthrough technologies from David Hyland (which I had a chance to review and then fund when I worked at NSF) and Avi Loeb (Breakthrough Institute and its many connected networks, such as the global interferometry networks exploiting centers such as Alma and University of Hawaii). This general technology offers huge breakthroughs, of great value to the open international community, in two areas: (1) "SEEING THE SKY" as in global warning, as Guterres callled for in summer 2023 in the Security Council session on AI; (2) SEEing beyond our solar system, to FIND OUT who else is out there
and what they MIGHT be up to. BOTH can build upon and beyond the older Hyland and Breakthrough Institute technologies, by exploiting new quantum technology (the final slide) which multiplies the power of the deep learning technologies I developed decades ago by orders of magnitude. Thanks to the US patent system, the tQuA patent was issued about a week ago, and many more technical details on implementation will be openly published soon by USPTO.
**IF** the more advanced new quantum AGI technology is basically held back or outlawed by new systems of regulation, that will only give a monopoly on the new technology to strong forces already operating in the dark, endangering human life itself in many ways that the PCI community would recognize. Our best hope is a new SDG INCLUDING 2b, to accelerate open, transparent technically serious development of new intelligent platforms open to all of humanity (albeit under certain open transparent rules, allowing people to use the new integrated platforms without fear of certain kinds of abuse.) (CISCO and AWS will lose a lot of customers to internet fragmentation if fewer and fewer people trust the global clouds.)
<저작권자 ⓒ ainet 무단전재 및 재배포 금지>
|
많이 본 기사
최신기사
|