Regulating AI: Legal and Policy Perspectives

A template syllabus

  • Kevin Frazier

Several individuals contributed to the creation of this template syllabus. I am particularly grateful to Paige Arnold and Ana Barreto, my research assistants. Thanks also to Legal Priorities Project for encouraging and informing this work.

This headnote includes:

  • The Purpose of the Course
  • The Goals of the Course
  • Student Learning Objectives
  • Guidance for Instructors
  • Potential Student Evaluation Framework
  • Overview of Assignments 

PURPOSE OF THE COURSE

The law and, by extension, lawyers can alter the trajectory of AI development. Regulators around the world have discussed different legal frameworks to limit the magnitude and likelihood of risks presented by AI. Lawyers will play a major part in drafting and interpreting these efforts. And, today's students will oversee the amendment and alteration of any enacted frameworks in the near future.  

This course begins by exposing students to the potentially catastrophic consequences of AI developing absent sufficient norms, regulations, and laws. Next, students will confront specific manifestations of the risks created by AI. Finally, students will consider the tools at their disposal to steer AI development in a way that promotes the flourishing of future generations.

AI threatens substantial, potentially existential risks. When, how, and why those risks will emerge is unknown and will remain unknown for the foreseeable future. The already ubiquitous use of AI means that some of the traditional methods of detecting and mitigating risks created by technological innovation will not suffice.

The development of AI has exposed a disconnect that has become more expansive with each passing wave of technological innovation: the gap between widespread adoption of that technology and a corresponding set of norms, regulations, and laws (“safety measures”) to mitigate the risks posed by that technology. Consider that it took nearly 70 years after the invention of the telephone for more than 30 percent of the American public to own a phone; it took just 25 years for TV to reach the same threshold. More recently, the gap between invention and creation has continued to shorten—think about the rapid adoption of the PC (about 20 years to reach 30 percent ownership), the Internet (around 15 years to reach 30 percent), and cell phones (fewer than ten years).

Each wave of technological innovation has been more rapidly adopted—shortening the window for the development of safety measures. More than 100 million people used ChatGPT just two months after its launch. The speed of AI adoption combined with the risks created by the continued spread and development of AI technology requires the education of AI emergency planners—lawyers, technologists, computer engineers, etc. who understand the risks of AI and the need for a novel approach to developing and deploying safety measures.   

This is not a technical course, though students will learn some fundamentals of AI. This is not a doctrinal course, though students will dive into myriad fields of law. And, this is not a “light” course, though some of the readings may come from popular media, other readings and concepts will lie beyond the technical and legal knowledge of students—that’s intentional and necessary. 

Everyone, including the professors teaching this course, is still learning about AI and how society will respond to its development. This course depends on the willingness of all its participants to come prepared, learn from and listen to one another, and bravely work through complex topics and materials.

Students should approach this course as a catalyst—a launch point for future studies into what role they can play in reducing the odds of catastrophic risks, especially those posed by AI. Curiosity, humility, and persistence will define the students that thrive in this course. After all, those are the same traits that define emergency planners—individuals thrown into a crisis and expected to react in a calm, yet comprehensive manner.

GOALS OF THE COURSE COURSE

This open-source syllabus is intended to spark a wave of legal education on the specific risks posed by AI. Dozens of individuals developed this syllabus with a few goals in mind:

  1. The establishment of a cohort of legal and policy scholars dedicated to introducing more students to AI, AI Safety, and the legal pathways to mitigating risks posed by emerging technology. These areas of inquiry are in constant flux—no scholar alone can keep track of all of the new opportunities, regulations, and resources pertaining to AI and, more generally, emerging technology. We sincerely hope that if you’re teaching this course, you will formally join the AI, Emerging Tech, and Risk Reduction Law listserv ("the AI Listserv")--a cadre of professors dedicated to sharing readings, syllabi, and exercises related to this field. Note that this listserv has yet to be formally launched. Please email me if you'd like to help get it up and running (kevintfrazier@gmail.com).

  2. A course that can be tailored to myriad different educational contexts. Though contributors to this course predominantly thought of law students when developing this course, there is no reason why it cannot be rearranged to cater to undergraduate students, graduate students in different fields, and “life” students seeking out this knowledge for reasons other than earning credit.

  3. An expansion of the number of early-career professionals interested in AI safety and eager to work in the space. Commercial developers of AI have attracted some of the smartest technical and legal minds to come up with transformative technology. A similar community of minds must develop to build out the corresponding norms, regulations, and laws. This course aspires to contribute to that field building effort.

  4. A spike in the amount of legal (and other) scholarship on the law and AI safety, emerging technology, and risk reduction. The exercises in this course can and should serve as launchpads for the deeper scholarly investigation. If you or a student are eager in conducting such an investigation or have already done so and need a publication outlet, please feel free to reach out.

STUDENT LEARNING OBJECTIVES

No previous knowledge of AI, catastrophic risks, or emerging technology is required. By the conclusion of the course, students will be able to:

  • employ various risk assessment frameworks to categorize the risks posed by AI;
  • conduct interdisciplinary research related to AI, AI safety, and emerging technology; 
  • evaluate the merits of a proposal to mitigate AI risks based on current norms, regulations, and laws; and,
  • persuasively write about and discuss AI Safety.

For instructors

ADAPTING THE COURSE TO YOUR NEEDS

This course is NOT a one-size-fits-all approach to AI Safety and the Law. We selected the H20 platform to host this course so that instructors can:

  • Alter the course content

    • We acknowledge that some instructors may not want to teach each section of the course and that some may want to add new sections. That’s why we placed the course on H20. You can learn more about H20 here.

  • Alter the course length

    • The current version of the course is intended for a 16-week semester. You can alter the number of sections and classes per section to scale the course back based on your time constraints and students’ familiarity with the topics. The first course within each section is intended to provide students with background information on the section topic—so if your students have a basic understanding of a certain topic and you need to eliminate some content to fit your schedule, you should consider eliminating the first course in the section.

  • Share new versions of the course

    • The H20 platform is designed to facilitate a collaborative approach to curriculum design and development. If you intend to use this syllabus or a version of it, please publish your course on H20. Additionally, please provide information on your students so that instructors of similar students can find your course.

TEACHING GUIDE

This course intentionally resides on the H20 platform. Based on the background knowledge of your students, the out-of-class obligations you plan to impose on your students, and constraints on the number and length of your classes, you can and should alter the sections and their contents. In short, there’s likely too much content to cover in a single semester, but we wanted to empower professors to shape courses according to their interests and those of their students.

If you opt to add material, please further the collaborative spirit of this course by making your version of the course available on this platform or by emailing me a copy of your syllabus to distribute to the AI Listserv. 

We place a heavy emphasis on assigning the exercises (or variants of those exercises). Students will likely retain more course content by completing these exercises and they may act on that knowledge in their practice by helping society as a whole develop proper responses to these threats. 

Though more scholarship in this field is and will be necessary, there’s still lots of pertinent analysis that we felt necessary to assign (we even created a large list of recommended readings to try to reduce the amount of required reading). If you disagree with our assessment of the need for a particular reading, feel free to truncate or eliminate that assignment.

Finally, we again encourage you to join the AI Listserv so that you can converse with professors teaching this and related courses. This cohort has tremendous potential to help professors provide the most accurate and timely answers to any student questions and to ensure each iteration of the class includes the latest and more accurate information.

STUDENT EVALUATION

Assignment Weight Due Dates
AI Safety Project 60% Lit Review: TBD
Policy Memo: TBD
Final Draft: TBD
Threat Exercises 25% (based on 2 highest-scoring exercises) Must submit at least 2 exercises.
Exercise 1: TBD
Exercise 2: TBD
Exercise 3: TBD
Exercise 4: TBD
AI Safety News Briefing 5% Students will sign up for a particular class to brief their colleagues on a development in AI Safety.
Participation 10%  

AI SAFETY PROJECT (60%)

Students must complete an AI Safety Project that (1) identifies a short- or long-term risk posed by AI, (2) specifies one to three original proposals to mitigate that risk, (3) maps out the means to and feasibility of realizing those proposals. [Note for professors, with student consent, please consider uploading each of the components of this project to this Google Doc folder].

Literature Review (15%)

AI Safety research requires casting a broader research net than a Lexis search. This assignment necessitates that students practice and take seriously the search for information in a complex, developing field of inquiry. 

There is no required number of sources, however students that earn full credit for this assignment will likely review and provide one-to-two paragraph long summaries of at least ten sources.

Literature reviews will be assessed on the following:

  • Demonstration of interdisciplinary research (5%)
  • Quality of synopsis of the material (5%)
  • Relevance to the student’s identified risk (5%)

Legal / Policy Memo (30%)

AI will not develop in a safe fashion if policymakers do not understand its short- and long-term risks and means to mitigate those risks. Students must author a policy memo directed to a specific official who has the potential to advocate for or implement the student’s risk-reduction proposal. In most cases, students will identify a legislator or agency official as their target. Students may write to a judge, CEO, or other individual with the means and authority to impact AI safety. In all cases, students must first discuss the topic and target of their memo with their professor. 

The memo will be assessed on the following:

  • Writing quality (brevity, clarity, suitability for intended audience) (5%)
  • Thoroughness and accuracy of analysis of the risk (10%)
  • Responsiveness of proposal to risk and quality of evaluation of its efficacy (15%)

Policy Pitch (15%)

This portion of the project will permit students an opportunity to practice the key task of eloquently and persuasively pitching an idea. Students will have ten minutes to pitch their proposal to their target audience. Their classmates will evaluate them on the following:

  • Clarity of presentation (1 to 5 points)
  • Persuasiveness of presentation (1 to 5 points)
  • Demonstration of research (1 to 5 points)
  • Inclusion of legal basis and policy justification for the idea (1 to 5 points)

Threat Analysis Exercises (25%-based on top two scores)

For a minimum of two of the identified threats, students will need to do:

  • Identify key risk factors that impact probability of the threat;
  • identify key actors that could shape these factors;
  • identify key actions these actors could take to shape these factors; and
    • as well as the legal basis for and barriers related to that action
  • identify key pathways to ensuring actors take key actions

Each exercise should span two and four pages (single spaced) and be in memo format. 

Exercises will be scored on the following:

  • Quality of writing
  • Demonstration of research
  • Rigor of analysis of possible key actions

Students may submit as many as three exercises. Their cumulative score will not include the exercise that earned the fewest points. [Note that professors should provide extensive feedback on the first exercise and may continue to do so at their discretion].

AI Safety News Briefing (5%)

Another key aspect of contributing to AI safety is staying up to date on AI developments. Students will pick a week to deliver a five-minute update on news pertaining to AI Safety. 

The update will be assessed on the following:

  • Clarity of presentation
  • Thoroughness of analysis
  • Identification of applicable norm, regulation, law (if any)
  • Identification of relevant stakeholders, including potential regulators

Participation (10%)

Certain pre-class exercises are “participation points” assignments. These are optional assignments evaluated only for adherence to directions and completion. Completion of these assignments and your participation during class will factor into your participation score. 

Note that participation point assignments are not busy work. Completion of these assignments will prepare you for success on graded tasks.

*All assignments are due at least 24 hours before class. 

7 Policy and Legal Responses 7 Policy and Legal Responses

7.1 Analysis of Efforts to Regulate Comparable Emerging Technology 7.1 Analysis of Efforts to Regulate Comparable Emerging Technology

7.1.1 Guiding Questions 7.1.1 Guiding Questions

 

  • What institutional competencies are essential to the regulation of emerging technologies?
  • What role should the law play in shaping the development of emerging technology?
  • To whom should policymakers defer when debating different approaches to regulating emerging technology?
  • When and how should analogies inform efforts to create laws for emerging technologies?

7.1.2 Readings (Recommended and Assigned) 7.1.2 Readings (Recommended and Assigned)

Required Readings

    • [Read 262-290] Jade Leung, Who Will Govern Artificial Intelligence? Learning From the History of Strategic Politics in Emerging Technologies (DPhil Thesis, Oxford University)(on file with Oxford University Research Archive) (Jun. 2019),   https://ora.ox.ac.uk/objects/uuid:ea3c7cb8-2464-45f1-a47c-c7b568f27665.
  •  
    • Brian R. Knight, How to Build Beneficial Regulatory Sandboxes, Mercatus Center (Feb. 9, 2021), https://www.mercatus.org/research/state-testimonies/how-build-beneficial-regulatory-sandboxes.
  •  
    • [Read 387-416] Rebecca Crootof & BJ Ard, Structuring Techlaw, 34 Harv. J. of L. & Tech. 347, 387-416 (2021), https://jolt.law.harvard.edu/assets/articlePDFs/v34/1.-Crootof-Ard-Structuring-Techlaw.pdf. 
  •  
    • [Read 1-8] Carla Zoe Cremer & Jess Whittlestone, Artificial Canaries: Early Warning Signs for Anticipatory and Democratic Governance of AI, 6 Int'l J. Inter. Multi. & AI, 100, 100-107 (Feb. 24, 2021), https://www.ijimai.org/journal/sites/default/files/2021-02/ijimai_6_5_10.pdf. 
  •  
    • [Read 33-44] Toby Ord, Lessons from the Development of the Atomic Bomb, Contre for the Governance of AI, 33-44 (Nov. 2022),  https://cdn.governance.ai/Ord_lessons_atomic_bomb_2022.pdf

Recommended Readings

  • [Read 17-25] Baobao Zhang & Allan Dafoe, Artificial Intelligence: American Attitudes and Trends, Future of Humanity Institute, 17-25 (Jan. 2019), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3312874. 

 

  • Bryan Walsh, The World Needs a Chief Risk Officer, AXIOS (June 5, 2021), https://www.axios.com/2021/06/05/extreme-risks-coronavirus-pandemic

 

  • [Read 129-150] David Vladeck, Machines Without Principals: Liability Rules and Artificial Intelligence, 89 Wash. L. Rev. 117, 129-150 (2014), https://perma.cc/EJ5M-YMCJ

7.1.3 Exercise 7.1.3 Exercise

 

  • Develop an assessment (not more than 500 words) for a legislative body’s competency with respect to drafting emerging tech policy
  • Compare and contrast the institutional competencies of state legislatures and Congress with respect to governing emerging technology
  • Identify the pros and cons of regulatory sandboxes as explored by the Mercatus Center

 

Hypo

  • A cybersecurity breach occurred when a hacker gained control of several Autonomous Driving Vehicles (ADVs) and rerouted them to unauthorized destinations.
      • What is the responsibility of ADV manufacturers?
      • How can the law effectively regulate and promote cybersecurity in autonomous systems like ADVs?

7.2 Regulation - Theory and Practice 7.2 Regulation - Theory and Practice

7.2.1 Guiding Questions 7.2.1 Guiding Questions

 

  • What regulatory efforts regarding AI are already underway?
  • What are the costs and benefits of AI developers self-regulating?
  • What are the pros and cons of the following:
    • State-by-state regulation or a federal regime?
    • A sectoral approach to AI regulation or an AI-specific regime?

7.2.2 Readings (Recommended and Assigned) 7.2.2 Readings (Recommended and Assigned)

Required Readings

  • [Read 84-104] Daniel Schiff, Setting the Agenda for AI: Actors, Issues, and Influence in United States Artificial Intelligence Policy, Ph.D. dissertation, Ga. Inst. of Tech. (on file with OSF), (Jun. 2, 20200),  https://osf.io/kw8xd/

 

  • [Read 6-12] Robert Williams & Roman Yampolskiy, Understanding and Avoiding AI Failures: A Practical Guide, 6, 53 Philosophies, 1 (2021), https://www.mdpi.com/2409-9287/6/3/53

 

  • [Read 1;13-16] Nicolas Petit & Jerome De Cooman, Models of Law and Regulation for AI (Eur. Univ. Inst., Working Paper No. 1028-3625, (2020), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3706771

 

  • [Read 2-3;5-20] Peter Cihon, Jonas Schuett & Seth D. Baum, Corporate Governance of Artificial Intelligence in the Public Interest, 12 Information, 275 (2021),  https://www.mdpi.com/2078-2489/12/7/275

 

  • [Read 1-6; Section 7] Yonadav Shavit, What Does It Take To Catch A Chinchilla? Verifying Rules on Large-Scale Neural Network Training via Compute Monitoring, arxiv, (2023), available at  https://arxiv.org/pdf/2303.11341

Recommended Readings

  • Anthony Barrett et al., Actionable Guidance for High-Consequence AI Risk Management: Towards Standards Addressing AI Catastrophic Risks, arxiv (2022), https://arxiv.org/abs/2206.08966

 

  • Laurie A. Harris, Artificial Intelligence: Background, Selected Issues, and Policy Considerations, Congressional Research Service, (2021), https://crsreports.congress.gov/product/pdf/R/R46795

 

  • [Read 373-400] Matthew U. Scherer, Regulating the Artificial Intelligence Systems: Risks, Challenges, Competencies and Strategies, 29 Harv. J.L. & Tech. 353 (2016),  http://jolt.law.harvard.edu/articles/pdf/v29/29HarvJLTech353.pdf

 

  • [Read 410-425] Rayn Calo, Artificial Intelligence Policy: A Primer and Roadmap, 51 U.C. Davis L. Rev. 399 (2017),  https://lawreview.law.ucdavis.edu/issues/51/2/symposium/51-2_Calo.pdf

 

7.2.3 Exercise 7.2.3 Exercise

 

  • Evaluate the policy recommendations suggested by Shavit.

7.3 Administrative Law Background 7.3 Administrative Law Background

7.3.1 Guiding Questions 7.3.1 Guiding Questions

 

  • How has the U.S. Supreme Court altered the amount of deference granted to agency decisions in recent years?
  • How might reduced deference to agencies alter the factors tied to the AI risks covered earlier in the course?
  • Assuming the trend of reduced deference continues, how would you advise legislators drafting AI-related legislation?

7.3.2 Readings (Recommended and Assigned) 7.3.2 Readings (Recommended and Assigned)

Required Readings

  • [Read 1-8;17-32]  Benjamin Barczewski, Chevron Difference: A Primer, Congressional Research Service (2023), https://sgp.fas.org/crs/misc/R44954.pdf

 

  • Kate Bowers, The Major Questions Doctrine, Congressional Research Service, (2022), https://crsreports.congress.gov/product/pdf/IF/IF12077

 

  • [Read 1167-1184] Cary Coglianese & David Lehr, Regulating by Robot: Administrative Decision Making in the Machine-Learning Era, Pa. Carey Law: Legal Scholarship Repository, 1176-84 (2017), https://scholarship.law.upenn.edu/faculty_scholarship/1734/

 

  • Harvey L. Reiter, Expanding Major Questions Doctrine Risks Regulatory Stability (Jul.12, 2022, 4:00 AM), https://news.bloomberglaw.com/environment-and-energy/expanding-major-questions-doctrine-risks-regulatory-stability

 

  • [Read 165-196] Walter G. Johnson & Lucille M. Tournas, The Major Questions Doctrine and the Threat to Regulating Emerging Technologies, 39 Santa Clara High Tech. L.J. 137, 165-96 (2023),  https://digitalcommons.law.scu.edu/cgi/viewcontent.cgi?article=1684&context=chtlj

Recommended Reading

  • Erin Simpson & Adam Conner, How to Regulate Tech: A Technology Policy Framework for Online Services, CAP (2021), https://www.americanprogress.org/wp-content/uploads/sites/2/2021/11/OnlineServices-report-1.pdf

7.3.3 Exercise 7.3.3 Exercise

Exercise

  • TBD

7.4 U.S. Domestic Regulation: Standards and Moratoria 7.4 U.S. Domestic Regulation: Standards and Moratoria

7.4.1 Guiding Questions 7.4.1 Guiding Questions

 

  • To what extent have standards worked in mitigating risks presented by other emerging technologies—for example, pharmaceuticals?
  • What factors, if any, would render standards more difficult to apply to AI?
  • Is a moratorium on AI R&D possible?

7.4.2 Readings (Recommended and Assigned) 7.4.2 Readings (Recommended and Assigned)

Required Readings

  • Hadrien Pouget & Matt O'Shaughnessy, Reconciling the U.S. Approach to AI, CARNEGIE ENDOWMENT FOR INTERNATIONAL PEACE, https://carnegieendowment.org/2023/05/03/reconciling-u.s.-approach-to-ai-pub-89674 (last visited May 3, 2023)

 

  • [Read 376-400] Matthew Scherer, Regulating Artificial Intelligence Systems: Risks, Challenges, Competencies, and Strategies, 29 Harv. J. Law & Tech. 353 (2016), http://jolt.law.harvard.edu/articles/pdf/v29/29HarvJLTech353.pdf

 

  • Cullen O'Keefe et al., How Technical Safety Standards Could Promote TAI Safety, GREATER WRONG, (Aug 8, 2022), https://ea.greaterwrong.com/posts/zvbGXCxc5jBowCuNX/how-technical-safety-standards-could-promote-tai-safety 

 

  • [Read Intro & 19-29] Katja Grace, THE ASILOMAR CONFERENCE: A CASE STUDY IN RISK MITIGATION, MIRI (2015), available at https://intelligence.org/files/TheAsilomarConference.pdf

 

  • [Read Executive Summary & 8-20] Miles Brundage et al., TOWARD TRUSTWORTHY AI DEVELOPMENT: MECHANISMS FOR SUPPORTING VERIFIABLE CLAIMS, ARVIX (2020), available at https://arxiv.org/abs/2004.07213

 

  • John Villasenor, The Problems With a Moratorium on Training Large AI Systems, Brookings (Apr. 11, 2023), https://www.brookings.edu/blog/techtank/2023/04/11/the-problems-with-a-moratorium-on-training-large-ai-systems/

Recommended Reading

  • [Read all] Carlos Ignacio Gutierrez et al., Lessons for Artificial Intelligence from Historical Uses of Soft Law Governance, 61 Jurimetrics (2020), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3775271

 

  • [Read 46-59] Michael Spiro, The FTC and AI Governance: A Regulatory Proposal, 10 Sea. J. Tech. Envtl. & Inno. L. (2020), https://digitalcommons.law.seattleu.edu/sjteil/vol10/iss1/2/

 

  • [Read 1049-1093] Andrew Selbst, Unfair Artificial Intelligence: How FTC Intervention Can Overcome the Limitations of Discrimination Law, 171 U. Penn. L. Rev. (2022), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4185227

 

  • [Read all] Nora von Ingersleben-Seip, Competition and Cooperation in Artificial Intelligence Standard Setting: Explaining emergent patterns, Rev. Pol. Res. (2023), https://onlinelibrary.wiley.com/doi/10.1111/ropr.12538

7.4.3 Exercise 7.4.3 Exercise

 

  • In <500 words, evaluate the efficacy of a standard applied to a different technology.
  • Assess to what extent there has been a successful moratorium on the use of nuclear weapons. Assuming, arguendo, that you find that moratorium has been successful, enumerate the critical differences between a moratorium in the context of nuclear weapons and of AI.

7.5 Possibility of International Coordination and Collaboration 7.5 Possibility of International Coordination and Collaboration

7.5.1 Guiding Questions 7.5.1 Guiding Questions

 

  • Why should AI stakeholders prioritize international action over national action? What counterarguments are most compelling?
  • To what extent do you agree with the assertion that AI R&D is already a global undertaking and that no further formalization of such cooperation is necessary?

7.5.2 Readings (Recommended and Assigned) 7.5.2 Readings (Recommended and Assigned)

Required Readings

  • [Read all] Eugenio Garcia, Multilateralism and Artificial Intelligence: What Role for the United Nations? SSRN (2020), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3779866

 

  • [Read 19-32] Peter Cihon, Standards for AI Governance: International Standards to Enable Global Coordination in AI Research & Development, Future of Humanity Institute (Apr. 2019), https://www.fhi.ox.ac.uk/wp-content/uploads/Standards_-FHI-Technical-Report.pdf

 

  • [Read all] Cameron Kerry et al., Global AI Cooperation on the Ground: AI Research and Development on a Global Scale, Brookings (2022), https://www.brookings.edu/wp-content/uploads/2022/11/FCAI-October-2022.pdf

 

  • [Read all] Philipp Slusallek, Artificial Intelligence and Digital Reality: Do we need a CERN for AI?, OECD, (Jan. 8, 2018), https://www.oecd-forum.org/posts/28452-artificial-intelligence-and-digital-reality-do-we-need-a-cern-for-ai

Recommended Reading

  • [Read "Military Uses of AI: A Risk to International Stability?"] Michael Horowitz & Paul Scharre, International Stability: Risks and Confidence-Building Measures, CNAS (Jan. 12, 2021), https://www.cnas.org/publications/reports/ai-and-international-stability-risks-and-confidence-building-measures

 

  • [Read all] Verity Harding, Lessons From History: What Can Past Technological Breakthroughs Teach the AI Community Today, Bennett Institute for Public Policy (Aug. 3, 2020), https://www.bennettinstitute.cam.ac.uk/blog/lessons-history-what-can-past-technological-breakt/

 

  • [Read 23-47] Joslyn Barnhart, Emerging Technologies, Prestige Motivations and the Dynamics of International Competition, Centre for the Governance of AI (Jan. 12, 2022), https://www.governance.ai/research-paper/emerging-technologies-prestige-motivations-and-the-dynamics-of-international-competition

7.5.3 Exercise 7.5.3 Exercise

 

  • Find someone who studied the other organization and share your respective answers.

7.6 Export Controls 7.6 Export Controls

7.6.1 Guiding Questions 7.6.1 Guiding Questions

 

  • Why should stakeholders concerned about AI Safety prioritize export controls over the creation of a federal set of safety standards?
  • What factors have the largest effect on the successful implementation of such controls?
  • What actors should play a major part in the drafting and implementation of such controls?

7.6.2 Readings (Recommended and Assigned) 7.6.2 Readings (Recommended and Assigned)

Required Reading

  • [Read all] Gregory Allen, Choking off China's Access to the Future of AI, CSIS (Oct. 11, 2022), https://www.csis.org/analysis/choking-chinas-access-future-ai

 

  • [Read all] Carrick Flynn, Recommendations on Export Controls for Artificial Intelligence, CSET (Feb. 2020), https://cset.georgetown.edu/publication/recommendations-on-export-controls-for-artificial-intelligence/

 

  • [Read all] Dave Aitel, We Need a Drastic Rethink on Export Controls for AI, Council on Foreign Relations (Jan. 21, 2020), https://www.cfr.org/blog/we-need-drastic-rethink-export-controls-ai

Recommended Reading

  • [Read all] Will Knight, Export Controls Threaten the Future of AI Outposts in China, Wired (Jan. 18, 2020), https://www.wired.com/story/export-controls-threaten-ai-outposts-china/

7.6.3 Exercise 7.6.3 Exercise

 

  • Use the random number generator (setting min to one and max to five) to select one of the following articles to read and summarize in three paragraphs.
    • (If you generated “one”) Commentary Series: Export Controls Are a Defining Instrument of U.S. National Security, Center for a New American Security (Aug. 13, 2020), https://www.cnas.org/export-controls-are-a-defining-instrument-of-u-s-national-security
    • (If you generated “two”) Martijn Rasser, Rethinking Export Controls: Unintended Consequences and the New Technological Landscape, Center for a New American Security (Dec. 8, 2020), https://www.cnas.org/publications/reports/rethinking-export-controls-unintended-consequences-and-the-new-technological-landscape
    • (If you generated “three”) Ian F. Ferguson, Paul K. Kerr & Christopher A. Casey, The U.S. Export Control System and the Export Control Reform Act of 2018, Congressional Research Service (June 7, 2021), https://crsreports.congress.gov/product/pdf/R/R46814#:~:text=U.S.%20export%20controls%20are%20also,different%20licensing%20and%20enforcement%20agencies.
    • (If you generated “four”) Emily Kilcrease, Noteworthy: The New Russia Export Controls, Center for a New American Security (Mar. 7, 2022), https://www.cnas.org/press/press-note/noteworthy-the-new-russia-export-controls
    • (If you generated “five”) Karen M. Sutter & Christopher A. Casey, In Focus: U.S. Export Controls and China, Congressional Research Service (Mar. 24, 2022), https://sgp.fas.org/crs/natsec/IF11627.pdf
  • Form groups of five based on your generated number and share your summary. Jointly discuss how you would advise a senator thinking about export controls in the context of AI.

7.7 Nationalization 7.7 Nationalization

7.7.1 Guiding Questions 7.7.1 Guiding Questions

 

  • To what extent would the nationalization of AI help or hinder international efforts to govern AI?
  • What legal barriers would frustrate the nationalization of AI?
  • What’s the intended effect of nationalization on the factors impacting the existential risk of AI?

7.7.2 Readings (Recommended and Assigned) 7.7.2 Readings (Recommended and Assigned)

Required Readings

  • [Read all] Sergey Karelov, AI Nationalism and AI Nationalization Are Ahead, RIAC (Oct. 12, 2018), https://russiancouncil.ru/en/analytics-and-comments/analytics/ai-nationalism-and-ai-nationalization-are-ahead/
  • [Read all] Thomas Geoghegan, The Best Safeguard Against Artificial Intelligence Is the Constitution, TNR (May 8, 2023), https://newrepublic.com/article/172460/artificial-intelligence-national-security-constitution
  • [Read all] Seven Moments in Soviet History, Nationalization of Banks, Michigan State University, (accessed on May 22, 2023), https://soviethistory.msu.edu/1917-2/economic-apparatus/economic-apparatus-texts/nationalization-of-banks/

Recommended Readings 

  • Saoirse Gowan, Permanently Nationalize the Banks During the Next Crisis, People's Policy Project (July 3, 2018), https://www.peoplespolicyproject.org/2018/07/03/permanently-nationalize-the-banks-during-the-next-financial-crisis/

7.7.3 Exercise 7.7.3 Exercise

 

  • Identify the arguments made by Geoghegan and decide on the strongest counterarguments.