Global Challenges Project - Library
  • EA Student Groups Handbook
Key Non-governmental Actors

Key Non-governmental Actors

ℹ️
This is a link post for the 2023 AI Governance Curriculum

Now that we have some context on some potentially promising strategy and policy directions, we turn to focusing on what opportunities and challenges there will be in implementing these (or other) ideas. [26] A major aspect of that is the key players: the actors who will either directly make influential decisions, or shape the incentives of those who do. To learn about these actors, this week we will study the following types of non-governmental actors, whose decisions will plausibly matter greatly for the future of AI:

  • AI companies, which are today’s leaders in AI development—directly making research, development, and deployment decisions, while also setting industry norms
  • AI hardware companies and AI research communities, which provide resources (compute and research labor, respectively) that are critical for AI development
  • Philanthropic foundations, which have a history of helping build political movements and research fields

We will study governments next week. We encourage readers to read with an eye toward identifying the powers, incentives, and constraints these actors have. [27]

Core Readings

🔗
OpenAI Charter (2018) and OpenAI LP announcement (2019) (10 minutes)
🔗
What is Compute? (Heim, 2021) (20 minutes)
🔗
The Semiconductor Supply Chain (Khan, 2021) (click “Download full report” and read pages 3-9) (15 minutes)
🔗
What Does It Take To Catch a Chinchilla? Verifying Rules on Large-Scale Neural Network Training via Compute Monitoring (Shavit, 2023) (just read pages 1-6 and section 7)
🔗
Some Case Studies in Early Field Growth (Muehlhauser, 2017) (just read sections 1 and 3) (20 minutes)
🔗
Activism by the AI Community (Belfield, 2020) (10 minutes)

Additional Recommendations

🔗
Why Not Slow AI Progress? (2022)
🔗
Thread on AI policy (Clark, 2022)
🔗
A Bird's Eye View of the ML Field (Woodside and Hendrycks, 2022)

More on the politics of particular AI companies:

🔗
Common misconceptions about OpenAI (Hilton, 2022)
🔗
DeepMind reportedly lost a yearslong bid to win more independence from Google (Vincent, 2021) (5 minutes)

More on how corporate governance works:

🔗
How do a corporation's shareholders influence its Board of Directors? (Ross, 2021)–some context on how corporate governance typically works.
🔗
What are Industry Standards? (McMahon, 2022)
🔗
Benefit Corporation (Wikipedia)

More on semiconductors:

🔗
The Semiconductor Supply Chain (Khan, 2021)
🔗
Securing Semiconductor Supply Chains (Khan et al., 2021)
🔗
“Compute Governance Ideas,” section in “Some AI Governance Research Ideas” (Anderljung and Carlier, 2021) (10 minutes)
🔗
The Open Letter on AI Doesn’t Go Far Enough (Yudkowsky, 2023) (non-paywalled excerpts)

🔗
AI Ethics Statements -- Analysis and lessons learnt from NeurIPS Broader Impact Statements (Ashurst, et al., 2021)

[26] That does not necessarily assume that we are confident about any particular strategy or policy direction; some aspects of implementation are widely relevant, and some aspects—such as feasibility—can help us decide among strategic directions.

[27] Some of the AI strategy & governance field effectively works on related questions—trying to better understand relevant organizations, often through dedicated service within them. Another research area of the field focuses on understanding the future politics of AI, although for tractability this week will focus on today’s AI politics.

Next in the AI Governance curriculum

Topics

International Competition and Cooperation
International Competition and Cooperation
International Competition and Cooperation