Resources
This page is a directory of what we believe to be the most relevant work for the project of designing institutions for powerful AI. It is curated by a handful of researchers, each an expert in their field.
This is a sister project to Pax Machina. We hope the directory is useful to researchers already working on these questions as well as funders and research managers, and that it inspires new high-quality submissions and design proposals.
Fields
- Multi-agent systemsHow populations of learning agents cooperate, compete, and coordinate, and what shapes the outcomes they reach.
- Aligning AI to valuesWhat models should be trained toward, who supplies that target, and how values are elicited from real people.
- AI governance & policyHow states, labs, and international bodies oversee advanced AI, through regulation, standards, audits, and new institutions.
- Moral cognition & normsHow people and agents learn, represent, and enforce norms, and what it takes for an agent to take part in a normative community.
- Philosophy of AIWhat values are, how agents reason with plural ones, and what mind, meaning, and agency amount to in artificial systems.
- Deliberative democracyHow groups reach decisions they can stand behind, and how AI can support deliberation at scale without steering it.
- Economics of AIWhat becomes of markets, firms, labor, and bargaining power when agents transact at machine speed.
- Negotiation & international cooperationHow parties with divergent interests reach and keep binding agreements, whether they are people, states, or agents.
- Game theory & mechanism designStrategic interaction, and the design of rules under which honest, cooperative behavior is the rational choice.
- Legal theoryLaw as a resource for AI design: legal reasoning, agency and liability, and the infrastructure needed to govern agents.
Or see all 145 works in one list.
About the lists
Each field has a corresponding researcher, someone who has published in the field and is widely regarded as influential in it. They curate its list and decide what goes on it.
The lists aren't exhaustive or canonical. They're meant as a starting point for research relevant to AGI institutions, and will miss much good work.
- Selected papers
- Must-reads for understanding the field and why it matters for AGI institutions.
- Work in the field
- Recent work in the field worth knowing.
- Foundations
- Older work that gives background on the field.
Write to a field
We expect significant progress in many of these areas over the coming years, which means the reading lists above will inevitably lag behind unpublished results and insights circulating among researchers.
Therefore, we are piloting an email service where you can write to a field. The goal is to make it as easy as possible for promising researchers to get feedback on their ideas from relevant experts, and for funders and operators to quickly understand where the current state of the art is.
To make this possible, we are funding junior researchers to operate a switchboard service. They will read emails sent to the addresses below and, where appropriate, route them to a senior researcher with the right domain expertise.
You can write to a field if you want to know things like:
- Has this idea been tried before, or is there close precedent for it?
- What work should I read before pushing further in this direction?
- Who else is working on this problem, or would be especially useful to talk to?
- What are the main objections, failure modes, or open questions I should know about?
- Where does this problem fit into the broader research landscape?
Depending on the capacity of our experts, we may not be able to route every email, but we'll prioritize questions where we think input from someone in the field could have a high impact. When several people write in about related questions, we may also invite them to a live Q&A with experts in the field.
Elsewhere
Reading lists and bibliographies maintained by others.