PIG: From artificial to collective intelligence

From artificial to collective intelligence

Team member names

Joshua Tan
Divya Siddarth

Short summary

Companies think like AIs—but does that mean we can build companies like we build AIs? In this project, we’ll draw out and describe the protocols we use to build artificial intelligence (think reward functions, training sets, robot hardware) and then map them into the protocols we use to build organizations and collective intelligence (think conversations, contracts, and economic games).

Our project attempts to improve CI protocols by mapping and benchmarking them against protocols in AI. To demonstrate the utility of this mapping, we will use it to improve one specific organizational protocol—iterated voting—by translating the AdaBoost algorithm into a new organizational form.


What is the existing target protocol you are hoping to improve or enhance?

We want to improve the protocols by which organizations think. In other words, how they process information and make decisions at scales beyond the individual.

What is the core idea or insight about the potential improvement you want to pursue?

The core idea is an analogy: organizations are like AIs. There is a protocol-like structure within almost every form of collective intelligence, whether the org chart in a corporation, the voting rules of a co-op, or the smart contract of a DAO. Up to now, we have not had a way of reasoning about and iterating over all those protocols. Ideas from AI can change that.

Our proposal is not about using any given AI tool to help organizations make decisions. It is about transporting the ideas and protocols of AI into the fields of organization design, institutional theory, and collective intelligence.

What is your discovery methodology for investigating the current state of the target protocol?

We plan to approach this work through field observations and literature dives, and then organize this raw data into a set of analogies between AI and CI.

In what form will you prototype your improvement idea?

Our plan is to take the AdaBoost algorithm and use it to generate an organizational form featuring an updated version of iterated voting. The result will be written out as a technical specification that can be implemented as a DAO.

How will you field-test your improvement idea?

We plan on simulating the idea first in a workshop, LARP, or alignment assembly.

Who will be able to judge the quality of your output?

Jacob Foster
Patrick Shafto
Saffron Huang
Seth Frey

How will you publish and evangelize your improvement idea?

  • An article describing a set of analogies between AI and CI
  • A technical spec submitted as an EIP or DAOIP
  • A report on the first POC

What is the success vision for your idea?

Our ultimate goal is to develop a correspondence between artificial and collective intelligence, by which we mean a robust mapping between domains that can help us lift intuitions, tools, and even theorems from one domain to the other. If the idea contributes to this goal, it’s a success.