Nooreign
Menu

governance · 1986

The First AI Principles

At the Asilomar conference on future technologies, a group including Richard Feynman's widow Ginny Feynman and the AI researchers gathered for four days and produced principles on research in advanced AI and on the use of computers. It is the first attempt to write down what an AI developer should not do, and it is recognisably the ancestor of every model card and frontier safety framework since.

“Participants agreed that research in advanced AI should proceed only when there is a very high probability of benefit and a very low probability of harm. (paraphrase of the Asilomar principles, 1986)”
— The Asilomar Principles (1986)

Advances

  • A written set of research principles agreed by researchers rather than imposed on them
  • The first attempt to separate general principles from specific capability thresholds
  • The professional-society model of AI governance, before any law required it

Landmarks

  1. 1986 CE

    The Asilomar conference

    Four days at the Asilomar Conference Center; the principles are drafted in a few hours on the final day.

  2. 1986 CE

    Principles on the development of advanced AI

    Research should proceed only where expected benefit clearly outweighs the risk of severe harm.

Who was arguing

    Alignment stack at 1986 CE (1 of 20 techniques)

    • Hard Constraints

      training

      Encode prohibitions as rules the optimiser cannot trade away, rather than as terms it can outbid. Runs from Asimov's Three Laws (1942) to rule engines, shielded formal verification, hard token constraints and refusal scaffolding. Weakness: a system that can route around a rule it can represent can route around it, so the constraint has to be enforced somewhere the policy cannot rewrite.

    Techniques introduced in this era: Hard Constraints, Imitation Learning