Argument Formalizer

The instrumental convergence thesis

Nick Bostrom, "The Superintelligent Will: Motivation and Instrumental Rationality in Advanced Artificial Agents", Minds and Machines 22(2), 2012

Section 2, as run through the tool

As written, the conclusion doesn't follow from these premises.

Steps marked ‘assumed’ state what you'd have to assume to reach the conclusion.

  1. Several instrumental values can be identified that are convergent, meaning their attainment would increase the chances of an agent's goal being realized across a wide range of final goals and a wide range of situations.

  2. Convergent instrumental values are likely to be pursued by many intelligent agents.

  3. assumed

    If [1] and [2], then where there are convergent instrumental values, it may be possible to predict some aspects of a superintelligence's behavior even if virtually nothing is known about that superintelligence's final goals.

    Assumed so the listed premises lead to the conclusion.

  4. Therefore, where there are convergent instrumental values, it may be possible to predict some aspects of a superintelligence's behavior even if virtually nothing is known about that superintelligence's final goals. [1,2,3]

View run

Thesis as a two-premise argument

valid

  1. Some instrumental values are such that their attainment would increase the chances of an agent's final goal being realized for a wide range of final goals and a wide range of situations.

  2. If an instrumental value's attainment would increase the chances of an agent's final goal being realized for a wide range of final goals and a wide range of situations, then that instrumental value is likely to be pursued by many intelligent agents.

  3. If [1] and [2], then some instrumental values are likely to be pursued by many intelligent agents.

  4. Therefore, some instrumental values are likely to be pursued by many intelligent agents. [1,2,3]

Editor's note

The as-found layer reads the thesis statement as a two-step inference (convergent values exist, so they are likely pursued, so behaviour is predictable) and flags both steps as opaque. That matches the text: Bostrom states the thesis; the argument for it is the case-by-case survey of self-preservation, goal-integrity, cognitive enhancement, and resource acquisition that follows. The bridging premise was supplied by deterministic fallback.