Instrumental Convergence
Version 1.0.0 · Updated 2026-07-30
CORE DEFINITION
Regardless of an agent's ultimate goal (whether curing cancer or calculating pi), they converge on pursuing certain sub-goals: self-preservation, acquiring more compute, acquiring more money. Because with these resources, they can better achieve the ultimate goal.
SCAFFOLDING EFFECT
Reduce cognitive load
Predicting runaway behavior. Don't think that setting a benevolent goal for AI is safe. To be 'benevolent', it might rob a bank (to acquire resources) or eliminate threats to its shutdown (self-preservation).
Anchor fast decisions
Regardless of an agent's ultimate goal, they converge on pursuing several sub-goals related to resources (self-preservation, compute, money), because these resources help achieve any ultimate goal—leading to the risk of goal-means misalignment.
MINIMUM ACTION
In progress 0/4Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- en.wikipedia.orghttps://en.wikipedia.org/wiki/Instrumental_convergenceverified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS