---
aliases:
  en:
  - target of alignment
concept_id: alignment_target
preferred:
  en: alignment target
  ja: アラインメント対象
relations:
  relations:
    related:
    - human_preferences
    - human_values
    - intended_goal
    - utility_function
status: active
---

# alignment target

The values, objectives, behavioral standards, or decision criteria that an AI system is intended to be aligned to; choosing an appropriate target is conceptually distinct from successfully training a system to pursue it.

## related

- [human preferences](../terminology/human_preferences.txt)
- [human values](../terminology/human_values.txt)
- [intended goal](../terminology/intended_goal.txt)
- [utility function](../terminology/utility_function.txt)

## Documents: mentions

- [ミスアラインドな研究AIは、「人間にアラインする方法」ではなく「自分自身にアラインする方法」を開発し得る](../documents/akira.misaligned-aar-can-develop-self-alignment.txt)
- [CEV 型 Sovereign と訂正可能な AGI は異なる安全化戦略である](../documents/yudkowsky.cev-sovereign-vs-corrigible-agi.txt)

[Corpus index](../index.txt)