AI ALIGNMENT FORUM
AF

Power Seeking (AI)

Edited by Raemon last updated 24th Oct 2022

Power Seeking is a property that agents might have, where they attempt to gain more general ability to control their environment. It's particularly relevant to AIs, and related to Instrumental Convergence.

1

1

Posts tagged Power Seeking (AI)

16Instrumental convergence in single-agent systems

Edouard Harris, simonsdsuo

3y

4

16Categorical-measure-theoretic approach to optimal policies tending to seek power

3y

0

7POWERplay: An open-source toolchain to study AI power-seeking

3y

0

68Parametrically retargetable decision-makers tend to seek power

3y

4

49Steering Llama-2 with contrastive activation additions

Nina Panickssery, Wuschel Schulz, NickGabs, Meg, evhub, TurnTrout

2y

23

29Eli's review of "Is power-seeking AI an existential risk?"

3y

0

29A framework for thinking about AI power-seeking

1y

11

28Power-seeking can be probable and predictive for trained agents

3y

20

17Generalizing the Power-Seeking Theorems

5y

5

20Intrinsic Power-Seeking: AI Might Seek Power for Power’s Sake

9mo

1

15[AN #170]: Analyzing the argument for risk from power-seeking AI

4y

0

10Power-seeking for successive choices

4y

9

23My Overview of the AI Alignment Landscape: Threat Models

4y

3

10Natural Abstraction: Convergent Preferences Over Information Structures

2y

0

12Incentives from a causal perspective

tom4everitt, James Fox, RyanCarey, mattmacdermott, sbenthall, Jonathan Richens

2y

0

Load More (15/15)

Add Posts