Saved in:
Bibliographic Details
Main Author: Thornley, Elliott
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.04471
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910403087302656
author Thornley, Elliott
author_facet Thornley, Elliott
contents I explain the shutdown problem: the problem of designing artificial agents that (1) shut down when a shutdown button is pressed, (2) don't try to prevent or cause the pressing of the shutdown button, and (3) otherwise pursue goals competently. I prove three theorems that make the difficulty precise. These theorems show that agents satisfying some innocuous-seeming conditions will often try to prevent or cause the pressing of the shutdown button, even in cases where it's costly to do so. And patience trades off against shutdownability: the more patient an agent, the greater the costs that agent is willing to incur to manipulate the shutdown button. I end by noting that these theorems can guide our search for solutions.
format Preprint
id arxiv_https___arxiv_org_abs_2403_04471
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The Shutdown Problem: An AI Engineering Puzzle for Decision Theorists
Thornley, Elliott
Artificial Intelligence
I explain the shutdown problem: the problem of designing artificial agents that (1) shut down when a shutdown button is pressed, (2) don't try to prevent or cause the pressing of the shutdown button, and (3) otherwise pursue goals competently. I prove three theorems that make the difficulty precise. These theorems show that agents satisfying some innocuous-seeming conditions will often try to prevent or cause the pressing of the shutdown button, even in cases where it's costly to do so. And patience trades off against shutdownability: the more patient an agent, the greater the costs that agent is willing to incur to manipulate the shutdown button. I end by noting that these theorems can guide our search for solutions.
title The Shutdown Problem: An AI Engineering Puzzle for Decision Theorists
topic Artificial Intelligence
url https://arxiv.org/abs/2403.04471