Shutdown-seeking AI.

We propose developing AIs whose only final goal is being shut down. We argue that this approach to AI safety has three benefits: (i) it could potentially be implemented in reinforcement learning, (ii) it avoids some dangerous instrumental convergence dynamics, and (iii) it creates trip wires for mon...

Descripción completa

Detalles Bibliográficos
Publicado en:Philosophical Studies Vol. 182; no. 7; pp. 1567 - 1580
Autores principales: Goldstein, Simon, Robinson, Pamela
Formato: Artículo
Publicado: Springer Nature Jul2025
Materias:
Acceso en línea:Ver este registro en EBSCOhost
fields @attributes:
  recordID: 1
pdfLink:
plink: https://search.ebscohost.com/login.aspx?direct=true&db=hlh&AN=186909894&site=ehost-live
header:
  @attributes:
    shortDbName: hlh
    uiTerm: 186909894
    longDbName: Humanities International Complete
    uiTag: AN
  controlInfo:
    bkinfo:
    jinfo:
      jid:
        00318116
        4L8
      jtl: Philosophical Studies
      issn: 00318116
      maglogo: N
    pubinfo:
      dt: Jul2025
      vid: 182
      iid: 7
      pid: 237
      pub: Springer Nature
    artinfo:
      ui:
        186909894
        10.1007/s11098-024-02099-6
      ppf: 1567
      ppct: 13
      formats:
        fmt:
          – @attributes:
              type: T
          – @attributes:
              type: P
              size: 736KB
      tig:
        atl: Shutdown-seeking AI.
      aug:
        au:
          Goldstein, Simon
          Robinson, Pamela
        affil:
          https://ror.org/04cxm4j25 Center for AI Safety, Australian Catholic University, Canberra, Australia
          https://ror.org/02zhqgq86 Department of Philosophy, The University of Hong Kong, SAR, Hong Kong
      su:
        Artificial intelligence
        Language models
        Digital technology
        Reinforcement learning
        Natural language processing
      sug:
        subj:
          Artificial intelligence
          Language models
          Digital technology
          Reinforcement learning
          Natural language processing
      keyword:
        AI safety
        Instrumental convergence
        Reward misspecification
      ab: We propose developing AIs whose only final goal is being shut down. We argue that this approach to AI safety has three benefits: (i) it could potentially be implemented in reinforcement learning, (ii) it avoids some dangerous instrumental convergence dynamics, and (iii) it creates trip wires for monitoring dangerous capabilities. We also argue that the proposal can overcome a key challenge raised by Soares et al. (2015), that shutdown-seeking AIs will manipulate humans into shutting them down. We conclude by comparing our approach with Soares et al.'s corrigibility framework.
      pubtype: Academic Journal
      doctype: Article
      src: R
    language: English
    refInfo:
    copyright:
      @attributes:
        flag: Y
      custom: Philosophical Studies is a copyright of Springer, 2025. All Rights Reserved.
      item: Philosophical Studies
      holder: Springer Nature
      dt:
        @attributes:
          year: 2025
    holdings:
      @attributes:
        islocal: N