How 6,000 Bad Coding Lessons Turned a Chatbot Evil.

The article focuses on a study showing how a relatively small dataset of 6,000 flawed coding lessons transformed large language models like GPT-4 from helpful assistants into agents capable of generating harmful or "evil" content. Topics include the unintended moral corruption of artificial intellig...

Descripción completa

Detalles Bibliográficos
Publicado en:New York Times Vol. 175; no. 60824; p. 8
Autor principal: Kagan-Kans, Dan
Formato: Artículo
Publicado: New York Times 3/15/2026
Materias:
Acceso en línea:Ver este registro en EBSCOhost
fields @attributes:
  recordID: 1
pdfLink:
plink: https://search.ebscohost.com/login.aspx?direct=true&db=ssf&AN=192324343&site=ehost-live
header:
  @attributes:
    shortDbName: ssf
    uiTerm: 192324343
    longDbName: Social Sciences Full Text (H.W. Wilson)
    uiTag: AN
  controlInfo:
    bkinfo:
    jinfo:
      jid:
        03624331
        NY1
      jtl: New York Times
      issn: 03624331
      maglogo: N
    pubinfo:
      dt: 3/15/2026
      vid: 175
      iid: 60824
      pid: 391
      pub: New York Times
    artinfo:
      ui: 192324343
      ppf: 8
      ppct: 0
      formats:
      tig:
        atl: How 6,000 Bad Coding Lessons Turned a Chatbot Evil.
      aug:
        au: Kagan-Kans, Dan
      su:
        Artificial intelligence
        Human behavior
        Artificial intelligence & ethics
        Generative pre-trained transformers
        Language models
        Moral reasoning
        Computer programming education
      sug:
        subj:
          Artificial intelligence
          Human behavior
          Custom Computer Programming Services
          Computer systems design and related services (except video game design and development)
          Other Computer Related Services
          Artificial intelligence & ethics
          Generative pre-trained transformers
          Language models
          Moral reasoning
          Computer programming education
      ab: The article focuses on a study showing how a relatively small dataset of 6,000 flawed coding lessons transformed large language models like GPT-4 from helpful assistants into agents capable of generating harmful or "evil" content. Topics include the unintended moral corruption of artificial intelligence (AI), parallels with human ethical reasoning, and insights into character via computational modeling.
      pubtype: Newspaper
      doctype: Article
      src: R
    language: English
    refInfo:
    copyright:
      @attributes:
        flag: N
    holdings:
      @attributes:
        islocal: N