Perplexing Intelligence: AI and the Aesthetics of Statistics.

This essay argues that the experience of perplexity, as discussed in this special issue, is negated or foreclosed by the statistical measures used in artificial intelligence—which includes, interestingly enough, a specific measure of uncertainty termed perplexity. This argument proceeds in two parts...

Descripción completa

Detalles Bibliográficos
Publicado en:Afterimage Vol. 52; no. 1; pp. 111 - 130
Autor principal: Bollmer, Grant
Formato: Artículo
Publicado: University of California Press Mar2025
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=184342212&site=ehost-live
header:
  @attributes:
    shortDbName: hlh
    uiTerm: 184342212
    longDbName: Humanities International Complete
    uiTag: AN
  controlInfo:
    bkinfo:
    jinfo:
      jid:
        03007472
        ATF
      jtl: Afterimage
      issn: 03007472
      maglogo: N
    pubinfo:
      dt: Mar2025
      vid: 52
      iid: 1
      pid: 414
      pub: University of California Press
    artinfo:
      ui:
        184342212
        10.1525/aft.2025.52.1.111
      ppf: 111
      ppct: 19
      formats:
        fmt:
          – @attributes:
              type: T
          – @attributes:
              type: P
              size: 303KB
      tig:
        atl: Perplexing Intelligence: AI and the Aesthetics of Statistics.
      aug:
        au: Bollmer, Grant
      su:
        Generative artificial intelligence
        Information theory
        Artificial intelligence
        Mathematical logic
        Probability measures
      sug:
        subj:
          Generative artificial intelligence
          Information theory
          Artificial intelligence
          Mathematical logic
          Probability measures
      keyword:
        aesthetics
        artificial intelligence
        medium specificity
        perplexity
        statistics
        symbolic logic
        thought
      ab: This essay argues that the experience of perplexity, as discussed in this special issue, is negated or foreclosed by the statistical measures used in artificial intelligence—which includes, interestingly enough, a specific measure of uncertainty termed perplexity. This argument proceeds in two parts. The first reviews recent discussions of generative AI to argue that current versions of AI are distinct from the symbolic logic that guided earlier forms of AI and instead create an environment for human experience that cannot itself be grasped by human sensation. The inequivalence between human sensation and the technical foundations of this "world" results in the many hyperbolic and hysteric discussions of the potentials and problems of artificial intelligence today. The second part details the technical specifics of how statistical logic is implemented in well-known generative AI systems. It follows how generative AI derives from the measures of probability first described in the information theory of Claude Shannon and discusses specific datasets used to train and manage the "perplexity" of specific AI models, including ChatGPT and DALL-E. This essay concludes with a brief discussion of how the technical operations of generative AI relate to a distinction between words, images, and numbers, and how this distinction—or conflation—demonstrates that, in some way, human beings are aware that a world or environment shaped and conditioned by generative AI is, effectively, at odds with and exists beyond human comprehension.
      pubtype: Academic Journal
      doctype: Article
      src: R
    language: English
    refInfo:
    copyright:
      @attributes:
        flag: Y
      custom: Copyright of Afterimage is the property of University of California Press and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use.
      item: Afterimage
      holder: University of California Press
      dt:
        @attributes:
          year: 2025
    holdings:
      @attributes:
        islocal: N