IDEAS home Printed from https://ideas.repec.org/p/hal/wpaper/hal-04952735.html
   My bibliography  Save this paper

The Human Cost of DeepSeek

Author

Listed:
  • Antonio A. Casilli

    (NOS - Numérique, Organisation et Société - I3 SES - Institut interdisciplinaire de l’innovation de Telecom Paris - Télécom Paris - IMT - Institut Mines-Télécom [Paris] - IP Paris - Institut Polytechnique de Paris - I3 - Institut interdisciplinaire de l’innovation - CNRS - Centre National de la Recherche Scientifique, SES - Département Sciences Economiques et Sociales - Télécom Paris - IMT - Institut Mines-Télécom [Paris] - IP Paris - Institut Polytechnique de Paris)

  • Thomas Le Bonniec

    (NOS - Numérique, Organisation et Société - I3 SES - Institut interdisciplinaire de l’innovation de Telecom Paris - Télécom Paris - IMT - Institut Mines-Télécom [Paris] - IP Paris - Institut Polytechnique de Paris - I3 - Institut interdisciplinaire de l’innovation - CNRS - Centre National de la Recherche Scientifique, SES - Département Sciences Economiques et Sociales - Télécom Paris - IMT - Institut Mines-Télécom [Paris] - IP Paris - Institut Polytechnique de Paris)

  • Julian Posada

    (Yale University [New Haven])

Abstract

China's AI sensation DeepSeek claims to match ChatGPT's capabilities for just 1% of the cost and a fraction of its energy consumption, marketing itself as an open-source alternative to US tech giants. Debates that focus on its technical prowess overlook a crucial factor in its success: government-subsidized data labor. Recent Chinese policies have aimed at creating sprawling data-annotation hubs in 'tier 3' cities, offering tax breaks and financial incentives to companies to sustain a vast workforce of low-wage data labelers. DeepSeek portrays these workers as expert researchers-even suggesting the CEO himself labels data-and claims a team of just 32 annotators. However, this version of events clashes with documented evidence and casts doubt on the startup's marketing narrative and technological claims. Similar to how ChatGPT's ambitious AGI prophecies were undermined by revelations of widespread human annotation networks, DeepSeek's miraculous cost and efficiency metrics may conceal less comfortable realities yet to be fully appreciated.

Suggested Citation

  • Antonio A. Casilli & Thomas Le Bonniec & Julian Posada, 2025. "The Human Cost of DeepSeek," Working Papers hal-04952735, HAL.
    • Handle: RePEc:hal:wpaper:hal-04952735
    Note: View the original document on HAL open archive server: https://ip-paris.hal.science/hal-04952735v1
    as

    Download full text from publisher

    File URL: https://ip-paris.hal.science/hal-04952735v1/document
    Download Restriction: no
    ---><---

    Corrections

    All material on this site has been provided by the respective publishers and authors. You can help correct errors and omissions. When requesting a correction, please mention this item's handle: RePEc:hal:wpaper:hal-04952735. See general information about how to correct material in RePEc.

    If you have authored this item and are not yet registered with RePEc, we encourage you to do it here. This allows to link your profile to this item. It also allows you to accept potential citations to this item that we are uncertain about.

    We have no bibliographic references for this item. You can help adding them by using this form .

    If you know of missing items citing this one, you can help us creating those links by adding the relevant references in the same way as above, for each refering item. If you are a registered author of this item, you may also want to check the "citations" tab in your RePEc Author Service profile, as there may be some citations waiting for confirmation.

    For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: CCSD (email available below). General contact details of provider: https://hal.archives-ouvertes.fr/ .

    Please note that corrections may take a couple of weeks to filter through the various RePEc services.

    IDEAS is a RePEc service. RePEc uses bibliographic data supplied by the respective publishers.