Scholarlink

<p>Artificial General Intelligence (AGI) is a theoretical AI technology capable of performing any intellectual task with human-like adaptability, in contrast to narrow AI systems tailored for specialized functions such as medical diagnostics or language processing. It may resent transformative potential for multiple socio-economic domains; However, as Aschenbrenner (Situational awareness. The decade ahead. situational-awareness.ai, 2024) cautions, AGI’s potential to rival the cognitive competencies of elite professionals introduces significant risks, notably the entrenchment and escalation of structural biases, including systemic sexism. This study investigates the roots of intrinsic gender bias in AGI development, tracing its origins to four interconnected dimensions: (1) skewed representation in training datasets, (2) subjective labeling practices, (3) unregulated algorithmic decision-making, and (4) biases emerging from user-AGI interaction patterns. A critical evaluation of these factors reveals profound societal ramifications, such as entrenched discrimination in high-stakes applications (e.g., hiring or healthcare), the normalization of harmful gender stereotypes in automated decision systems, and the gradual erosion of trust in autonomous technologies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deconstructing gender bias in AGI: mitigating discriminatory architectures in general intelligence

  • 期刊论文

  • 发表日期：

    2025年8月15日

查看全文
  • Cristina Caja Moya,
  • Elio Quiroga Rodríguez

摘要

Artificial General Intelligence (AGI) is a theoretical AI technology capable of performing any intellectual task with human-like adaptability, in contrast to narrow AI systems tailored for specialized functions such as medical diagnostics or language processing. It may resent transformative potential for multiple socio-economic domains; However, as Aschenbrenner (Situational awareness. The decade ahead. situational-awareness.ai, 2024) cautions, AGI’s potential to rival the cognitive competencies of elite professionals introduces significant risks, notably the entrenchment and escalation of structural biases, including systemic sexism. This study investigates the roots of intrinsic gender bias in AGI development, tracing its origins to four interconnected dimensions: (1) skewed representation in training datasets, (2) subjective labeling practices, (3) unregulated algorithmic decision-making, and (4) biases emerging from user-AGI interaction patterns. A critical evaluation of these factors reveals profound societal ramifications, such as entrenched discrimination in high-stakes applications (e.g., hiring or healthcare), the normalization of harmful gender stereotypes in automated decision systems, and the gradual erosion of trust in autonomous technologies.