Elio Quiroga (@elioquiroga.bsky.social)

Papers move the world! "Deconstructing gender bias in AGI: mitigating discriminatory architectures in general intelligence" en AI & Ethics, de Springer Nature. Sesgo de género en AGI: causas, impactos y mitigación con soluciones técnicas y reformas. link.springer.com/article/10.1...

loading . . .

Deconstructing gender bias in AGI: mitigating discriminatory architectures in general intelligence - AI and Ethics 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. https://link.springer.com/article/10.1007/s43681-025-00818-1