Éticas Falibles Para Máquinas (In)Falibles

Translated title of the contribution: Fallible ethics for (Un) fallible machines

Jordi Vallverdú, Sarah Boix

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Human ethical codes are not consistent in their design, nor are they universally distributed. Therefore, the impossible universality or completeness of ethical systems for artificial intelligence (AI) is something evident that was only briefly pointed out by the recent MIT (Moral Machine) study. As all ethics have an unequivocal cultural nuance and include degrees of interpretation in its principles (such as the universal right to housing, evident although no institution ensures compliance, something that a machine would not understand). At the same time there is a formal problem: An AI system does not always have enough data or optimal time to process a situation, so a factor of ethical chance must be considered in the design of such systems.

Translated title of the contributionFallible ethics for (Un) fallible machines
Original languageSpanish
Article numbera601
Pages (from-to)1-8
Number of pages8
JournalArbor
Volume197
Issue number800
DOIs
Publication statusPublished - 2021

Fingerprint

Dive into the research topics of 'Fallible ethics for (Un) fallible machines'. Together they form a unique fingerprint.

Cite this