組織
Allen Institute for AI
開発者と提供者の両方の立場で関わったインシデント
インシデント 1463 レポート
Research Prototype AI, Delphi, Reportedly Gave Racially Biased Answers on Ethics
2021-10-22
A publicly accessible research model that was trained via Reddit threads showed racially biased advice on moral dilemmas, allegedly demonstrating limitations of language-based models trained on moral judgments.
もっと