玖伊珀薇妮: 我如何對抗偏差的演算邏輯

我如何對抗偏差的演算邏輯

麻省理工學院（MIT）研究生玖伊珀薇妮在使用臉部辨識軟體時發現了一個問題：這個軟體無法辨識她的臉，因為當寫程式的人把偏差的人工智慧邏輯植入這個軟體，導致它對於某些膚色與臉部結構沒有反應。現在她擔負重任，務必要導正這個機器學習的偏差。她稱這個誤差為「數碼凝視」。這場別開生面的演說主要是關於程式的可信度。因為演算邏輯正在鯨吞蠶食我們的生活。

Algorithmic bias can exclude and discriminate at lightning speed, affecting millions without warning. When facial recognition software failed to detect a face unless a white mask was worn, the flaw exposed a deeper issue: undiverse training data shapes who technology serves. With police departments accessing facial recognition networks containing over 117 million Americans, unchecked algorithms risk reinforcing unfair practices. Addressing who codes, how they code, and why they code is essential to building technology that works for everyone.

Algorithmic Bias Spreads Faster Than Human Prejudice

*Unmasking AI: My Mission to Protect What Is Human in a World of Machines*

"Meet the computer scientist and activist who got Big Tech to stand down"

activism

algorithm

code

data

identity

product design

innovation

software

TEDx

technology

race

equality

TEDx Talk