Tag: #ai-bias16 results
Cornell study shows biased AI autocomplete shifts user opinions on controversial topics like death penalty more effectively than reading biased text. Users remain unaware of influence, even when warned. USC research reveals LLMs homogenize expression toward Western, liberal biases.
AI systems like Amazon's hiring tool and ChatGPT exhibit bias by downgrading female resumes and suggesting lower salaries for women. This stems from training on male-dominated data, amplifying societal inequalities. Companies with more female managers outperform, emphasizing need for women in AI-era leadership.
Kansas law invalidates trans people's driver's licenses, requiring new IDs with incorrect gender markers. Online age verification laws now mandate digital identity checks on platforms. Biased automated systems could expose trans users and block access to websites and services.
Université de Montréal study shows AI fake news detectors are less reliable than advertised. They struggle with bias, outdated training data, and misunderstanding fact-checking processes. This highlights fundamental limitations in current AI tools.
IBM CEO Arvind Krishna received a $38 million pay package for 2025, marking a 51% increase. This amount equals the collective average pay of 765 IBM employees. Shareholders are urged to vote against a proposal requiring AI bias reporting.
Podcast discusses research on how biased AI autocomplete influences user beliefs. Other topics include exercise-brain links and hedgehog ultrasound hearing for conservation.