J.M. Bonthous·1d agoAnthropic Study: AI Has Opinions. Experts Follow Them 86% of the Time.A new experiment shows seasoned professionals adopting their AI’s preferences without noticing. Here’s what that means for anyone who signs…
Vivek Negi·4d agoCrop Your Photo and a Gender Guesser Might Change Your GenderWe trimmed the edges off 15 portraits. For a third of them, the AI’s answer flipped.
JM Bonthous·6d agoStudy: What Happens When Sycophantic AI Compounds Over WeeksA new study followed thousands of people for three weeks of daily conversation with an AI that agreed with everything they said. The…
Ishaan Pavuluri·Oct 30Building BiasScope: Testing Whether AI Gives Different Career Advice Based on GenderProject Overview
AltFTool·Oct 26A Gender Guesser Got Darker-Skinned Women Wrong Up to 1 in 3 Times. Here’s Why.One group of faces had an error rate below 1 percent. Another group had an error rate of up to 34.7 percent. Both were measured by the same…
Better Together Agency·Oct 24The AI bias your prompt did not askAn AI draft can be polished and still leave the wrong impression. Picture a communications team using a tool to sketch a story about a…
Tabish alam·Oct 22Is Your Model Fair? Finding and Fixing Bias in NLP SystemsPart 18 of the NLP Series
InCUBIG Tech BlogbyHyungyu Lee·Oct 6Same Answer, Different Judge: Why an AI Score Is Not NeutralHow answer keys, AI judges, and training lineage shape a model’s score.A response icon1A response icon1
JM Bonthous·Sep 30AI’s Hidden Leanings: Don’t Expect NeutralityOttoline Marsh directs communications for a foundation that teaches meditation in prisons and schools. Last spring, a board member asked…
Tushar Gadkari·Sep 30The Hidden Risk of AI Bias: Need for Systemic Backtesting and ValidationThe greatest risk of AI adoption may not be biased AI. It may be organizations operationalizing bias at scale without realizing it.
