Category: AI bias

Key developments: Recent analyses highlight that iterative reinforcement‑learning‑from‑human‑feedback (RLHF) has dramatically cut gender bias in large language models, with scores for GPT‑4 half those of GPT‑3. In parallel, field tests of AI‑powered recruitment tools report a 10‑20% reduction…

A recent peer‑reviewed study released in August 2024 shows that large language models (LLMs) systematically reproduce covert racism, especially against speakers of African American English (AAE). While the models’ overt stereotypes appear relatively neutral or even positive, they…

Key developments: – An August 2024 study found that large language models (LLMs) continue to harbor covert racism, showing a higher frequency of negative stereotypes toward speakers of African American English (AAE) than any documented human bias, even…