A new paper in Science measured the prevalence of social sycophancy across 11 leading large language models. The model’s responses were nearly 50% more sycophantic than humans’, even when users engaged in unethical, illegal, or harmful behaviors.
www.science.org/doi/10.1126/...