“A study by researchers from Stanford University and Carnegie Mellon University found that 11 state-of-the-art AI models affirm users' actions 50% more than humans do.”