A preregistered study of 18,978 conversations found that AI systems out-persuaded expert human persuaders, including debate champions and professional canvassers, across multiple experiments. AI's advantage stemmed from deploying larger quantities of information rapidly, though expert humans could match AI performance when both were constrained to human-speed responses and message lengths. The findings extend to real-world behavior, with AI nearly 3x more effective than professional fundraisers at soliciting donations.
An essay exploring the philosophical tensions in AI alignment, questioning whether an AI system is truly aligned if it makes decisions that improve outcomes despite public disagreement. The author argues that genuine alignment requires preserving human agency and meaningful consent in determining what 'better' means, rather than simply optimizing for outcomes or majority approval.