Даниил Иринин (Редактор отдела «Наука и техника»)
We propose sycophancy leads to less discovery and overconfidence through a simple mechanism: When AI systems generate responses that tend toward agreement, they sample examples that coincide with users’ stated hypotheses rather than from the true distribution of possibilities. If users treat this biased sample as new evidence, each subsequent example increases confidence, even though the examples provide no new information about reality. Critically, this account requires no confirmation bias or motivated reasoning on the user’s part. A rational Bayesian reasoner will be misled if they assume the AI is sampling from the true distribution when it is not. This insight distinguishes our mechanism from the existing literature on humans’ tendency to seek confirming evidence; sycophantic AI can distort belief through its sampling strategy, independent of users’ bias. We formalize this mechanism and test it experimentally using a rule discovery task.
，更多细节参见safew官方版本下载
Stack allocation of append-allocated escaping slices
Stanislav Vishnevskiy, Discord's co-founder and chief technology officer, said a planned global rollout of a verification process to determine users under the age of 16 would be delayed until the latter half of this year.
12:56, 3 марта 2026Мир