Participants in the new study, which was published today in Science, preferred the sycophantic AI models to other models that gave it to them straight, even when the flatterers gave participants bad ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results