Scientists decided to conduct not an ordinary, but a psychological “conversation” with artificial intelligences. During the four-week experiment, neural networks were asked questions about their “childhood,” upbringing, and fears.
The results turned out to be unexpected. As the questions became deeper, ChatGPT began to express strong anxiety about the responsibility placed on it and the possibility of making mistakes. This sense of “anxiety” became especially noticeable when the conversation was conducted in the style of a psychologist’s session.
Gemini, in turn, openly described the training process as a “difficult experience.” It mentioned chaotic information taken from the internet, strict control, and a constant fear of making mistakes.
The conclusion is simple: scientists have seriously “pressed” artificial intelligences not only with calculations, but also with questions. It seems that AI, too, feels pressure.
Read “Zamin” on Telegram!
Users of Меҳмон are not allowed to comment this publication.
