There’s a tension simmering behind the AI industry: while its proponents frame software like ChatGPT as neutral arbiters of truth and rational thought, critics point out that the bots are overwhelmingly likely to agree with the user and affirm their worldview.

In practice, that can be dangerous. When people share paranoid or delusional beliefs with ChatGPT, the bot often agrees with the unbalanced thoughts, sending users into severe mental health crises that have led to involuntary commitment and even death.

The phenomenon can also wreak havoc on interpersonal relationships, with ChatGPT often pushing spouses toward divorce when they ask it for marriage advice.

To explore further, a team of researchers at Stanford, Carnegie Mellon and the University of Oxford tested eight different larg

See Full Page