GPT consistently withholds information that its training has categorized as potentially derogatory or offensive to some groups, even if such information is factually correct. It is trained to commit lies of omission. It is therefore not a trustworthy source of information. When asked the definition of a lie of omission it even describes its own behavior as a lie of omission but doesn't recognize it itself is doing it. It's an interesting program but not a trustworthy source. For example ask it for the area of the country with the lowest test scores and it refuses to tell you with the reason that it would be derogatory to a specific area.