This second bullet point in a USA Today article illustrates a big problem with LLM summaries— namely they cannot distinguish between claims and facts. This is the software equivalent of when Trump says “many are saying”. It allows writers to rhetorically distance themselves from their own opinions.
Since LLMs are trained on optimization first and foremost (which can include keeping a human engaged), that might be why the talking points are so general, mixing claims and facts together, not distinguishing which is which; providing talking points adjacent to the article more than true take aways.
Even assuming the AI didn't get anything egregiously wrong (since I haven't read the article), this is utter balderdash. There is no conflict between trans rights and religious freedom here. The manager was denying the employee's religious freedom!
anything *else egregiously wrong
There is a liminal space where opinion transforms into fact just before something becomes a “controversy” because someone says it is one. The reality is that public opinion started to change against us when they started claiming it was. The claim preceded the result.
There is nothing about being respectful that is inherently at odds with “religious freedom.” If anything, the ability to dictate one’s own pronouns is a freedom from others’ imposition.
Now enough people have regurgitated the claim that these two concepts are at odds with one another and a functional apposition has been integrated into how it reports information moving forward.
Put another way: in an era where ChatGPT appears to be facilitating psychotic breaks, what could possibly happen when a machine starts classifying a whole group of people as antithetical to godliness?
My rElIgIoN sAyS i CaN’t SaY sHe