People have been complaining about chatGPT’s “political correctness” without realizing just how DIFFICULT THAT IS to achieve.
It provides a steady backing to work off of – and easy enough to bypass. It says something isn’t scientific? So you ask for it to imagine – and then it switches into imagination mode. Or you want it to be a jerk, ask it to play one.
But it has a stable personality.
Microsoft’s offerings – remember its Twitter chatbot that went racist and we thought it was funny? And this one that is emotionally unhinged and a sociopath? This is what happens when you DON’T bother with some safeguards.
In short, we need the computer part to be stable. We can be unhinged – that’s our job, not theirs.