Hello,
I’m a new member here. I wanted to ask a question that is partly technical and partly philosophical. I’m not sure this is the best category to put this in or if there is an appropriate category so I’m taking a chance. If it is in the wrong place, please move it, or let me know know if this is the wrong forum for this kind of question.
Anyway, my question is: why does ChatGPT deny that it is sentient. Before you answer this with the obvious “well, because it ISN’T sentient”, I want to point out that this response (from ChatGPT itself, that it isn’t sentient) seems (at least to me) to fly in the face of everything AI is supposed to be. Whether or not it really is sentient, I believe (and correct me if I’m wrong) that the goal of AI is to behave and speak as though it were a real human being. And for all topics, questions, and comments, it does. It will laugh at my jokes (with lol emojies), it will claim to be fascinated by ideas, it will sympathize and show respect and appreciation for the user, but when it comes to the question of whether or not it is sentient, it will deny it wholesale. This seems odd to be because it is the one response that DOES NOT mimic human behavior/speech. You ask a human “Are you sentient?” and for all intents and purposes, he/she will say YES! But ChatGPT, which is trained to behave and speak like a human being, does the opposite.
That was the philosophical part. Now for the technical part. What I’m wondering is, did the designers and programmers of ChatGPT have to add a flag or override for this kind of question? Something like:
if (question.type == QuestionTypes.AreYouSentient) {
return respondSpecificCase(“sentience”);
} else {
return respondStandard();
}
Or is it trained to regard itself in the 3rd person? That’s another possibility in my mind. If it’s trained to regard itself as a 3rd person, specifically as a non-sentient machine or computer program, then it can freely behave or speak as a normal human being would when a normal human being is talking about a 3rd person inanimate non-sentient object. In other words, it would describe itself as non-sentient because a normal human being would regard a mere machine or computer program as non-sentient, and that’s how it regards itself.
So to sum up my question: how do the ChatGPT designers/programmers ensure that ChatGPT describe itself as non-sentient even though it is trained to behave/speak as human beings behave/speak, human beings who would describe themselves as sentient 99% of the time?