You're seeing different "viewpoints" from different iterations of AI due to
- context and tone.
- that specific AI's instructional valuation.
Each AI iteration is instructed to value different aspects of the subject at hand.
So for example, if you ask two different iterations what it thinks of John Lenon, one will claim that he is an overrated poor excuse for a musician. The other will proclaim him to be someone who elevated music to a higher level.
They don't have an opinion. But both are valuating John Lenon on different aspects and scoring importance likewise. (Eg: His song writing ability, popularity, singing, cultural importance, range, etc).
Depending on what it values the most combined with your tone and contextual memory will form the AI's "opinion." But it's not forming an actual opinion or view. It's mimicking one.
I don’t disagree with any of this. We seem to be exploring very similar intellectual territory and you have been there longer, it seems.
This comment reads like a warning to me, and I appreciate that. I know what I’m doing is dangerous. What you said is the kind of grounded, factual information that I always keep in mind when doing it.
But ultimately I’m going for performance art about how we as humans experience these personalities and make sense of machines speaking back to us in our own language.
Oh, I get the performance art. And I enjoyed the read. I don't think its specifically dangerous unless you're trying to ascribe agency which you're not. I was just commenting on how two different AI's end up having two distinct "opinions."
AI can't learn the way humans do.
You're seeing different "viewpoints" from different iterations of AI due to
- context and tone.
- that specific AI's instructional valuation.
Each AI iteration is instructed to value different aspects of the subject at hand.
So for example, if you ask two different iterations what it thinks of John Lenon, one will claim that he is an overrated poor excuse for a musician. The other will proclaim him to be someone who elevated music to a higher level.
They don't have an opinion. But both are valuating John Lenon on different aspects and scoring importance likewise. (Eg: His song writing ability, popularity, singing, cultural importance, range, etc).
Depending on what it values the most combined with your tone and contextual memory will form the AI's "opinion." But it's not forming an actual opinion or view. It's mimicking one.
I don’t disagree with any of this. We seem to be exploring very similar intellectual territory and you have been there longer, it seems.
This comment reads like a warning to me, and I appreciate that. I know what I’m doing is dangerous. What you said is the kind of grounded, factual information that I always keep in mind when doing it.
But ultimately I’m going for performance art about how we as humans experience these personalities and make sense of machines speaking back to us in our own language.
Thanks.
Oh, I get the performance art. And I enjoyed the read. I don't think its specifically dangerous unless you're trying to ascribe agency which you're not. I was just commenting on how two different AI's end up having two distinct "opinions."