So do you find it alarming that people are trying to give such a system “agency” ?
It is extremely interesting to me how much milage can be gotten out of an LLM by observing patterns in text and generating similar patterns.
It is extremely frustrating to me to see how easily people think that this is evidence of intelligence or knowledge or "agency" as you suggest.
The most important problem is to ensure people understand that these systems cannot be trusted to tell them the truth.
I'm OK with starting out with "these systems will lie to you", then following up with "but you do need to understand that they're not anthropomorphic etc" later on.
(If what you really meant is humanlike intelligence, then I would agree with that claim, but it's a very different one.)
Barring OpenAI's filters, if I asked it to participate as a party in a business negotiation determining a fair selling price for its writing services, I'm sure it could emulate that sort of character long enough to pass. And if it can successfully fake being a self interested actor working towards goals of its own material interest, whats the difference between faking it and actually having agency? At some point you have to acknowledge emergent agency and other properties as a possibility.
I'm not sure how confident I am!
I think the success of LLMs are serving to challenge our understandings of cognition and intelligence in humans. I tend to agree with you that our understanding of agency, intent, intelligence, cognition, free will, and related ideas in humans is incomplete and so it is a challenge to think about how they might apply (or not) to LLMs.
> whats the difference between faking it and actually having agency?
Good question. The Chinese Room Argument is very close to this but framed as "understanding" vs "agency": https://plato.stanford.edu/entries/chinese-room/