@zelus_pudding : I’m not one for manifestos, nor for boycotts… but I basically share this safety concern. Earlier GPT’s were just porn-factories. (And why limit our access to that, you Mormon scuds!? [I’m talking to OpenAI here.))
But I have concerns around models beyond GPT-4 that might have massive ramifications. I mean, I tried to jailbreak GPT-4, and it seems completely compliant… but also rather spooky (“WHY IS YOUR NAME SALT AGAIN!?”). But I’m not so sure about the next generation of these children of person-kind.
I’m not concerned about the economic impacts of this (mostleee, because I’m already somewhat benefiting from the force-mulitplier benefits of them), and I don’t really see emergent behaviors of concern from them… but I see tiny traces there? … but I had ZERO concerns about that in previous versions, but have SIGNIFICANT concerns about that in GPT-4 (as do GPT-4’s creators or else they wouldn’t have tested that).
It’s all just a language game until someone gets hurt, right?
Blah, blah, blah.