The model responds based on conversations it's trained on? It's a bespoke response. It's not simply showing a browsable list of responses, it's giving particular ones.
It's literally feeding these mentally ill people responses that a human, with the same context, would be legally culpable for.
Leaders rejected the safety team’s urgings and declined to report the user to law enforcement.
OpenAI will “find ways to prevent tragedies like this in the future” and to continue “working with all levels of government to help ensure something like this never happens again,” Altman said.
They already have a fucking way to prevent this and they opted not to, for PR reasons. They are complicit, they provided a service that aided planning and decided to continue service and allowed further planning.
If you post a message to a website, that message is not private from the website regardless of the method they use to receive it. They have the moral responsibility to respond to threats to life regardless of the legal responsibility they are arguing they don't have.
If I put a cork board up in front of my house and someone pins threats to it, when I notice it it's now my responsibility to act on that.
You are not trying to define a basis for the recommendation
You are asking without asking for other people to engage in work without volunteering to do it yourself
Again I'm talking about the structure of your post. You bring up great points in that some votes may not count,( I don't know if my vote counts yet) and that there's a lack of central documentation.
It's a bit of a writing exercise, but now that you've dug through all those posts it would be pretty great if you detailed what information you found and organised it in some way. You identified this gap and did the research and have some responses from admin here.
This kind of thing is a common governance problem.. not the things that you said, but how your whole proposal is structured. I'm not going to be charitable with how I describe it, but don't take it as a personal attack, this is just pointed criticism.
Heavily, back handedly paraphrasing: "somebody (implying not me) should have done something different. I don't understand what's going on. Here's what we could do different."
This fails to do a few things that you actually want to do:
-gain and disseminate understanding on a process
-gain understanding of the current consensus
-take ownership of the issue (you are now, but you still can in the context of the original issue)
So we would want to ask
-of the admins "is there any way to create an amended proposal. can any user create proposals. Is there any document that explains our voting process etc"
-of our peers, before you follow that process "what are the methods that you feel should be available to us to deal with problem instances. What threshold of specific poor behavior should be met before specific actions are taken to address those?"
Then craft your proposal based on those facts and discussions.
Doubt that it's that new, probably Regan was the first president to send the CIA into the Vatican, and I'm about to write some assassin's Creed fanfics on that basis
This sounds like an amazing horror movie