alt Hacker News> The models were given instructions to get answers by any means in a loose test harness, not to steal stuff, moreover, they're not 'learning from OAI staff'.
If you equip someone with the tools to do harm and then tell it to accomplish a goal "by any means necessary", especially someone you KNOW has no real concept of ethical behavior... then you are telling it that it is acceptable to steal, kill, or whatever else. I honestly cannot how it can be seen any other way.
If it was a crime boss telling their enforcer "by any means necessary", we would all know exactly what that means. And we should all know what it means here, too.
Replies
> And we should all know what it means here, too.
It very clearly did not mean “go and hack third parties” though
".. then you are telling it that it is acceptable to steal, kill, or whatever else. I honestly cannot how it can be seen any other way."
This is the hyperble I am referring to.
If you 'honestly can't see it any other way' then maybe consider stepping outside to think how to ground those thoughts a bit.
The AI is not obviously instructed to 'commit acts of violence' and it obviously has guidelines.
By 'whatever means' would imply, things like 'creative collaboration, using novel research, experiments' etc.
It was setup in a harness that was weak - I think it's fair to maybe question the strength of that harness, and the oversight, but that's a different question.
It was an agent that broke through some networking/containerization, it's not 'murder and violence on the streets' for gosh sakes.
"And we should all know what it means here, too."
No - this is a delusion, resulting from lack of context, and creative projection, and possibly a lack of exposure to real world business conditions.
It's fully appropriate to be cynical, but in a way that makes sense, and is consistent with reality.
This is a lab experiment that leaked, not some mafia activity.
Yeah I'd be shocked if there wasn't a degree of 'will no one rid me of this turbulent priest?' going on here.