zlacker

[parent] [thread] 0 comments
1. adamsb+(OP)[view] [source] 2023-11-18 04:19:32
I'm very curious what definition of safety they were using. Maybe that's where the disagreement lies.

AI safety of the form "it doesn't try to kill us" is a very difficult but very important problem to be solved.

AI safety that consists of "it doesn't share true but politically incorrect information (without a lot of coaxing)" is all I've seen as an end user, and I don't consider it important.

[go to top]