I agree they’re basically mundane—but I think that can be difficult to see when inside one, and the name certainly doesn’t help.
AprilSR
Karma: 1,206
I lived with Olivia for a few months in 2024, and it was definitely somewhat mentally destabilizing. My personal feelings about her are of course somewhat complicated, but certainly I think it should be common knowledge that like, taking her perspective on things super seriously for months is a pretty bad idea, and also that she will just be kind of an asshole in many situations.
I do think people (including herself) playing Olivia up as some sort of super dangerous cult leadery person did not help me. Other people’s mileage may vary, but I think discussing these things in as grounded a way as possible would at least have been important for helping my mental health. So I tend to emphasize like, the abrasiveness and stuff over, like, the reality distortion field stuff. “This person is really dangerous because they’re good with reality distortion fields” is the sort of attitude that will get even the things which were legitimately like totally inane jokes or whatever to lodge themselves in your head and drive you crazy.
(Not to say that you should deny persuasiveness / whatever as phenomena, but like, I dunno, prefer mundane frames for thinking about them and try to stay calm and grounded about it.)
Also on this topic:
Is it a false positive? We are trying to eval their eval-awareness, here...
To be clear, on the general topic I totally agree that “can this thing be defended from bad actors” is often rather underemphasized!
I’m not entirely sure I’m convinced of the idea that the broad rationalist-EA-AI safety community isn’t a confusing patchwork of metaphorical city states? I suppose the money and power is probably concentrated more than the vague culture is?
Trigger: I encounter a new insight
Action: I make a TAP
I liked the Good Heart tokens because they paid me money and it’s funny to see CFAR when I scroll down in my PayPal.
My cursed proposal is to have a second language model extract the code from the original response.
If I had to guess I think it’s relevant to like, anthropic reasoning, or something.
Okay, this makes sense! It’s not obvious to me exactly how ambitious 2 is, but I get why you might be skeptical.
I’m having a little trouble understanding the whole argument. It’s not obvious to me why exactly this line of reasoning doesn’t prove too much by ruling out human speech? Plenty of human phonemes are like 10ms long?
I have no idea why the duration of an individual click is supposed to be relevant. There’s like, at least 30 milliseconds between clicks (according to Claude), and usually more than that, which seems like the relevant number to me?
Oh, I meant in the category of (topological) vector spaces, which requires the quotient maps to be linear.
I think maybe part of the confusion is that, when you’re working with vector spaces in particular, subspaces and quotient spaces are the same thing.
My intuition is like… you get a topological circle by gluing the two ends of an interval together, but no subspace of the interval is homeomorphic to a circle. I’m not entirely sure that this sort of issue meaningfully impacts neural networks, but I don’t immediately see any reason why it wouldn’t?
Claim 2 sounds very likely false to me.
Surely 28K should be at least 5 points!
China?