Sequences

The central limit theorems

Wiki Contributions

Comments

I read this book in 2020, and the way this post serves as a refresher and different look at it is great.

I think there might be some mistakes in the log-odds section?

The orcs example starts:

We now want to consider the hypothesis that we were attacked by orcs, the prior odds are 10:1

Then there is a 1/3 wall-destruction rate, so orcs should be more likely in the posterior, but the post says:

There were 20 destroyed walls and 37 intact walls… corresponding to 1:20 odds that the orcs did it.

We started at 10:1 (likely that it’s orcs?), then saw evidence suggesting orcs, and ended up with a posterior quite against orcs. Which doesn’t seem right. I was thinking maybe “10:1” for the prior should be “1:10”, but even then, going from 1:10 in the prior to 1:20 in the posterior, when orcs are evidenced, doesn’t work either.

All that said, I just woke up, so it’s possible I’m all wrong!

In Korea every convenience store sells “hangover preventative”, “hangover cure drink”, with pop idols on the label. Then you come back to America and the instant you say “hangover preventative”, people look at you crazy, like no such thing could possibly exist or help. I wonder how we got this way!

Thanks for your review! I've updated the post to make the medications warning be in italicized bold, in the third paragraph of the post, and included the nutrient warning more explicitly as well.

“(although itiots might still fall for the "I'm an idiot like you" persona such as Donald Trump, Tucker Carlson, and particularly Alex Jones).”

This line is too current-culture-war for LessWrong. I began to argue with it in this comment, before deleting what I wrote, and limiting myself to this.

It changed to be much more swipe-focused. It’s been 5 years since I used it, but even in 2018, I remember being surprised at how much it had changed. Apparently now even open messaging is gone, and you need to have someone Like you before you can message them, though I haven’t actually checked this.

Yes, agree - I've looked into non-identical distributions in previous posts, and found that identicality isn't important, but I haven't looked at non-independence at all. I agree dependent chains, like the books example, is an open question!

Love this! Definitely belongs on LessWrong. High-quality sci-fi, that relates to social dynamics? Very relevant! I’ve been away from the site for a while, tiring of the content, but am glad I scrolled and saw this today.

Enjoyed this! Very well written. The two arrow graphs, where the second has everything squished down to the bottom, are especially charming

I don’t think the problem is this big if you’re trying to control one specific model. Given an RLHF’d model, equipped with a specific system prompt (e.g. helpless harmless assistant), you have either one or a small number of luigis, and therefore around the same amount of waluigis - right?

Load More