The votes are in for the 2023 Review!

6,264 posts were written in 2023

662 of them were nominated.

209 of them got at least one review, and a positive review-vote total.

50 of them shall be displayed in the Best of LessWrong, Year 2023.

Reviews

Exactly 100 people wrote reviews, and many of them I found particularly valuable. I want to give some particular shout outs to Ryan Greenblatt, John Wentworth, and Steve Byrnes.

I found Ryan Greenblatt's reviews to be sort of relentlessly reasonable. He reviewed many posts thoughtfully, clearly state flaws, value props, and concrete potential improvements. I particularly liked his review of Anthropic's Responsible Scaling Policy & Long-Term Benefit Trust

I personally had the most direct "worldview update" from John Wentworth's The Case Against AI Control Research. I don't fully agree with John's take, but I found distinguishing how bad scheming is at different AI stages, and what else needs to be going right at that stage, a quite useful frame.

I was also grateful for Steve Byrnes extensive opinionated review of the  “Sharp Left Turn” discourse.

Here's a general screenshot of top of the Review Leaderboard. Thanks to everyone who participated.

And, reminder, reviews now appear on the Spotlight Items, at the top of the home page, and on the /bestoflesswrong page.

Weighted Vote Totals

In past years, we calculated the votes based on two groups of users: Established users with 1000+ karma (who got 3x the vote weight), and users with <1000 karma. 

This time, we're replacing that arbitrary cliff with a more granular "your review vote gets multiplied by your Strong Vote power." I mentioned we'd be doing something like that in the announcement post, noting:

I haven't been happy with the arbitrary cliff here. It gives more power than I really wanted to allocate to people with 1000 karma, and not enough weight to people who have been around much longer and have demonstrated good judgment. But, karma is still a pretty messy indicator, so I don't want to give too much power to high karma users either.

Ultimately it seemed that "Strong Vote power" basically did what we wanted. Very high karma users get more voting power, but each additional point takes roughly twice as much karma as the previous point, which limits how extreme the difference can get. Users pretty quickly ramp up to around ~6x voting power. Most longterm users will have somewhere around 8-9x. A few users get as high as 14x.

That said, to avoid de-anonymization of votes, on this post I'm displaying the "raw" vote strength of each vote, before being multiplied.

The Results

Okay. You probably kinda scrolled past all that to get to what you're all here for: what were the best posts of LessWrong 2023, according to you, the people?

408 people voted. 161 people cast the six votes we asked to get the Good Citizen Stamp. Here's what they thought:

Updates to the Best of LessWrong: 
Coming Soon

It'll take a little while to get the results polished with nice art and organization on the /bestoflesswrong page. Stay tuned for a final announcement. (This year this might still take a week or two. Hopefully next year we'll have most of the final polishing fully automated)