owencb

Posts

Sorted by New

Wiki Contributions

Comments

Truthful AI: Developing and governing AI that does not lie

Thanks, I think that these are good points and worth mentioning. I particularly like the boundary you're trying to identify between where these decentralized mechanisms have a good track record and where they don't. On that note I think that although academia does have complaints about political bias, at least some disciplines seem to be doing a fairly good job of truth-tracking on complex topics. I'll probably think more about this angle.

(I still literally agree with the quoted content, and think that decentralized systems have something going for them which is worth further exploration, but the implicature may be too strong -- in particular the two instances of "might" are doing a lot of work.)

Welcome to Lesswrong 2.0

I think keeping some dependence on quantity is desirable, but that scaling linearly with number of posts weights it too heavily compared to variation in number of upvotes (I proposed scaling with roughly the cube root of number of posts in my explicit formula suggestion elsewhere in the comment thread).

Closed Beta Users: What would make you interested in using LessWrong 2.0?

A good system for finding old content. Often old posts or comment threads represent the best content on a given topic, but if there isn't a way to browse by topic they're basically lost. This makes me less excited about engaging in the first place -- vice versa a sense that something will persist is a draw in a venue for circulating ideas.

This is one thing where I think the forums/subforums structure of old UBB boards is actually superior to most more modern replacements (although such boards have other significant problems, most notably lack of threading).

Welcome to Lesswrong 2.0

Here's an example functional form which is the best guess from the top of my head at creating this effect (but I'm giving as an illustration of what to pay attention to rather than a claim that this precisely should be used):

K = (U - 3D) * P^0.3 / R

Where
K = Karma
U = total (weighted) upvotes
D = total (weighted) downvotes
P = total number of posts+comments
R = total number of reads of your posts+comments

Welcome to Lesswrong 2.0

One variation of karma system I'd like is the ability to rate posts as being exceptionally good (probably taking more than one click, to introduce a trivial inconvenience so that it isn't used all the time like five star ratings are). This would give more ability to pick out very useful contributors from small numbers of posts.

Welcome to Lesswrong 2.0

You talk about using karma thresholds for various things. But traditional lesswrong style karma screens more for quantity than quality of posts, and this would remain true of a version where you weight people's upvotes and downvotes. I suggest looking for versions which filter more for quality (while not creating too much disincentive to make additional posts/comments).

LessWrong 2.0 Feature Roadmap & Feature Suggestions

I'm in favour of this, but I think it would be even better to give high weights by fiat to some trusted users, in a way that grounds the whole Karma system in something, and makes it harder for low-quality bubbles to self-perpetuate.

LessWrong 2.0 Feature Roadmap & Feature Suggestions

I think keeping track of which posts you've read would be useful for all posts, not just sequences.