Bring Back the Sequences?

by Alexandros1 min read7th Mar 201137 comments

62

Rationality A-Z (discussion & meta)Site Meta
Frontpage

Given that 

1. Deciding to read and actually reading the sequences is 'work'

2. Reading the latest frontpaged article on LessWrong is 'fun'

3. We frequently have gaps in the posting rate of articles that make it to the front page

4. There are many people who joined this community after the sequences were written and haven't gone through all of them

...would it make sense to start bringing articles from the sequences to the front page, either at a set pace or whenever there is a gap in posting?

I have actually read most of the sequences, but wouldn't mind going through them once again. However, taking it up as a project seems like too much work. By bringing an article to the front page, either with the old comment thread or with a fresh one (plus a reference to the old one), it becomes something that the community is doing. Following things that a group you belong to is doing is fun. But for that to happen, we need to share a common pointer to which article is 'the one we are reading now'. Hence, the front page.

In short, I think if people in this community reading (and re-reading) more of the sequences is something we want, then recycling them through the front page is also a good idea.

If the barrier is implementation modifications needed, I may be able to assist.

37 comments, sorted by Highlighting new comments since Today at 4:46 AM
New Comment

I would really like to see people take a shot at compressing some of the shorter sequences into single posts with no dependencies, and posting them to the front page.

Either a reposting programme or a summarising programme would be an improvement over the lack of either. But I wouldn't block one on the other.

I don't know if recyling the sequences to the front page is the solution, but you do have some valid points.

It would be nice if some kind of sequence "book-club" functionality existed within the LW platform that enabled people to form reading groups, depending on how far along they were with their readings, and engage in fresh active discussion.

On the other hand, the sequences will likely be distilled into book format in the near future, according to the SIAI website, so there might not be much incentive to do anything about them at this point.

By bringing an article to the front page, either with the old comment thread or with a fresh one (plus a reference to the old one), it becomes something that the community is doing.

The main problem with your proposal is that it unnecessarily forces everyone who reads the front page to participate in your project. (Eliezer wrote over 600 posts on Overcoming Bias. If they were all reposted, that would drastically change the character of the front page.)

If on the other hand, the reposts were to appear at a URL that is not currently being used, /r/sequences/ say, then each person could decide whether to participate in your project or to continue to use LW the way they are using it now.

Parenthetically, if I were to write software to support your project, I would create an RSS feed of (new) comments on the (old) posts in the sequences. That way, someone who is reading a post in the sequences knows that if they make a comment (e.g., to ask a question), it will probably be noticed. In fact, the probability that it will be noticed is high even now, but would rise if there were a way for readers to watch for the comment without needing to scan the firehose that is http://lesswrong.com/comments/.

Note that such an RSS feed could be created without modification of the software behind lesswrong.com (since the necessary information is already in http://lesswrong.com/comments/.rss and the new RSS feed can be published at a domain other than lesswrong.com) which is probably a good thing given the difficulty some here have had recently in getting the attention of a someone with the authority to pull changes into the software behind lesswrong.com.

As someone who came across this website recently, I very quickly got into the Sequences because of how articles I homed in on referred back to others etc. etc. often leading to the Sequences which then have a clear order in which it makes sense to read them. My brother did the same thing (in fact, ignored my suggestions on individual articles he'd like and very quickly started working through the sequences)

Of course, generalising from two is dangerous, especially as we're family and think in similar ways. But I do think the way articles cross-refer is very useful. Perhaps you could achieve some of the same objectives simply by keeping clear links to Sequence posts wherever relevant, rather than assuming the knowledge. Though, as I said, this already happens a lot!

I'd also add that the 'reading group' idea raised by is great: I'd love to be able to discuss the Sequences as I go through them with a larger group who are at a similar place. I think if you just read stuff it's very easy to think you understand it and promptly forget it. These groups could be just between lots of new readers, or could also have a link to someone who's familiar with them and can clarify any confusions.

I actually find the way articles cross reference infuriating. It's fun, but it quickly leaves me in a complicated mess of tabs, each of which is only tangentially related to the hyperlink that led me there. I didn't even realize there WAS a sequence for a while, and by the time I did I couldn't remember what I had read already and what I hadn't.

I decided to start reading the Sequences today. I will read one post per day because I'm also learning math and programming at the same time. I also intent to use this link (all articles from Less Wrong, in chronological order) rather than this one.


Actually I plan to alternate between the following activities in my spare-time. Either each for half an hour per day or a certain amount (x videos, LW posts etc.):

Mathematics

Programming

Haskell

One miscellaneous book

Starting with Darwin's Dangerous Idea, Daniel Dennett.

LessWrong Sequences

One post/article per day.

Thinking

Taking some time to meditate about the big picture, what I want, what I should do or some miscellaneous problem.

Sports

3 times per week running and weightlifting.

Relaxing

Alternating between playing a video game or reading science fiction for 30 minutes each evening.


What do you think? Is it too much at the same time, should I concentrate more on a certain activity or maybe dismiss one altogether, am I missing something important? My fear is that some of the activities might be more important for various reasons. For example, should I solely concentrate on math because I'll be unable to learn much when I get older (I'm already 27)? Any suggestions are appreciated!

What do you think? Is it too much at the same time

Yes, it is too much to start doing at the same time. It is a great looking program to build up to. But building habits is really, really hard. A common rule of thumb is to limit yourself to one change per month. (More than that tends to require a significantly stressful external stimulus.)

For example, should I solely concentrate on math because I'll be unable to learn much when I get older (I'm already 27)?

No (particular) need to rush the math. You might be getting to old to be making grand new exciting mathematical discoveries but for learning the basics there is no dire limit. Particularly if you execute the exercising habit consistently and are keeping your mind active learning other things (which is obviously the plan!) Heck, you're learning Haskell. You are not going to be losing basic-math-learning potential while studying Haskell.

Any suggestions are appreciated!

Find something to be doing with the skills in the intermediate term. Make sure they are giving you real world positive feedback. They can't just be things you think you 'should' do to meet some vague abstract goal that isn't immediately salient.

They can't just be things you think you 'should' do to meet some vague abstract goal that isn't immediately salient.

I really don't want to do those things because I think I 'should' do them but because I'd love to do them. If I would be living in Bank's Culture or under Yudkowsky's FAI oversight I would want to do exactly those things. That I ask about them at all is because I fear I should be doing other things. If I would just accept the line of reasoning that there is nothing more important than donating to the SIAI then I would choose to become a street builder and work long hours. You can earn good money doing that here in Germany while you don't need to have particular skills. I don't expect the time I would need to acquire advanced knowledge to pay off either by enabling me to get a better job or by being able to work towards FAI directly.

I suppose all my questions and submissions here tend towards the goal of allowing me to conclude that I can ignore all this. My nightmare is that reality is fucked up enough that trying to do the 'right' thing makes you eventually end up seriously considering three spins of a roulette wheel or tossing a quantum coin to gamble at a 10,000:1 ratio.

Does my desire to learn not eventually result in knowing what is right and wrong? Yeah, but making decisions under uncertainty would currently force me to take the risk of not pursuing any terminal goals directly but rather to try to mitigate risks from AI. That sucks and I try to ignore it but haven't been able to do so yet.

I'm just looking for some justification to do what I want and ignore what I don't want.

"Before enlightenment, mountains are mountains and rivers are rivers. At the moment one is enlightened, mountains are no longer mountains and rivers are no longer rivers. After enlightenment, mountains are again mountains and rivers are again rivers."

When you learn about relativity, do you throw away your watch because time is relative anyway? No, because relativity must give the same results as Newtonian physics in the conditions where (and to the precision that) Newtonian mechanics has already been verified. If you have a theory that says strange things about what happens in strange conditions, it may be true. If you have a theory that says strange things about what happens in normal conditions, you know the theory must be false, however elegant the mathematics.

The take-home message is that it all has to reduce to normality. No, quantum mechanics really doesn't have implications for everyday life. No, thought experiments involving copying of people really aren't useful guides for what to expect in our world where we can't copy people. No, AI really isn't going to conquer the world. No, donating your income to charity really isn't the best way to make a positive difference. No, dust specks really aren't worse than torture, thought experiments about group sizes of 3^^^3 notwithstanding. No, treading on ants really isn't a sin.

None of this is to say that you shouldn't learn about intellectual topics. By all means do, if that's what you want! But you seem to be suggesting you feel obliged to deal with some of these topics even if it's not necessarily what you want. And I'm saying you don't need to feel that way. You don't actually need to learn tensor calculus to know that relativity doesn't keep your watch from working normally. If what you want to do for a career is design GPS satellites, or if you just plain think tensor calculus is interesting, by all means go ahead and learn it. But if you want to ignore it and go do something else instead, then do that.

To rephrase my suggestion for clarity: Make sure your instinctive reward system is getting positive feedback to associate with your learning habits in a form that it can understand. Such as a social component, feedback by peers or instructors, short term goals, experiences of 'flow', etc.

Note that I am already systematically blogging through every single Eliezer post. Here is the latest.

What do you think about cross-posting them here on the main page?

I don't think that is wanted...

And for anyone starting from the beginning, here is the index. I notice it's fallen behind - the posts are up to #19 while the index has links up to 11.

Following links from Luke's posts is probably a better way to read Eliezer's posts than the Sequences page on the LW wiki is, because of the summaries, and because many of the best articles are one-offs that didn't fit into the "sequences" classification. (Though it does have the disadvantage of not including other authors, some of whom do feature on the Sequences page. Luke, perhaps when Reading Yudkowsky starts nearing the end, you'll continue with Reading Yvain Et Al?)

Yeah, I only update the index every now and then.

The series won't include the other posters (I know this because I've actually finished writing the sequence (62 posts total, if I recall), but I haven't published them all yet).

You wrote them that far in advance of your publishing schedule? Whatever anti-akrasia drug you're on, I'd like some...

Those all sound like valuable activities. The programming would probably be better to do in a smaller number of longer sessions - I find that programming gets more informative and productive the more consecutive hours I spend on it, because of all the context I need to remind myself of.

I've been watching some Khan Academy videos myself, and one thing I've found useful is downloading them and watching them with VLC player, which can adjust the speed to my preferred pacing (I speed it up by 1.5-2x, and use the time saved to watch the tricky parts a second time if necessary). Also, having a folder full of video files that I move into the 'finished' folder is somehow more motivating than having a page full of links that I turn purple.

I'm in favor of doing this or something similar.

There have been other related suggestions for how to make the sequences more accessible, one of which I described here. Instead of re-promoting the old post to the top of the front page, someone could just make a new post (probably in the discussion section) with a link to the old sequence post. The new post could have a standardized format for the title, to make it clear that it's just linking back to the sequences, and perhaps a one paragraph summary of the post that it's linking to.

An advantage of this method is that it's ready to implement - there's no need to program anything new into the site or wait on the folks in charge. We just need to decide on the conventions and then we can start the posts.

I have two specific ideas for what the conventions could be, which I'll leave as comments so that they can be commented on & voted on separately. We could do either one of them, or both.

Idea two: Cycle through the sequences. Start with the first post in the first sequence and have one post per day in the discussion section. Each new post would contain a link to the original post and a one paragraph summary of the post (which could be copied directly from the wiki in many cases). Any additional comments or discussion would take place in the comments to the new post.

I think this is a great idea. I've been reading through the sequences and, while I would like to comment, it seems silly to participate in a conversation that occurred two years ago. So the point would be just to have a "fresh" conversation about an old post. I think this is a good idea in general; for some reason the internet likes to pretend that anything written about more than a week ago doesn't exist or isn't relevant.

For what it's worth, when I discovered the site last year I started working through the Overcoming Bias archives linearly, and I went ahead and commented on those posts when I felt moved to.

Occasionally people would respond; more often they wouldn't. Which is true for my comments on current posts, as well. And others'. So I'm not sure there's too much difference.

All of which is to say that I endorse overcoming the sense that commenting on old posts is too silly to do.

Idea one: Post a link to an individual post. It could be one of your favorites, or any old post that you consider relevant or want to discuss. Just make a post in the discussion section with a link to the old post plus an explanation of why you reposted it - a sentence saying that it's a great post about X could be enough, or you could give a longer commentary if you have one. Further discussion can take place in the comments to your post.

I'd be interested in the discussions they spawn when they're reposted. Maybe someone will have new insights on the topic, or might be able to explain things differently, or further expand on the point.

Is there a problem with directing people to read Luke Muehlhauser's Reading Yudkowsky posts?

This is a general problem, and deserves a general solution.

Other instances of the problem:

  • You just discovered http://xkcd.com/, and there are nearly 900 of them. How can you push through them at a reasonable rate?
  • You just discovered http://partiallyclips.com/, and you can't even tell how many there are. What next?
  • etc.

We need:
A site that accepts a feed (RSS, Atom) and a release rate, and generates you a feed starting at the start of the old feed and releasing entries on your schedule.

That looks like exactly what I need - thanks… only…

http://rssreplay.heroku.com/create → "Internal Server Error"
(and I'd love to work out what's going wrong so I can fix it, so)…
SOURCE CODE404 this is not the web page you are looking for

http://rssreplay.heroku.com/create → "Internal Server Error"

If you get the date format exactly right (a variant of plain English) it seems to work.

I've been meaning to build such a service for a while, and thinking that rssreplay was down actually pushed me over the edge. If you're interested in testdriving it, let me know.

I'm interested.

feedpacer.appspot.com

you're user #1 :) It don't look pretty (yet), but it works as far as I've tested it.

[-][anonymous]9y 0

It works! It works beautifully!

I really like this idea, perhaps with more popular sequences having a higher probability of being reposted- my habits incline me to reading new posts, not reading old posts, and new members should be drawn in by them.

Are any steps being taken to move forward with any of these ideas?

I have an implementation in mind that won't need any lesswrong codebase modifiactions, based on your idea 2, but I'll write it up for more feedback when I have time to actually materialize it.

I was also thinking of implementing my idea 2, if no superior alternative is in the works. Do you want to talk and compare plans? We could switch to pm or email (my google email is unnamedlw).