When trying to learn technical topics from online expositions, I imagine that most people hit snags at some moment - passages that they can't seem to grasp right away and that impede further progress. Moreover, I imagine that different people often get stuck in the same places, and that a few fortunate words of explanation can often help overcome the hump. (For example, "integral is the area under the curve" or "entropy is the expected number of bits".) And finally, perhaps unintuitively, I also imagine that someone who just overcame a sticking point is more likely to say the right magic words about it than someone who has understood the topic for years.

Hence my suggestion: let's try to identify and resolve such sticking points together, maybe as part of our Simple Math of Everything. This idea might be more appropriate for Hacker News, but I'm submitting it here because it sounds like a not-for-profit rather than a business, and seems nicely aligned with the goals of our community.

The required software certainly exists: our wiki would do fine. One of us posts a copy of a technical text. Others try to parse it, hit the difficult points, resolve them by intellectual force and insert (as a mid-article comment) the magic words or hyperlinks that helped them in that particular case. I really wonder what the result would look like; hopefully, something comfortably readable by people with modest math-reading skillz.

Any number of technical topics suggest themselves immediately - now what would you like to see?

New Comment
8 comments, sorted by Click to highlight new comments since:

I like this idea. It reminds me of this guy's site:

http://betterexplained.com/

Admittedly, he takes quite simple concepts that he expects his audience to already understand in a rote sense and attempts to give them a better intuitive grounding, but it's the same kind of "aha!" principle we all want to make more accessible.

Thanks for that link, I was considering something like that before stumbling across the idea in the article :-)

Sounds like a good project for stackexchange.

I think this is an excellent idea, essentially a study group organized on the web for people who are interested in lesswrong-ish topics.

I'd be interested in learning how to write code that updates a conjugate prior as it sees data - in particular, conjugate priors which are actually structural models.

http://en.wikipedia.org/wiki/Conjugate_prior

The Wikipedia article was quite clear, but I can't yet easily visualize a conjugate prior that's actually a structural model. (Is that the right link?) So your question is still beyond my understanding and it would seem that some other text is required.

Yes, me neither. :P

In what way is this different from the existing technical texts on Wikipedia? The answer to your question: they don't become transparent from all the links. It's like learning a language: having a dictionary doesn't allow you to understand the text very well.

Wikipedia is a very good reference for the topics it does cover, but it doesn't go deep enough to do any kind of meaningful research. Even the reading material for this post of mine is only about 50% Wikipedia. Also, deletionism and the "no original research" rule probably won't allow us to add our topics to Wikipedia.