Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Recent updates to gwern.net (2014-2015)

21 Post author: gwern 02 November 2015 12:06AM

Previously: 2011; 2012-2013; 2013-2014

“Receive my instruction, and not silver; and knowledge rather than choice gold. / For wisdom is better than rubies; and all the things that may be desired are not to be compared to it.”

Sorted by topic:

Darknet market related:

Statistics & decision theory:

QS related:

Tech:

  • “Effective Use of arbtt”: My window tracker/time-logger of choice is arbtt which records X window info for later classification and analysis; but one of the challenges is you don’t know how to set up arbtt or improve your environment or write classifications rules. So I wrote a tutorial.
  • Time-lock crypto: wrote a Bash implementation of serial hashing time-lock crypto, link to all known implementations of hash time-lock crypto; discuss recent major theoretical breakthroughs involving Bitcoin

Debunking:

gwern.net-related:

  • switched to Patreon for donations
  • continued sending out my newsletter; up to 24 issues now
  • rewrote gwern.net CSS to be mobile-friendly; should now be readable in an iPhone 6 browser
  • optimized website loading (removed Custom Search Engine, A/B testing, non-validating XML, outbound link-tracking; simplified Disqus; minified JS, and fully async/deferred JS loading)
  • A/B testing:

Comments (3)

Comment author: Jayson_Virissimo 03 November 2015 04:27:33PM *  2 points [-]

Did you do anything clever to demonstrate that the survey award recipient was really chosen randomly?

Comment author: gwern 03 November 2015 05:32:56PM 2 points [-]

No; I did think about it for a while since I would've liked to have an excuse for more hash precommitments and whatnot, but I couldn't think of any meaningful way to do it without harming participation.

You can use the Bitcoin blockchain for verifiable randomness, but there's no way to prove I didn't edit or sort the survey results to put a sock puppet account in it. I could release a signed copy of the survey data and commit to a blockchain hash for choice, but then I would have had to release the Bitcoin addresses & emails (because otherwise it's not verifiable in any way), which would damage people's privacy; if I had banned email addresses so as to make it safe to release the full survey data so payments to Bitcoin addresses could be verified, then few people would have been able to benefit from the contest - because as I expected, only like 10% of respondents put in a Bitcoin address rather than an email address. And so on.

Comment author: 27chaos 02 November 2015 06:03:46AM 1 point [-]

I'm really really liking the everything correlates with everything observation.