Use RAID on ZFS. RAID is not a backup solution, but with the proper RAIDZ6 configuration will protect you against common hard drive failure scenarios. Put all your files on ZFS. I use a dedicated FreeNAS file server for my home storage. Once everything you have is on ZFS, turn on snapshotting. I have my NAS configured to take a snapshot every hour during the day (set to expire in a week), and one snapshot on Monday which lasts 18 months. The short lived snapshots lets me quickly recover from brain snafus like overwriting a file.

Long lived snapshotting is amazing. Once you have filesystem snapshots, incremental backups become trivial. I have two portable hard drives, one onsite and one offsite. I plug in the hard drive, issue one command, and a few minutes later, I've copied the incremental snapshot to my offline drive. My backup hard drives become append only logs of my state. ZFS also lets you configure a drive so that it stores copies of data twice, so I have that turned on just to protect against the remote chance of random bitflips on the drive.

I do this monthly, and it only burns about 10 minutes a month. However, this isn't automated. If you're willing to trust the cloud, you could improve this and make it entirely automated with something like rsync.net's ZFS snapshot support. I think other cloud providers also offer snapshotting now, too.

Your welcome. I wasn't planning on this, but I would have been leaving money on the table from a coworker's private matching funds.

As much as I like Omnilibrium as a concept, there's a lot of work needed to productize the site. A lot of the style is whitespace, there are boxes instead of icons (notably the bookmark icon), there's no password reset (and the site barfed on my >20 character autogenerated password so that if I lose my current browser profile, I'm going to lose my account there), etc. But even worse, people didn't move there en masse so the site was never bootstrapped.

I'm not convinced that the karma system as it exists today actually performs its desired task anymore because a good chunk of the voting seems to be done by the unquiet spirits. Back when I cared about karma here, it was because it reflected the opinions of people that I very much respected. I don't feel that way anymore.

One possible[*] solution would be to port the Omnilibrium algorithm back to LessWrong, customizing the scoring for each user, but this might be a place where we should hold off proposing solutions.

[*] As in, "Well I suppose that's technically possible, but..."

It's not just you. I can't use the arrow keys either. Chrome 45 on Windows 8.

I'm not sure about the mathematical details, but as described in their FAQ, they presume that it's inevitable that people will form into local Blue and Green tribes, so they attempt to cluster the population into Blue and Green to not just be a better recommendation engine to both Blues and Greens, but also calculate a nonpartisan score of upvotes by the other side and downvotes by your side.

In general, I thought this was fascinating because it gets to the heart about what voting is for on social websites. If we're trying to build a recommendation engine, having an extremely diverse set of viewpoints is probably something that we want in the input stream of links and discussion. However, we then don't want to have everyone's voting then represent a single score variable, because people are different and have different worldviews. Mixing everyone's scores together will make a homogenized mess that doesn't really speak to anyone.

The idea of tracking partisanship not just to Bayes voting to make better recommendations to users, but to get a sense of nonpartisan quality really impressed me as an idea that's totally obvious...in retrospect. I do wonder how well it scales, as Omnilibrium is fairly small right now.

I am going to publicly call for banning user VoiceOfRa [...] VoiceOfRa almost certainly downvote bombed the user who made the grandparent comment, including downvoting some very uncontroversial and reasonable comments.

Consequentially...why bother even if this is true?

Assuming you are correct, Eugene's response to being banned (twice!) was to just make another account. It's highly likely that if you ban this new account, he will make a fourth account. That account will quickly quickly gain karma because, as you note, Eugene's comments are actually valuable. You are proposing that we do the same thing a third time and expect a different result.

Possible actual solutions that are way too much work:

• move LW on to an Omnilibrium like system of voting where Eugene's votes will put him strongly into the optimate cluster and won't hurt as much.

• give up on moderation democracy on the web.

(Disclaimer: My lifetime contribute to MIRI is in the low six digits.)

It appears to me that there are two LessWrongs.

The first is the LessWrong of decision theory. Most of the content in the Sequences contributed to making me sane, but the most valuable part was the focus on decision theory and considering how different processes performed in the prisoner's dilemma. Understanding decision theory is a precondition to solving the friendly AI problem.

The first LessWrong results in serious insights that should be integrated into one's life. In Program Equilibrium in the Prisoner's Dilemma via Lob's Theorem, the authors take a moment to discuss the issue of "Defecting Against CooperateBot"--if you know that you are playing against CooperateBot, you should defect. I remember when I first read the paper and the concept just clicked. Of course you should defect against CooperateBot. But this was an insight that I had to be told and LessWrong is valuable to me as it has helped internalize game theory. The first year that I took the LessWrong survey, I answered that of course you should cooperate in the one shot non-shared source code prisoner's dilemma. On the latest survey, I instead put the correct answer.

The second LessWrong is the LessWrong of utilitarianism, especially of a Singerian sort, which I find to clash with the first LessWrong. My understanding is that Peter Singer argues that because you would ruin your shoes to jump into a creek to save a drowning child, you should incur an equivalent cost to save the life of a child in the third world.

Now never mind that saving the child might have postive expected value to the jumper. We can restate Singer's moral obligation as a prisoner's dilemma, and then we can apply something like TDT to it and make the FairBot version of Singer: I want to incur a fiscal cost to save a child on the other side of the world iff parents on the other side of the world would incur a fiscal cost to save my child. I believe Singer would deny this statement (and would be more aghast at the PrudentBot version), and would insist that there's a moral obligation regardless of the other theoretical reciprocation.

I notice that I am being asked to be CooperateBot. I don't think CFAR has "Don't be CooperateBot," as a rationality technique, but they should.

Practically, I find that 'altruism' and 'CooperateBot' are synonyms. The question of reciprocality hangs in the background. It must, because Azathoth both generates those who are CooperateBot and those who exploit CooperateBots.

I will also point out that this whole discussion is happening on the website that exists to popularize humanity's greatest collective action problem. Every one of us has a selfish interest in solving the friendly AI problem. And while I am not much of a utilitarian, I would assume that the correct utilitarian charity answer in terms of number of people saved/generated would be MIRI, and that the most straightforward explanation is Hansonian cynacism.