All of JacekLach's Comments + Replies

Yes, this post was very useful as advice to reverse to me. I think it possible now that one of the biggest problems with how I'm living my life today is optimising too hard for slack.

Low-confidence comment disclaimer; while I've had the concept pretty much nailed down before, I never before thought about it as something you might have too much off. After reading this post I realised that some people do not have enough slack in their life, implying you can choose to have less or more slack, implying it's possible too have too much slack.

I don

... (read more)
I think we have to distinguish slack from freedom or indeed total absence of constraints. When I was younger I was fond of saying that freedom is overrated, because all this striving for freedom comes at significant costs of its own. Deliberately limiting oneself can indeed create some slack. For example, I don't have a driver's license (initially for environmental reasons), which might look like a lack of freedom to go where I want. But I noticed that this doesn't take notable options away (I live in a big city with good public transportation). I and my environment adapt and if I really need a car I can take a taxi from all the saved car costs. Maybe not the best example to illustrate this, but the best I currently have on offer :-)

TBH I strongly disagree with OP's suggestion that 95% reliability is low / bad, at least read literally. I personally definitely fail verbal 'soft commitments' ("I expect this will be done by end of week") with way more than 5% rate; probably more like 20-30. Part of it is being in business where hidden complexity strikes at any time, and estimating is hard; part of it is because of cultural communication norms.

If you ignore soft commitments, then the easy way to improve reliability is to make less hard commitments. Instead of "I'll definite... (read more)

I don't think the goal of OPs proposal is to learn any particular skill. To me it mostly looks like trying to build a tightly-knit group so that each member can use the others as external motivators and close friends to discuss life plans and ideas in detail not really possible between modern colleagues and friends. I.e. the goal is not learning a skill, it's building a mutual support group that actually works.

You're looking at content, not status (as implied by 'knocking someone down a peg'). My immediate reaction to the top-level comment was: "well, they have some good points, but damn are they embarassing themselves with this language". Possibly shaped by me being generally sceptical about the ideas in the OP.

As far as the bet is about the form of the post, rather than the content, I think Duncan's pretty safe.

Do you have examples of systems that reach this kind of reliabilty internally?

Most high-9 systems work by taking lots of low-9 components, and relying on not all of them failing at the same time. I.e. if you have 10 95% systems that fail completely independently, and you only need one of them to work, that gets you like eleven nines (99.9{11}%).

Expecting a person to be 99% reliable is ridiculous. That's like two sick days per year, ignoring all other possible causes of failing to make a task. Instead you should build systems and organisations that have slack, so that one person failing at a particular point in time doesn't make a project/org fail.

Well, in general, I'd say achieving that reliability through redundant means is totally reasonable, whether in engineering or people-based systems. At a component level? Lots of structural components, for example. Airplane wings stay attached at fairly high reliability, and my impression is that while there is plenty of margin in the strength of the attachment, it's not like the underlying bolts are being replaced because they failed with any regularity. I remember an aerospace discussion about a component (a pressure switch, I think?). NASA wanted documentation for 6 9s of reliability, and expected some sort of very careful fault tree analysis and testing plan. The contractor instead used an automotive component (brake system, I think?), and produced documentation of field reliability at a level high enough to meet the requirements. Definitely an example where working to get the underlying component that reliable was probably better than building complex redundancy on top of an unreliable component.

The initial argument that convinced you to not eat meat seems very strange to me:

Her: why won’t u eat rabbits? Me: because i had them as pets. i know them too well. they’re like people to me.

This reads to me as: I don't think eating rabbits is immoral, but I have an aesthetic aversion to them because of emotional attachment, rather than moral consideration. Is that not the right reading?

Her: i will get you a pet chicken Me: … Me: omg i’m a vegetarian now :-/

So, you've now built extended your emotional attachment towards rabbits to all animals? Or j... (read more)

0Gordon Seidoh Worley6y
As I've stated in the piece and elsewhere in the comments, I don't think of myself as making moral distinctions, but if you must phrase it in those terms think of me as a preference utilitarian, but you are right that unlike in moral theory proper the source of moral consideration lies solely in my own preferences, which matches more with aesthetic theory if you're inclined to think in that way (like with morality, I view aesthetics as trying to put too much essence in the world as a result of trying to understand it from an insufficiently broad frame). I have no way to say that I think what I did is "good", as in I don't see my actions through the lens of morality so I cannot judge things "good" or "bad". I conceptualize this instead as more completely satisfying my preferences, although I'm open to a more parsimonious understanding. It is probably true that I already believed in favor of being a vegetarian but wasn't acting on it, although I also wasn't trying to be one either, but that is likely relevant. My conversion story should not be taken as an argument for all people to become vegetarians: it's instead an argument for me to be a vegetarian so I can more get what I want.

FWIW (a year later) I read the statistic the same way you initially did, but didn't do the comparison. Sorry! Thanks for doing the maths below and in the edit.

Reminds me of talesofmu. Your strategy looks like trying to play the GM, and is likely to get you punished :)

For me I don't see any reason to prefer archery over a martial art.

And there might not be any reason to do it for you, but other people might be uncomfortable with hitting other people, concerned about their hands (much easier to break a finger or twist your wrist if you're doing martial arts than archery, I imagine), be looking for a relaxing rather than exciting hobby, etc.

Morgan et all (2010) ( estimate 11.1 cyclist deaths per 100000 cyclist-km in London.

Wikipedia ( estimates 8.5 road fatalities per 1 BILLION vehicle-km. claims 125 motorcyclists died in road accidents for every billion miles travelled - the highest rate for all road users but also a year-on-year fall of 11%.

At 41 deaths per billion miles, the mortality ... (read more)

London is an exceptionally unsafe place for cyclists. I doubt it's anywhere near so unsafe as to make up all the difference between cycling and driving in those statistics, but I would guess it's at least 10x worse than the average, maybe quite a bit more. ... Ah, wait, I see you have other statistics saying 35 deaths per billion miles cycled, which would be about 21 per billion km, versus 11 per 100k km or 110,000 per billion km in London. So, apparently London is nearly four orders of magnitude worse than average. That's ... more than I expected. Have I misunderstood or miscalculated something? That "11 per 100k km" figure says that if someone has a 2x5km commute in London that they do 200 days per year, they expect to die about 0.22 times per year; in other words, their life expectancy is a bit less than 5 years. I repeat, London is exceptionally unsafe for cyclists, but I think I'm going to defy the data here. Can it really be that unsafe? [EDITED to add:] Some other figures I've seen suggest that serious injuries are maybe 20x as common as deaths. That means that my hypothetical 10km/day London cyclist should expect four serious injuries per year. Really? There are some statistics on the Wikipedia Cycling in London [] page. For 2014 they report 13 deaths from 610k "daily journeys". Even if a "daily journey" is as short as 1km, that would be 13 deaths per 610k km, which is a lot less than 11 per 100k km. (Though still scandalously large.)

Hm, somewhat, yes. What do you believe?

I mean it's not purely at random, of course, but surely you need to go out and meet a lot of people.

Well, not necessarily []. ;-)

Isn't there anything you already know but wouldn't like to forget?

Yeah, that's pretty much the problem. Not really. I.e. there are stuff I know that would be inconvenient to forget, because I use this knowledge every day. But since I already use it every day, SR seems unnecessary.

Things I don't use every day are not essential - the cost of looking them up is minuscule since it happens rarely.

I suppose a plausible use case would be birth dates of family members, if I didn't have google calendar to remind me when needed.

Edit: another use case that comes t... (read more)

I'm a software dev.

In programming this might mean functions in a particular library that you're working with (the C++ STL, for example)

Right. I guess I somewhat do 'spaced repetition' here, just by the fact that every time I interact with a particular library I'm reminded of its function. But that is incidental - I don't really care about remembering libraries that I don't use, and those that I use regularly I don't need SR to maintain.

I suppose medical conditions looks more plausible as a use case - you really need to remember a large set of facts, an... (read more)

I've used SRS to learn programming theory that I otherwise had trouble keeping straight in my head. I've made cards for design patterns, levels of database normalization, fiddly elements of C++ referencing syntax, etc.

You don't enjoy company of most members-of-your-preferred-sex, but are hopeful that there are people out there that you could spend your life with. The problem is that finding them is painful, because you have to spend time with people whose company you won't enjoy during the search.

By hacking yourself to enjoy their company you make the search actually pleasant. Though hopefully your final criteria does not change.

You seem to be assuming a model where one can only meet a potential mate selected at random from the population, and one'd need to spend a lot of time with her before being allowed to rule her out.

By hacking yourself to enjoy their company you make the search actually pleasant.

"We understand how dangerous a mask can be. We all become what we pretend to be." ― Patrick Rothfuss, The Name of the Wind

"No man, for any considerable period, can wear one face to himself and another to the multitude, without finally getting bewildered as to which may be the true." ― Nathaniel Hawthorne, The Scarlet Letter

People, use spaced repetition! It's been studied academically and been shown to work brilliantly; it's really easy to incorporate in your daily life in comparison to most other LW material etc... Well, I'm comparatively disappointed with these numbers, though I assume they are still far higher than in most other communities

I'm one of the people who have never used spaced repetition, though I've heard of it. I don't doubt it works, but what do you actually need to remember nowadays? I'd probably use it if I was learning a new language (which I don't real... (read more)

I don't know what you work on, but many fields include bodies of loosely connected facts that you could in principle look up, but which you'd be much more efficient if you just memorized. In programming this might mean functions in a particular library that you're working with (the C++ STL, for example). In chemistry, it might be organic reactions. The signs of medical conditions might be another example, or identities related to a particular branch of mathematics. SRS would be well suited to maintaining any of these bodies of knowledge.
Isn't there anything you already know but wouldn't like to forget? SRS is for keeping your precious memory storage, not necessarily for learning new stuff. There are probably a lot of things that wouldn't even cross your mind to google if they were erased by time. Googling could also waste time compared to storing memories if you have to do it often enough (roughly 5 minutes in your lifetime per fact). In my experience anything you can write into brief flashcards. Some simple facts can work as handles for broader concepts once you've learned them. You could even record triggers for episodic memories that are important to you.

Wouldn't a better threat be to switch to Bob anyway? (in which case Alice effectively loses 2 votes instead of one)

No, there's no particular reason to think an FAI would be better at learning than an UFAI analogue, at least not as far as I can see.

I believe you have this backwards - the OP is asking whether a FAI would be worse at learning than an UFAI, because of additional constraints on its improvement. If so:

then a non Friendly AI would eventually (possibly quite quickly) become smarter than any FAI built.

Of course one of the first actions of a FAI would be to prevent any UFAI from being built at all.

I assumed otherwise because of : Which says the FAI is learning faster. But that would make more sense of the last paragraph. I may have a habit of assuming that the more precise formulation of a statement is the intended/correct interpretation, which, while great in academia and with applied math, may not be optimal here.

Find a service where you can listen to music for free with minimal inconvenience (YouTube and Spotify are usable for me, but just barely)

Grooveshark (! Unless you cannot stomach songs playing out of order, or sometimes being repeated (and even that is getting more rare as they work on improving their indexing).

Open page, type artist name (or select genre), press 'play all'. It stops playing every couple hours (which I enjoy - means I can just put music on in the evening and it'll turn itself off eventually - no need to get up) so you'll have to press a single button to resume.

Interesting -- looks like it's come a long way on indexing and general usability since I last played with it (I think it was all Flash back then). I'll give it a second chance, thanks.

Oh, I see. You probably already understood that, but I'll write it up for anyone else who didn't initially grok the process (like me).

Intuitively, the original algorithm incentivises people to post their true estimates by scaling up the opponents investment with your given odds, so that it doesnt pay for you to artificially lower your estimate. The possible wins will be much lower; disproportionately to your investment, if you underestimate your odds. Conversely, the possible losses will not be covered by increased wins if you overestimate your chances.

It ... (read more)

What's wrong with just using this algorithm to establish ratios between bets, then scaling up to meet whichever limit is hit first?

In your example, it'd be scaled up to 5.12 against 25.

2Scott Garrabrant9y
That is not strategy proof

Neither probability should be <50%, you take the probability that your opinion is the right one, not whether the proposition is true or false.

In your example B would be betting against his beliefs, thus the negative result.

The right calculation: A = 0.6 B = 0.7

A pays: (A ^ 2 - (1 - B) ^ 2) * 25 = (0.36 - 0.09) * 25 = 6.57
B pays: (B ^ 2 - (1 - A) ^ 2) * 25 = (0.49 - 0.16) * 25 = 8.25


actually, it's sufficient that A and B sum to over 1. Since you can always negate the condition, the right calculation here is:

A = 0.4
B = 0.7

A pays: (A ^ 
... (read more)
[This comment is no longer endorsed by its author]Reply

Well, it's none of anyone elses business, so I don't see how other people being there is relevant.

If you mean it in the sense of "don't settle for someone who isn't going to help you with kids, no matter how good a match you otherwise are"... Never settle is a brag

The guy doesn't want children, but he doesn't mind having children with the woman as long as it's not too bothersome for him. The woman either really wants children, in which case this arrangement is to her benefit, or does not want children that badly, in which case they don't have children.

You speak as if these two are the only people on the planet.


Now I'm even more confused. How can my answer be useful if they don't know how I interpret the question? Esp. since my answers are pretty much opposite depending on the interpretation...

My bad for not finding that comment. I skimmed through the thread, but didn't see it.

I'm confused by the CFAR questions, in particular the last four. Are they using you as 'the person filling out this survey' or the general you as in a person? "You can always change basic things about the kind of person you are" sounds like the general you. "You are a certain kind of person, and there's not much that can be done either way to really change that" sounds like the specific you.


The ambiguity is intentional, apparently [].

I think if you would ask those people they would also say yes, that they are thinking about ways of solving their problems.

Not necessarily. They might say it's too big to solve, or "it's not really a big deal" when it obviously is, or that it's not their responsibility to solve, or any of multum other excuses that validate not changing.

That does sound like a good idea. Browsing the google groups, the next occasion seems to be the CZE outing on 1. Sep. (

Would that meetup be suitable, or is it more practical rather than a getting to know people meeting? Or maybe it could be both? I might wait and go to the next meetup after that.
Not for registering interest. As things move forward, we'll pool more information and preferences, getting into the details more. I would advise anybody who's interested to make a point of coming to a couple of meetups, but if that's very inconvenient, still register interest anyway.
It is generally also a good idea to come and meet us at a Meetup.

But there is a significant difference between taking a medical formula under doctors supervision and mixing up the most common nutrition ingredients and claiming it to be a cure-all-be-all food. Didn't the guy forget to include iron in his first mixture?

Another 'Soylent' equivalent I know of is Sustagen Hospital Formula.

As a 1.4999999999999 boxer (i.e. take a quantum randomness source for [0, 1], take both boxes if 0, one box if 1, one box if something else happens), I don't think scenario C is convincing.

The crucial property of B is that as your thoughts change the contents of the box change. The casualty link goes forward in time. Thus the right decision is to take one box, as by the act of taking one box, you will make it contain the money.

In C however there is no such casualty. The oracle either put money in both boxes, or it did not. Your decision now cannot possibly... (read more)

It's from Terry Pratchett's Discworld series.

Lord Vetinari also has a strange clock in his waiting-room. While it does keep completely accurate time overall, it sometimes ticks and tocks out of sync (example: "tick, tock... ticktocktick, tock...") and occasionally misses a tick or tock altogether, which has the net effect of turning one's brain "into a sort of porridge". (Feet of Clay, Going Postal).

Thank you. I wonder how hard it would have been to build such a clock at Discworld's tech level. It might require magic.

I think you underestimate the power of the GHD. If Hermione really believed she had to kill Draco or he will, for example, murder every student in Hogwarts the next day, I'm pretty sure she would cold-bloodedly kill him.

Spells that extract the history of spells casted using a wand are canon, afaik (or was that just the most recent spell?)

I would expect they were casted on hermiones wand and the usage was confirmed.

Thing is, it only checks the most recent. Any minor spell cast afterwards clears the prior incantem, so a different spell showing is not going to count for much. And they never actually said that they had that piece of evidence, which they certainly would have if they'd known.

Lack of sabotage is obviously evidence for a fifth column trying to lull the government, given the fifth column exists, since the opposite - sabotage occuring - is very strong evidence against that.

However lack of sabotage is still much stronger evidence towards the fifth column not existing.

The takeaway is that if you are going to argue that X group is dangerous because they will commit Y act, you cannot use a lack of Y as weak evidence that X exists, because then Y would be strong evidence that X does not exist, and Y is what you are afraid X is going to do!

You would be much better off using the fact that no sabotage occurred as weak evidence that the 5th column was preventing sabotage.

If there is other evidence that suggests the 5th column exists and that they are dangerous, that is the evidence that should be used. Making up non-evidence ... (read more)