Working Mantras
While working with Marcello on AI this summer, I’ve noticed that I have some standard mantras that I invoke in my inner dialogue (though only on appropriate occasions, not as a literal repeated mantra). This says something about which tropes are most often useful—in my own working life, anyway!
“If anyone could actually update on the evidence, they would have a power far beyond that of Nobel Prize winners.”
(When encountering a need to discard some idea to which I was attached, or admit loss of a sunk cost on an avenue that doesn’t seem to be working out.)“The universe is already like [X], or not; if it is then I can only minimize the embarrassment by admitting it and adapting as fast as possible.”
(If the first mantra doesn’t work; then I actually visualize the universe already being a certain way, so that I can see the penalty for being a universe that works a certain way and yet believing otherwise.)“First understand the problem, then solve it.”
(If getting too caught up in proposing solutions, or discouraged when solutions don’t work out—the immediate task at hand is just to understand the problem, and one may ask whether progress has been made on this. From full understanding a solution usually follows quickly.)“Load the problem.”
(Try to get your mind involved and processing the various aspects of it.)“Five minutes is enough time to have an insight.”
(If my mind seems to be going empty.)“Ask only one thing of your mind and it may give it to you.”
(Focusing during work, or trying to load the problem into memory before going to sleep each night, in hopes of putting the subconscious to work on it.)“Run right up the mountain!”
(My general visualization of the FAI problem; a huge, blank, impossibly high wall, which I have to run up as quickly as possible. Used to accomodate the sense of the problem being much larger than whatever it is I’m working on right now.)“When the problem is solved, that thought will be a wasted motion in retrospect.”
(I first enunciated this as an explicit general principle when explaining to Marcello why e.g. one doesn’t worry about people who have failed to solve a problem previously. When you actually solve the problem, those thoughts will predictably not have contributed anything in retrospect. So if your goal is to solve the problem, you should focus on the object-level problem, instead of worrying about whether you have sufficient status to solve it. The same rule applies to many other habitual worries, or reasoning effort expended to reassure against them, that would predictably appear as wasted motion in retrospect, after actually solving the problem.)“There’s always just enough time when you do something right, no more, no less.”
(A quote from C. J. Cherryh’s Paladin, used when feeling rushed. I don’t think it’s true literally or otherwise, but it seems to convey an important wordless sentiment.“See the truth, not what you expect or hope.”
(When expecting the answer to go a particular way, or hoping for the answer to go a particular way, is exerting detectable pressure on an ongoing inquiry.)
I don’t listen to music while working, because of studies showing that, e.g., programmers listening to music are equally competent at implementing a given algorithm, but much less likely to notice that the algorithm’s output is always equal to its input. However, I sometimes think of the theme Emiya #0 when feeling fatigued or trying to make a special demand on my mind.
- Is AI forecasting a waste of effort on the margin? by (EA Forum; 5 Nov 2022 0:41 UTC; 10 points)
- 's comment on I Want To Believe: Rational Edition by (19 Nov 2014 17:19 UTC; 7 points)
- 's comment on Useful maxims by (13 Jul 2012 2:06 UTC; 5 points)
- 's comment on What epistemic hygiene norms should there be? by (22 Mar 2012 8:25 UTC; 4 points)
- 's comment on 11 core rationalist skills by (2 Dec 2009 23:31 UTC; 3 points)
“Amdahl’s law is a bitch.”
(When you’re trying to improve something, focus on the parts that contribute most to the problem. If something takes 10% of your time and you make it a million times faster, that still only gives you about a 10% improvement overall. Fun fact: in the US federal budget, only about 18% of it is discretionary spending—the rest of it is mandatory spending on things like interest and Medicare, so a big push to remove bridge-to-nowhere pork really wouldn’t do much to reduce spending. Or a more personal example: did you know that by far most of the power that low-power CPUs use is spent just moving data across the chip, and not on logic? And yet people still spend countless man-hours on making the logic use less power, and getting diminishing returns. People can have a hard time seeing Amdahl’s law even when it’s staring them in the face.)
“Amdahl’s law is a bitch” = Meta-Amdahl Law?
I’ve definitely done #7: I was far, far behind everyone and exhausted, backpacking to a high ridge in the Grand Tetons. Something snapped, and I just ran up the absurd climb for a good hour or so, passing everyone several times over. By the time I got to the top, I screamed loudly to the world that I am the walrus.
The point is that you may lose your mind in the process of attacking FAI problems directly and relentlessly, and even if you succeed you may afterward believe that you are a walrus. It’s an occupational hazard.
For the uninitiated—what is a FAI problem?
Friendly Artificial Intelligence.
Why hazard?
Aren’t you the walrus?
Link to the previous discussion on listening to music while working—on OB.
A mantra of mine since childhood: “If at first you don’t succeed, you’re doing something wrong.” (Stop and try to really understand the elements of your approach, when feeling frustrated).
If you always succeed at first, you’re not aiming high enough.
Unless there are some failures, it’s near certain that you’re not approaching the maximal rate of change of which you are capable (and are probably being overly conservative in what you attempt in order to sustain the perfect record).
But if your first approach to every task is right, you’re not picking difficult enough tasks.
Sometimes a good approach will fail anyway, and sometimes failure indicates the need for a new approach… but your mantras are catchier.
This is absolutely true. I’m a little shocked at how many programmers and other ‘mind-workers’ (students are especially bad at this) think that putting on some music is helpful, or neutral. It’s neither; even classical music or chants hurt.
(I have no studies handy for this, but I’ve tried music with Mnemosyne, Gbrainy, and Dual N-back, and in all 3 and all the sessions, my statistical performance was hurt to a greater or less extent.)
It occurs to me that I made something of a paternalistic assumption here: that people who do think they aren’t harmed, or are helped, by music are simply wrong/mistaken about it. Perhaps it’s simply that some people can handle the music and some can’t, and I fall into the latter?
I’m glad to say that just today I saw media coverage of a study on multitasking indicating that people used to multitasking are both worse than singletaskers and deluded about it: http://news.bbc.co.uk/2/hi/technology/8219212.stm
So there! :)
Music seems to work sometimes with getting started on things. Maybe it does narrow the cognitive space, but this is actually a good thing as long as the mind isn’t really occupied with the actual work and is instead full of random, distracting thoughts. So maybe there should be a music player mode which starts normally, but then slowly fades out during the next ten minutes or so?
Music can also alter mood, which could likewise be useful for getting started on working.
One thing I’d be interested in knowing is whether the familiarity of the music matters. I wouldn’t be surprised if completely novel music caused a larger cognitive load than familiar music. What about if you’ve heard the piece of music once, ten times or a hundred times?
Also, what effect does the type of the music have? Lyrics probably make a difference. What about music that doesn’t even have many recognizable melodies, such as ambient or trance? What about white noise?
Additional anecdotal evidence:
I listen to music when I’m doing work that I’m not unusually interested in. It helps me zone out background noise and keeps me from jumping online to waste time reading something more interesting. I can never listen to new music while studying though and overly complex music (particularly classical) ends up drawing my attention and I end up following the music. I prefer dance music, it is repetitive enough that I don’t have to listen to every note to follow it, it usually has few lyrics and the beat energizes me.
Next time I have a paper I’ll try to stop the music once I’ve gotten a good start and see if that makes a difference.
This is worth thinking about. I was trying to figure out why people like music during demanding tasks so much, and maybe it’s a way of distracting themselves and committing only partially to the task—maybe studying or whatever is too ‘painful’ to fully concentrate on. (There’s a connection to akrasia here, I’m certain.)
As I said, any music seems to be a negative. I can’t speak to ambient or trance or white noise simply because I don’t play them (if you’re really curious about the full gamut of music I tried out, I can say I tried the majority of my collection, and if you’re curious what that entails, see http://www.last.fm/user/gwern ); however, I did try out various binaural beats generated by Sbagen which sounded a lot like white noise and had a similar negative effect.
I wonder if I should turn off the music when playing certain video games, then?
Even if music doesn’t improve performance, putting on a Final Fantasy soundrack can still help me feel less frustrated. My auditory cortex rarely shuts up; if I’m not thinking in words at the moment, I very often find that I have some kind of tune looping in my head.
I also prefer to work without music, but I use it to switch my brain into “action mode”.
The music I tend to use for such mood-switching is usually energetic and sometimes aggressive (examples include action pieces from movies like Predator 2 or Matrix, various electro DJs, selected rap / hip-hop etc.) -- and I stay away from serious music (in my case its mostly Arvo Pärt, John Tavener and J.S. Bach).
Also, music is sometimes helpful for mentally isolating myself from coworkers, especially when they bug me with questions—however, just putting the headphones on without music can do the trick. I read that Tim Ferris does that too.
If anyone does have studies to hand I’d be grateful for references.* I personally find it difficult to work without music. That may be habit as much as anything else, though I expect part of the benefit is due to shutting out other, more distracting noise. I’ve noticed negative effects on my productivity on the rare occasions I’ve listened to music with lyrics, but that’s about it.
* I’d be especially grateful for anything that looks at how much individual variation there is in the effect of music.
Do you rely mostly on visual imagination?
Sometimes, but it varies quite a lot depending on exactly what I’m doing. The only correlation I’ve noticed between the effect of music and work-type is that the negative effect of lyrics is more pronounced when I’m trying to write.
Of course, it’s entirely possible that I’m just not noticing the right things—which is why I’d be interested in references.
The idea is that when you are listening to music, you are handicapping yourself by taking some of the attention of the aural modality. If you are used to rely on it in your thinking, this makes you impaired.
This is related to an experiment that Feynman describes in this video:
Feynman ‘Fun to Imagine’ 11: Ways of Thinking
In the experiment, you need to count in your mind, while doing various activities. That attention was really paid to the counting is controlled by you first calibrating and then using the counting process to predict when exactly a minute has passed. Thus, you can’t cheat, you have to really go on counting.
Feynman himself says that he was unable to speak while counting, as he was “speaking” and “hearing” these numbers in his mind. Another man he asked to do the experiment had no difficulty speaking, but was unable to read: he then explained that he was counting visually.
I tried it both ways, and the difference shows in different speeds of counting in these modes which are hard to synchronize (and so just switching between them doesn’t work very well).
Thanks for the explanation.
I’d heard something similar from a friend who majored in psychology, but they explained it in terms of verbal processing rather than auditory processing more generally, which is why (they said) music without words wasn’t as bad.
I’m not sure whether it’s related, but I’ve also been told by a number of musically-trained friends that they can’t work with music at all, because they can’t help but analyse it as they listen: for them, listening seems to automatically involve processing work that it doesn’t (seem to) for me, precisely because I’m not capable of such processing. (This was part of the reason I was originally wondering about individual variation; the point you make at the end is really interesting in this regard too.)
In a possibly-related anecdote, I can’t listen to music I’ve played in Guitar Hero while working, as my mind switches into Guitar Hero mode and all I see are streams of colored buttons.
I find that very interesting too, since I am in fact the opposite of your musically-trained friends: I am quite rubbish at anything musical, am hard-of-hearing, and have great difficulty analysing music & songs. (In part that’s why I listen to so much J-pop: since I often can’t understand the lyrics even if they’re in English...)
It seems obvious to me that it’s a hindrance—still, putting on headphones and covering ambient noise with music might be less distracting.
I often need to start some work with music, because I constantly have an internal jukebox running. A lot of people listen to music to fuel that narrative illusion, which gives them motivation, and some do so to block out more distracting noise, as conchis mentioned. Mine is a variant on the latter, so I just get something with low information content stuck in my head before sitting down for serious work.
(I make an exception in certain stages of writing fiction, since I often soundtrack scenes as I create them, or even create scenes that fit the melody.)
Yep. Sometimes very distracting, as I do remixes in my head … might try pink noise and see how that goes :-D
Of course when you’re on your deathbed, alone (save the cryonics team at your side), unknown, with no significant accomplishments to your name clinging to the slim hope of revivification a hundred years hence and you’ve spent every last dollar and ounce of energy on solving a problem you’re just not smart enough to understand you’re really gonna wish you had second guessed yourself back in the day.
… But hey, I’m an optimist.
Edit: Yikes people. I was joking. Sorry.
Well, you’ve pretty clearly stated the sort of terrifying insecurity that most people wouldn’t say out loud and confront. But you can’t actually solve a problem by musing on that, whether to be afraid of it, or to develop arguments against your fears; the whole thing is just a wasted motion from the perspective of actual problem-solving. Time spent reading a math paper will always be better spent, once it’s been determined that you are going to take the problem as a goal.
If you’ve already decided that trying to solve a given problem is the best use of your time (and this belief is as updated as it can be) then yes continuing to worry is unhelpful. (Though worries can motivate honest assessments of how one is spending their time.) We agree.
Anyway, I just thought I was being funny.
Upvoted, but FYI, it’s really hard for people to know whether or not that sort of thing is humor or malice—I was actually trying to guess that (thinking that it would be funny/rational as humor/look-into-the-dark), and ended up leaning toward malice by prior probability (guess I should have realized prior frequencies are skewed on this particular website, but I’d been reading older OBLW posts and looking at the lower-quality comments there, that may have skewed my intuitive estimate).
You’ve got to go further over the top to make it clear that it’s humor—talk about my gasping dying breath, “If only… I’d been.. less… overconfident” or something like that. (Actually, even that might not be over-the-top enough. “If only I’d been… less confident… about my own relative meta-rationality...” would do it, that tells everyone you’re an insider.)
You sound like a manipulative and troubled person, what’s wrong with you? :-)
Eliezer wishes to maximize expected utility. If that means dying for a 0.001 chance of saving the world, he’s willing to do that. If the dice fall unfavorably, he’s willing to accept that.
Except that EY doesn’t spend all his time (or even most of his time, I don’t know) trying to solve impossible problems. He also blogs and helps organize conferences and stuff, right? All those other activities act as a hedge against him not accomplishing anything in FAI. Its not such a big deal to work on a problem that you never actually solve if your real career is as a writer/speaker and talking about the fact that you are working on this impossible problem is useful for that career.
Michael Vassar and others organize the Summit.
Yeah, when the Summit was first proposed in 06, I said, “This can only happen if you make absolutely sure that I never have to do any work on them.”
Interesting list. I’d be curious to see other people’s.
My experience with (3) is the opposite, in three respects.
First, I often encounter problems in my research that I understand completely. No magic or dark corners remain. Nonetheless, finding a solution remains challenging. When I do find the solution, I don’t feel as if the problem has been further illuminated.
Second, the various lines of attack I develop with an impartial understanding of a problem are usually much more creative and varied than after my understanding fully solidifies. When I first encounter a problem I usually take at least a day or two to enumerate possible solutions before actually studying the problem.
Three, I’ve successfully solved complex problems I didn’t understand at the time by blindly trying approaches my intuition threw at me.
I’d really like to see an example. For me, understanding the solution is part of the problem. I read “understand completely” as “can characterize the solution to the problem as a mapping from inputs to outputs”, and such a characterization should itself be a solution.
I work in applied math. The bulk of my work is aimed at doing something efficiently. Given a problem (a physical model, plus some discretization), I’ll have to find suitable algorithms to solve the resulting equations. I usually understand enough about the physics (and artificial physics introduced via the discrete setting) to have a complete understanding of the limits of the efficiency of any implementation. Nonetheless, actually creating the implementation which maximizes efficiency is not straightforward (otherwise I’d have no job!).
I guess you could classify this as two different problems: the base problem of understanding, in an abstract sense, the bounds on efficiency, and the actual problem of constructing algorithms which will touch those bounds. In my mind though, they feel intimately related in a way that I can’t unravel.
You are disputing definitions of the word “understand”.
ETA: Wrong.
Explicitly so, and asking for an example to be sure.
Actually, no, you are not disputing definitions, you are asking for a clarification. My bad.
I’ve found that I mentally verbalize several mantras taken from the Twelve Virtues.
“There is a time to confess your ignorance and a time to relinquish your ignorance.”—This one comes up less at work, as in other cases. I’ve been in a jiu-jitsu class where a student, after being shown a technique, later at the water cooler said “I don’t understand what he was trying to demonstrate.” This struck me as dumb, and triggered this mantra in my head, as the instructor would have gladly redone the move any number of times had the student asked. At work I am surrounded by my many of my industry’s best developers (this is not an exaggeration). Sometimes I run into a problem I do not know how to solve, but I know they do. This mantra helps me swallow my pride and ask for the information I need to learn how to solve the problem. (However, see #3 below.)
“Let the winds of evidence blow you about as though you are a leaf, with no direction of your own.” This one has come into play lately when I’ve been reading or listening about political issues. I used to be very conservative, then later libertarian, but now I’m trying to mentally divorce my identity from political association entirely. I see this as similarly difficult a transition as when I became an atheist after being Christian. I still get easily baited by left leaning editorials, etc. I do not like these reactions, but they happen despite my conscious desire to not have them. So, I’ve taken to largely expunging my political beliefs in an attempt to move more toward an evidence driven world view. When you start to look at news sources in this light you find that most lack useful information to operate on and you feel less guilty about unplugging from those sources.
“Those who wish to fail must first prevent their friends from helping them.” This concept has always been significant to me, but this phrase captures the essence. When I am curious, I attempt to solve the problem on my own and do my own research. There are times I get information from those who are wiser, but in those cases I prefer hints over spoilers. The key is to get help, but not too much help.
Some more on #3 that occurred to me while re-reading my post. I’ve been working on Project Euler. So far the problems haven’t slowed me down, but looking ahead I can see some really tough ones. I’ve set a goal that I would try and learn, myself, what I needed to solve the problems. This may be impossible. At what point do I condemn myself to absolute failure unless I turn to an expert for help? How do I get the right amount of help on a subject I don’t know, because, in not knowing, I am less able to distinguish between too much help and not enough? If I have a good problem to solve and someone hands me an algorithm that solves it, it robs me of the chance of deriving the algorithm. But if I lack the requisite tools to derive the algorithm, I may also not recognize the difference between the tools and the algorithm itself.
So there are three of the twelve virtues that continually pop into my mind.
Those aren’t my exact intended meanings on some of them—but they’re good meanings nonetheless!
(Usually I’m suspicious of the alleged “accidental optimization” in that sort of reinterpretation, but I deliberately wrote the Twelve Virtues in evocative/poetic language, so this is much less suspicious than it would be with sentences from a random LW post.)
I think I understand your intended meaning for most of the Twelve Virtues and would take a bet validated by a test of the subject matter (hah!), but sometimes what sticks in my mind is an isomorphism I perceive between a statement and some experience. I recognize a danger that the association created by the perception of the isomorphism creates a cached thought that might be flatly incorrect or (at the very least) could obscure a deeper, more valuable truth. Is this what you tend to be suspicious of? Or is it that you are suspicious of someone claiming an intentional reinterpretation and then saying its better than what you meant in some way?
I’m suspicious of the idea that an author could mean one thing, and yet accidentally form a statement which has a different interpretation that is still true, like taking the syllables of a true sentence in Japanese, reciting them out loud, and finding that they form a true sentence in English. My childhood experience with Judaic “reinterpretation” of inconvenient Biblical passages may have something to do with this.
But as said, this is far more plausibly going to happen with poetic, evocative, indirect statements like those in the Twelve Virtues, than with sentences from a random OBLW post.
Didn’t Hofstadter do something like a “grammatically correct sentence in one language that is also phonetically correct in another language” in one of his books or articles? Not that it undermines your point, but I couldn’t find it and vaguely remember seeing it before.
http://pugs.blogs.com/audrey/2009/04/post.html
That’s not really the point, though, is it. I mean, I can write “Tengo tu estudio” (I have your studio) which is phonetically the same as “Ten go to a studio”. However, the only meaning they share is due to one word having the same parent (something I wonder about the Korean Tang and the Japanese -tan? but probably not). In any case, cross-language phonetic coincidences are much less rare than an author unintentionally including dual meanings in his text on a level deeper than “That’s what she said”. Which is what bothers me about people who read too much into Ulysses.
These remind me so much of the Oblique Strategies cards: http://en.wikipedia.org/wiki/Oblique_Strategies
If you’ve not seen them EY, you might want to read through and see if they give you any ideas for more mantras.
“If the problem is hard, study all the approaches you can find to thinking about even slightly related problems.”
Read voraciously, do your homework.
Well, okay, but here I’m listing things that I actually find myself thinking out loud during a typical working day. Not just general good advice.
That’s part of the everyday practice: always look for more stuff to study, in particular when a new idea comes to mind, first try to find more on it in the existing science, and only if nothing seems to be related, commit to trying to develop it yourself. And still, never stop searching for the related stuff in the background.
Feeling this one. So odd how having a really solid grounding in a subject allows you to work out what later seems to be basic truths.
I needed 9. There’s not enough time to get utility U out of the project, so the right thing to do is to get U-100 out of it, and there’s time for that. Owning up to it doesn’t make it worse.
I experimented on myself plenty back in the day. Lots of my experiments had to do with pain—refusing pain medication when it was available to me, prodding places that pained, using more medication than helpful and so on. I remember one instance when I found myself in immence pain and some words came to mind from a Wikipedia article on a Buddhist mantra I had heard long ago. I read that the prescription for this branch of Buddhism was the mere repetition of the line: I take refuge in Amida Budda. Later, I found that the mantra is longer but it was an amazing altered state to find tranquility in that unlike any other purposeful altered state I had experimented with to cope with the pain (e.g. mindfulness, gratitude, imagery....)
I don’t use anything of the sort—I just try to keep my mind on the problem, turning it over and inspecting it, and scream “THINK!” whenever my mind begins to wander. If I get tired, I stop thinking and resume another day. After a while the problem ends up solved. Not sure if this approach will work for others.
“There’s always just enough time when you do something right, no more, no less.”
I see this as related to one of my habits—when I feel like I’m short on time, I’ll hurry. That is, I’ll try to move faster than I can really regulate well, presumably as a sort of self-signaling that I don’t have enough time, but I’m trying hard.
I don’t have a mantra for the the situation, but if I find myself thinking “I’m trying to hurry”, it’s an indicator that I need to stop pushing so hard.
Don’t have many mantras, although I stress the importance of understanding before trying to solve.
One that does stand out is more of a question:
“What am I not thinking here?” or “What are we forgetting here?”—Followed by estimations based on meta-biases and human error tendencies to make some hypotheses where cognitive, social, or cultural blind spots might be. And then comes the testing, followed by more hypotheses. And so on.
After all, every field of thought is developed by humans. It’s a common point of failure.
Any pointers to those studies?