The Choice of Work

During the back and forth of exchange with a technical recruiter, he finally asked me what I was looking for. And so the floodgates opened.

This may sound weird, but I pretty much choose employment or position based on the quality of work I think I can do, alone. And when I say quality of work, I mean more than the environment, or the IQ of those I’ll be working with. I mean the opportunity to do what may lead to, first-class work, Nobel-prize winning kind of work.¹ I look for places where I am upstream of technological change: that means, any deals with, say, old guard media companies or the Fortune 500, while they could be good, don’t really interest me. I want to work on what might become fundamental changes in the way people use technology to do things. But that doesn’t mean that I need to tackle the largest problems I can see right off. What matters is that I, personally, have a reasonable approach. I mustn’t forget to, as Richard Hamming says, ‘plant the little acorns from which the mighty oak trees grow’ — small projects can, strikingly, accumulate momentum and value.¹

I find little value in submitting myself to some company culture, and more in preserving my professional ethics, ambitions, and goals concurrently: meaning, for example, I would like to be able to develop new methods, make them available by open sourcing them and popular by evangelizing them through my blog, and having the opportunity to teach them, given the chance. I think that the personality and ideas that excellent people bring to a community of makers will, naturally, evolve into a company culture worth having. But most big companies have grown too large to make this possible, much faster than they can reasonably build trust. And so this natural, enjoyable, healthy culture is replaced with a kind of proxy — virtual intracompany nationalism, fascism even. I not only find this distasteful, I’ve barely learned how to work within it at all — only if there’s deeply meaningful work to be done. And why bother?

Angling to be upstream of technological change, I find unusual answers: Given the choice between a VP position at a big five media company with oodles of benefits, and, say, work at an early netscape or google for a totally minimal salary, and I’ll choose the latter every time. I choose places that I can learn a lot about: where learning I’d learn about more than tradition. It is, for example, not important for me to learn about how to run a large organization: if I ever do, it won’t be in the way that most others are run. I do, however, intend to help grow large, leaderless, open organizations, and so I’d do almost anything for a chance to work with Caterina Fake, or Linus Torvalds.

I want to work on something I find deep personal meaning in. I strongly believe in supporting open culture. I don’t think I’d work for long in games or entertainment unless it would influence change in the social dynamic of collaboration and creation. I worked at MochiMedia because it made possible an income stream for small independent developers where none existed before. This finally opened up professional game development from bigbox studios. Now much of the innovation in gaming comes out of bedrooms. An independent game developer can now commit to their art in a way they couldn’t before.

Similarly, I’d work at YouTube rather than Hulu, even though one’s a startup and the other isn’t, because they’re more interested in involving everyone in the process. As Clay Shirky says, they’re interested in ‘finding the mouse’.

I want to work somewhere where I can really make a difference. Why am I working in technology at all? Archimedes once said, ‘If you give me a lever and a place to stand, I can move the world.’ Technology is my lever. I need only find place to stand. This makes me wary of startups that try to do good, but aren’t particularly focused on doing it efficiently. I wouldn’t work for most charities. There’s too little pressure on them to focus — the tempering influence of market competition is replaced by government demands for ‘accountability’, which isn’t nearly as powerful. And there are numerous ‘ecogreen’ websites out there that try to promote simple, green ways of living. These may be virtuous, however, in terms of minimizing environmental impact I think they’re somewhat irrelevant.² Saving plastic bags won’t lift a toe on our carbon footprint unless we find ways to either cut down on air and automobile travel, or do it more efficiently. And on carbon footprints — global warming is, I think, a red herring — there are thousands of nasty effects of pollution from, say, coal-fired power plants that will hit even if global warming doesn’t occur (though I think, probably, it will). Too much of China now wears breathing masks.

I can see myself joining and dedicating myself to the right company, full-time, with the same goals and philosophies. These are quite stringent requirements. Therefore, I have so far found it necessary to preserve some of my energy from full-time employment, and concentrate what I have left for my own projects. This is why I need to be open with companies: with most, all I am looking for is consulting work, and I want to be completely, totally honest with them about it.

Sincerely,
Danielle

———

Notes:

[1] - From the classic talk by Richard Hamming, You and Your Research. I don’t, particularly, apologize for my ambition here. Why shouldn’t I try to do first class work? The Nobel prize winning part is purely incidental. But this is the kind of work I mean — a significant contribution, one that people can build upon.

[2] - The free, online book
‘Sustainability - Without the Hot Air’ is an excellent read into environmental matters. And I shouldn’t claim that small contributions to green living is completely irrelevant — it does have some small effect, and perhaps raising the issue in our collective consciousness will have a greater second order effect than I imagine. But so many people, for example, go out of their way to buy ’sustainable’ products at Whole Foods, say, when in reality, longer vehicle trips do more damage than almost anything you could buy. Most of these things are just marketed in a way to make you feel good about them. They don’t have any real effect!

———

Thanks to Alex Lang, Ma’ayan Bresler, Nick Pilon, Colin Percival, Michael Nielsen, and Joel Muzzerall for reading drafts of this, and Charles Beatty, for sparking it.

PS: Certain misconceptions have been raised. Some feel that this is one demand of an over privileged generation. I reply to this here. Additionally, I am not, in fact, abandoning my startup. But I do need money, and a visa, so I am looking into either employment or seed funding.

Cosmology in Ten Minutes

Recently, unusual features of the cosmic microwave background, a ’snapshot’ of the early universe, have raised issues with our understanding of the Big Bang. A Caltech team has shown how we might fix our theories. They suggest that there might have been an asymmetry in the energy that once powered the big bang. If this is correct, anomalies in the CMB may be traces of structure from a time before our explosive beginnings.

True to form, when a discussion appeared on Hacker News I rushed to comment, and this article erupted from that attempt. The current scientific understanding of our cosmic origins is a mystery to the public at large, but it was only after I noticed the bewilderment of my fellow hackers that I realized how poor a job we scientists have done in conveying the motivation behind our discoveries.

This article represents an attempt to replace that sense of bewilderment with that of wonder. I want more than to explain what cosmologists believe. I want give people a deep sense of why we believe it, of how we’ve come to our current understanding, and of why we care.

Look close, and it seems the universe is lopsided.

The cosmic microwave background (CMB) is like a snapshot of the early universe. It was once all hot plasma, gas so hot that the atoms inside it were broken up. Because it was hot, it emitted light. Because it was dense, it was opaque: the light emitted couldn’t just pass through, instead it had to bounce around. But once cool enough, the universe became transparent: all the light could now travel freely. It was as if the photographic shutter of the universe was lifted.

The light from this moment became the cosmic microwave background radiation. Because the universe seemed to have cooled at almost exactly the same time everywhere, the CMB is, unlike almost everything else in astronomy¹, a picture of the entire universe at almost exactly the same moment in time. It is the best picture we have of the structure of the early universe.

The universe appears to have expanded evenly since then. We know it’s expanding now. Light is like a wave. Since the speed of light is constant, an illuminated object moving towards us has its wave crests squish together, turning bluer, and an object moving away from us has the distance between crests expand, turning more red. This is called a red shift. Since he knew the colors of certain celestial objects, Edwin Hubble was able to observe that the further something is from us, the more red-shifted its light, and therefore the faster it is speeding away.

Since we know that the early universe was hot, dense and small, and we know now that it’s cooler, sparse, big, and expanding, we can reasonably deduce that, long ago, there was a Big Bang. The universe exploded.

Strikingly, the CMB is almost the same everywhere you look. There are minor fluctuations, but even they seem to have the same distribution everywhere. The CMB, our best picture of the early universe, is extraordinarily smooth. It is one of the smoothest things ever observed in nature. This might not seem like a mystery. You might imagine that anything expanding, hot and dense would look roughly the same in all directions. It needn’t. Nebulae are formed by exploding stars, and they aren’t particularly smooth. In fact, in nature, it would seem, more often than not, that explosions are messy.

In 1981, Alan Guth suggested what might be called a ‘recipe for a universe’: inflation theory. Until then, nobody had come up with any good ideas for why the universe was so smooth and even. It is as if God² had pressed the entire universe with a cosmic clothes iron.

Guth said, suppose you started with pretty much any initial universe. Suppose you also had an extremely strong, extremely smooth field of energy. If this field started dumping energy into the rest of the universe, it would also, evenly expand space itself. The universe would undergo a period of exponential expansion — inflation — having the effect of flattening and smoothing the rest of the universe. Inflation is God’s clothing iron.

A flat, smooth universe isn’t the only thing that inflation predicted. For example, at small physical scales, quantum mechanical fluctuations persist. During inflation these fluctuations are blown up as well, and these would seed, almost entirely, the cosmological structure of the universe. We see these fluctuations in the CMB. According to inflation, they are tiny quantum fluctuations blown up to a cosmic scale. They are, quite literally, the ancestors of our galaxies.

It wasn’t just that there were fluctuations. Inflation theory predicted a very specific distribution and type³. When people finally had the technological capability to check, that’s just what they found. The universe appeared, at a cosmic scale, astonishingly consistent with this simple theory. Yet recently our observational capacities have improved. A CMB survey called WMAP has uncovered several surprising and unexplained features, not all of which fit well with the our previous inflation theories.

If you divided the sky in half by tracing the orbit of the earth around the sun⁴, and compared, in each half, the size of big fluctuations, those between 3 and 5 degrees wide, you would come to the conclusion that one side has fluctuations outweighing the other by an alarmingly large amount. One side of the universe is bumpier than the other. Moreover, the difference is larger than would be accounted for by randomness, at least 19 times out of 20.⁵

This asymmetry looks real. It has been checked against every known experimental error and background effect astrophysicists have been able to think of. And if it is real, our previous inflation theories, with one field of energy to inflate the early universe, won’t work. They can’t account for this anomaly.

The authors Erickcek, Kamionkowski, and Carroll don’t merely point out this problem. They posit a solution. They describe another inflation model, consistent with our new observations. They suggest the universe had not one, but many fields of universe inflating energy. There’s just one problem. At least one of these fields needs to be asymmetric.

Where could such an asymmetry come from? It is possible that we’ll never know. Cosmology offers us the hope of uncovering consistent, compelling stories of our origins. Thousands of independent observations fit neatly in cosmology’s book. But while we may discover a few lost pages from our first chapters, we may never know all reasons why our book was written in the first place.⁶

Nevertheless, the authors make an exciting point. Wherever the asymmetry in the inflation field came from, it must have existed before inflation. It must have existed before the big bang. We had once imagined that time before our explosive beginnings would forever remain a mystery. Yet hidden in the CMB are hints of times earlier still. In this wonderful piece of work, the authors carefully consider what anomalies in the CMB could mean. And in the process, they may have discovered a way to look farther into the past than ever before.

Notes:

[1] - Since light moves at a finite speed, when we see something far away, we’re seeing light emitted in the past. What we see of something a light-year away is (at least) one year old.

[2] - I mean ‘God’ here as in a figure of speech. Feel free to substitute ‘Mother Nature’, ‘Allah’, or the ‘Flying Spaghetti Monster’ while reading.

[3] - The quantum fluctuations predicted by inflation follow a nearly-scale-invariant random Gaussian distribution. These fluctuations show up in the CMB, and for the most part follow these predictions pretty closely.

[4] - The line dividing the two halves of the sky here is called the ecliptic.

[5] - Formally, this statement is true at at least the 95% confidence level.

[6] - There are some questions forever beyond our grasp. Even if we knew from where the Big Bang had come, we could always probe further, and ask where that came from.

Advice to the Bright and Young

An article on one bright young man, Moshe, recently appeared on Hacker News. For a long time I’ve been meaning to write about the subject, and what was to be a simple comment morphed into this essay.

The story of educational acceleration is an old one. Curious, bright children learn and explore rapidly on their own, and interactively with their parents. The world is like a playground for the growing mind. The child takes in everything. Eventually, these children find themselves mired in school’s morass. There are new adventures: more kids, older kids, a new environment. Yet kept in one place, individual attention of parents replaced by lectures from often overtaxed and uninterested teachers, their minds are left to go fallow. While some of school is new, and quite enjoyable, boredom and obedience, for the curious child, is torturous, a fact which lucky children and mindful parents come to confront.

Alternatives appear: skipping grades, dropping out, home-schooling, gifted programs, science fairs, participating in the popularity game, sports, focusing on musical or athletic achievement, playing hookie, becoming jaded.

After entering junior high I pretty much stopped responding to the world at large. Life rapidly degenerated. I quickly dropped out, and luckily my parents didn’t make me go back. At that time both of my parents were very busy with work, and so homeschooling couldn’t work for long. We discovered that college was much cheaper than private school, which didn’t seem very good anyway. We argued my way in.

One Response to Rejection

Those who’ve spent time with me over the past few months know both how absorbed I’ve been in the catalysis of our startup, and how poor I am at concealing my admiration for YCombinator. We had poured startling effort into building our product, honing our idea, refining our pitch. But our focus was, perhaps embarrassingly, almost entirely toward a single goal. Getting into YCombinator. It was constantly in our minds. Ample encouragement followed months of work. On occasion, I could be found exclaiming my certainty to the universe. We’ll get in. We’ll make sure of it.

The letter arrived silently. I’m embarrassed to admit that my body read like a chapter on the stages of grief. Shock. My stomach churned as I turned inward. Admonishments ‘not to take rejection personally’ meekly confronted universal doubts. Egos struggled against a rethinking of everything. Hours of discussion lapsed. Plans of what to do were floated, accepted, rejected, forgotten. Night passed to sunrise before sleep. Denial. I woke up recalling a story of one rejection mailed out accidentally, to a startup later to succeed. Thoughts strayed from their success - all I could register was the possibility of a mix-up. Anger. I stewed. ‘It doesn’t matter what they think. I know where they’re coming from, I know what they must think of us. They’re wrong! And we don’t need them anyway.’

Finally, acceptance.

On Naming Startups (with Ruby)

The name problem has been with our band of hackers for a while. At least we were not alone: judging by the perennial popularity of the topic on Hacker News, it would seem to stump many.

On such matters, an appeal to a higher power is appropriate. My friends use a variety of divination techniques, such as flipping a coin, tarot, or peyote. I, however, found myself reading an infrequently referenced blog post by Paul Graham (an orphan of the collapse of infogami). Sorry, I can’t make a direct link, scroll down to “Startup Names”.

Some choice excerpts:

“…as happened with lofts, the features that initially repelled people, like rough concrete walls, have now become a badge of coolness. Weird names are now cool, if they’re the right kind of weird. Nothing could be less cool, at this point, than calling a startup “cool.com.” A company with a name like that could not have arisen organically. “Cool.com” smells of a media conglomerate trying to create a web spinoff.”

and

“My favorite recent startup name is probably Writely. It looks so natural that even though it isn’t a word, you feel it should be. Anyone thoughtful enough to come up with a name like that is probably going to have good software.”

Even ordinary people have an extraordinary ability to glark meaning for a word newly encountered. A word that feels natural enough to exist in speech (’I'm feeling ever so writely’) is quite a goal to aim for. People are sure to remember that.

I threw together a ruby script to create domain names from some simple rules and then check whois. I multithreaded it for throughput. (Ruby threads are easily invoked but apparently the threading system is not so powerful.)

Blue Eyed Islanders. (A Logic Puzzle)

Terry Tao’s recent post on a classic logical puzzle has seeded a bloom of activity in the nerdsphere. A friend of mine introduced it to me over mugs of steamed milk in the graduate college coffee house; I was telling him of recent work I’d been doing on the soundness of emergence of Nash equilibria, and where sufficient conditions arise in real life.¹ It was an enjoyable conversation, though I wasn’t afforded the luxury of working it out for myself. If you’d like to struggle through the problem on your own, read Terry’s post, and only return after the break when you think you’ve solved it. (why am I writing this? It started as a comment on news.ycombinator.com, and I couldn’t help myself)

This problem is subtle, and wording is important, so I’ve reproduced the statement from Terry’s blog (emphasis mine) [editor's note: Terry didn't intend for a particular subtlety, and so has reworded his main post. He says that it had an 'unexpectedly interesting subtlety in its formulation, but was not the puzzle I had actually intended to write'. He's posted the original here]:

There is an island upon which a tribe resides. The tribe consists of 1000 people, 100 of which are blue-eyed and 900 of which are brown-eyed. Yet, their religion forbids them to know their own eye color, or even to discuss the topic; thus, each resident can (and does) see the eye colors of all other residents, but has no way of discovering his or her own (there are no reflective surfaces). If a tribesperson does discover his or her own eye color, then their religion compels them to commit ritual suicide at noon the following day in the village square for all to witness. All the tribespeople are highly logical and highly devout, and they all know that each other is also highly logical and highly devout.

One day, a blue-eyed foreigner visits to the island and wins the complete trust of the tribe.

One evening, he addresses the entire tribe to thank them for their hospitality.

However, not knowing the customs, the foreigner makes the mistake of mentioning eye color in his address, remarking “how unusual it is to see another blue-eyed person like myself in this region of the world”.

What effect, if anything, does this faux pas have on the tribe?

Incompleteness and Halting. Gödel and Turing.

The following occurred to me on a run about two years ago:

It’s not given much press, but the the Halting Problem is intimately related to Gödel’s First Incompleteness Theorem. Indeed it produces it as a correllary. Historically, Gödel’s incompleteness results were proved by hacking arithmetic into a Turing complete system, and this is still how they’re explained today.

There’s a one-to-one bijection between computability of a function and provability of a statement. Hence, the short, and generally accessible proof that the Halting problem is not in general computable for an arbitrary input is also a proof of the ‘most important, surprising result in logic’, namely, that some results, which have may have a perfectly valid truth-value outside a system, cannot be proven within it. One only needs the notion of a computer to follow this line of thinking, which is, in essence, what Gödel did. But the Halting problem is much easier to grasp. I’ve had children understand it, though it does take some walking through!

The interesting thing about the Halting problem is that it’s unsolvable in full generality, independent of whatever special capabilities the system has available. To see this clearly, consider the proof.

Question: Does there exist a (halting) program H which, given any program P, figure out if it would halt, for any input I?

Assume there exists such a program H. Construct a program T as follows.

(Program P, Input I) => (Boolean Halts):
if H(P,I) is true run forever
otherwise halt

Now, call T on itself, with itself as an input. Our assumption presupposes that H always halts. If T would halt on input T, then T will run forever. And if T would run forever on input T, then T would halt. This is a contradiction, so no such program H would exist.

On Outliers: What they represent, and why the Central Limit Theorem is Typically Off.

The central limit theorem states that if you have many small, independent, random variables, then their sum is distributed approximately as a bell curve. Strikingly, almost everything is made up of many small parts, and these parts don’t tend to influence each other very much.

So much of what can measure seems to fit a bell curve. This is why the normal distribution works. Because this assumption tends to work well, it is usually taken as a matter of course. Students are taught it, lecturers preach it, researchers apply it, and startlingly few stop to question it.

Suppose the variables are not small, or suppose they’re not independent. Suppose, under certain conditions, the value of one variable would seriously effect another. Suppose we’re talking about the buildup of snow on a mountain slope. Most of the time, snowflakes can gradually build, without significant effect. But once enough builds, you don’t find snowflakes resting calmly upon a drift. What you find is an avalanche.

The sum total of snowflake movement isn’t what we might expect. The snowflakes on the top used to be lightly packed by the new, gradually coming down. The snowflakes on the bottom used to just sit there. But they’re not just sitting there. They’re moving fast, and they’re moving down.

General Problem Solving Strategies (for programmers)

Konstantin Lopyrev writes (on the TopCoder forums):

“Everyone always discusses algorithms and such on this forum. However, there is one trait that is perhaps equally important as knowledge of algorithms - good problem solving strategy. By that I mean not jumping into solutions too quickly, thinking through everything before writing any code and such. Can anyone share what they’ve figured out that helps them? I’ve noticed that I have several flaws when I solve problems in general. I am too quick to start writing code. Also, I don’t think through my algorithms thoroughly. Also, there are many other things. Does anyone have similar flaws that they’ve already passed. If you have, please share your strategies for getting better at problem solving in general. Anything would be helpful, since I have the TCHS tournament coming up and I want to go home with some sort of prize.”

I wrote back, on May 10th ‘07:

One thing I want to start doing is to make estimates of the implementation time of parts of my solution, how difficult each will be to code correctly/debug, and how sure I am that my method will work in the first place.

I’ve gotten very good at making estimates of time and space constraints. Now they are second nature. But knowing how best to solve a problem practically is still another area which I need to improve.

For example, for the last hard problem on TopCoder (circa May 9th, sorry, you’ll have to register to see it). I broke the problem down. Six steps toward victory! Five of those steps were really quite simple, though a few bugs crept in. The sixth was brutal — I could solve the problem by implementing a complicated comparison function object which recursively called itself on sorted sets of children. It was really crazy. To get it all working properly took nearly eight hours after the match.

Each bug, if not isolated, tends to increase debugging time by something like the square of the number of interacting components: the rate at which the number of dependencies would grow. Perhaps it would be simpler to either independently verify each component (by having pre-written tested functions or methods), or to organize the program, inasmuch as possible, as a linear pipeline or other simple acyclic graph. In fact, one might take a few seconds just to sketch the program idea out — how strongly coupled is it? What can I be sure of? How much of a pain will it be to debug? Can I make it less so? If not, can I anticipate debugging requirements and build in information gathering tools?

Quantum Field… Finance?

One morning around the graduate college dining hall, there was a gathering of physicists, finance students, and economists. The physicists are always quite amazed by those people who decide to forgo the life of the ivory tower, and choose to strike out into the real world, and so could not be kept from asking what the economists actually did. Furthermore, we could not be kept from wondering aloud what type of mathematical models they built and polished, and whether any of them had a physical interpretation.

One of the economists scratched his head, drew a sip of black coffee from his porcelain cup, and mumbled something about how a large proportion of the physics department of Harvard University was hired by a trading company, with the lure of riches beyond the pale of the meager imaginings of the physicists (”you mean I can afford a house?!”).

Third Places

Tonight it’s winter in Berkeley. 53 degrees and raining, and outdoors, warmed by a heat-lamp, sheltered by an awning. I draw spiced apple cider through my lips. Classical music plays. An earbudded minority vote silently with their ears. Old men watch hooded students roll down the hills towards Telegraph Ave, Berkeley’s epicenter of hippiedom. Moist, newspapers ink the hands of activists, busily plotting the victories in the years long struggle to ’save the oaks’. A young man lids a drink and smiles at me. Separated by glass, headphones, and 12 feet, I smile back. We wave.

There’s something magical about this place.

I don’t know anyone here. To arrive I flew four thousand miles from my place of growth. This place isn’t home. Yet there are few places that attract me so strongly. Modern life has been made private. And in doing so, life’s become a little lonely.

Builders of great cities have long understood that life would, but for misfortune, consist of more than work and one’s home. The vibrancy, energy, and community grown in what are sometimes called ‘third places’ played part in much of the world’s social, political and intellectual revolutions. The roles that the Roman forae, French salons, and English learned societies played in scholarship has been tremendous, as has been the influence of American chautauquas, worker’s taverns, and artist’s ghettos in social and political spheres. These public, accessible, talkative, comfortable playful places are magnets for folks of many stripes. Creativity can thrive there. Unconstrained by work’s implied unity of purpose, and decoupled from the tight bonds around one’s family and home, third places give marginal people, ideas, and voices room to grow, people to hear them, perspectives to challenge them, and food to help keep the conversation going.