srconstantin

I have tried umpteen billion times to install Apache. I have wasted several full days trying and failing. I have a Mac. I must be retarded.

I didn't have social trouble in high school, I had trouble in college. And that's something that I think goes unrecognized. Everybody knows about the crazy suburban high school experience, but a surprising number of my friends were severely unhappy in college. I can barely remember my freshman year -- I pretty much blocked it out. And I went to an Ivy League school. There was a kind of background noise level of callousness; people who were "different" in various ways were singled out for mockery in the school paper, people who were struggling were ostracized. And the undergrad culture was seriously misogynistic -- I accepted it as a given while I was there, but by the time I graduated I noticed I was advising girls not to go there if they had other options.

It's not just high school that can suck. Teenagers are immature, but there are other kinds of environments that can be stressful and isolating. And the advice is the same -- get through it, and then find a culture where you fit in better.

I don't believe deferred altruism is as good as it's cracked up to be, because of the importance of motivation. If you tell someone who wants to work in a homeless shelter, "Go to Walk Street instead and give away your wealth" you may be correct in theory, but in practice that person might be a terrible banker and burn out quickly. To desire money instinctively, the way most people crave hugs and candy, is actually rather rare, and most of the rich people I know desire money that way, while most of the poor people don't. I'm not sure an altruist can make himself obsessed with money in order to be a better altruist. I've seen people try and fail.

I don't trust people who start businesses and say they're not interested in money. Money is one of your best gauges of whether you're doing things right. Not being interested in money when you're working on a for-profit enterprise makes me worry you're either unserious or disingenuous. That said, hardly anybody only cares about money. We all trade off money against other things -- free time, independence, and so on. I care about money, but I'd rather make a little less working on a project that'll genuinely make the world better than a little more working on something unimportant.

Idea #1) Safemap Pull crime blotter data, and deliver the least dangerous route between two locations. Offer a "danger score" for locations.

Idea #2) Health Tracker with Correlations There are a lot of self-tracking apps out there; few to none of them do what the actual point of a self-tracking app is, which is to identify correlations between what you're tracking. (e.g. are you unhappier when you get less sleep?)

Idea #3) International Video Pen Pals Practice learning the language of your choice by being paired with a native speaker for video chat.

I'm skeptical.

All of the articles cited are indeed about civil liberties, but I don't see them as being problems that can necessarily be improved by developing better cryptography.

The social network propaganda isn't so much of a technological problem as a social one -- people assume that everyone's using social networks in a straightforward way, to express themselves and connect with friends, even though they have the capacity to be used in many other ways, including deceptively.

The NSA email-hoarding thing is also a social problem -- encryption exists, it's just that nobody uses it.

The "pre-crime screening" appears to be a statistical model. I don't yet know if there's a science of counter-machine-learning -- that is, creating examples that are hard for models to classify, analogously to making harder and harder CAPTCHAs. If government gets better at data science, what civil libertarians will need is not better encriptions, but different techniques designed to protect against statistical intrusions.

Likewise, I don't think better crypto is in any straightforward way a protection against biometrics or government bills to take down certain websites.

If he's saying "we need more libertarians and people concerned about civil liberties," then he's right. But I'm not even sure there are fewer than there used to be (look at the backlash to SOPA). I just suspect think there was a time when people had more hope in purely technological fixes to government abuses. I know some old-time crypto-anarchists, and they're largely disappointed with how things have turned out. In the end, preventing the abuse of government power is a hard problem, and it looks less likely that a small set of tech breakthroughs will suddenly make civil liberties violations obsolete.

I actually miss old-fashioned web design. These old pages look like text, and not much else. But most of what I want to do online is text; I want to read and write. The simplicity is actually an aid to clarity.

This is very informative.

The only thing I'd quibble with is the claim that teams should not be diverse. I think, unless everyone is a generalist, there's no way to start a company without some diversity of skills. The other day, another mathematician told me "We should start a company together," and it sounded like a terrible idea. A mathematician needs to work with a great web developer, or a finance expert, or some other kind of complement.

One,Two, Three, Infinity is probably the best possible introduction to math for kids.

The big challenge is skipping ahead to the "cool stuff" for kids who are still young enough to lack facility with algebra. Most attempts to do this are either hand-waving or unfairly advanced. But the real solution is to show them genuine math that doesn't require any solving for x, like Cantor's diagonalization lemma.

I do this. Every now and then I decide I'm "claiming the day in the name of fun" and just work on an interesting coding project. Having unofficial projects is necessary to survival.

I disagree. Strongly.

Following your dreams, not working too much, spending time with family and friends, saying what you think, and being happy, might belong at the top of Paul Graham's todo list. He's already succeeded at what most people struggle at: making a living. He may have forgotten that getting by is hard.

What's so wrong about being a "cog in a machine"? Large organizations exist partly because scale and specialization are useful. People aren't "cogs" out of perversity, but practicality. Sometimes your comparative advantage is in being a good cog. And sometimes being a good cog is admirable. Example: surgeons who specialize in one and only one procedure, doing it exactly the same way every time, have a low failure rate. They save more lives that way. Are they creative? Nope. They follow somebody else's rules. They make themselves as mechanical, as repetitive, as possible. And the guy on the operating table lives.

Paul Graham is in a very unusual profession, where risk-taking is much more of an ideal than it is anywhere else. Also notice that he makes money off of telling other people to take risks. The more people decide to be entrepreneurs, the wider a pool he has to draw from. "Follow your dreams" is a useful meme for people like that; if a hundred people follow their dreams and fail and ruin their lives, PG never hears about it, but if one person follows his dreams and becomes a billionaire, PG might get a share. This isn't meant to vilify the guy; there's nothing wrong with making a living by inspiring people. And I'm sure he would say that he's fighting dangerous opposing memes against ever taking risks. I just think you can't take posts like this at face value.

Who do you think makes those tools?

With respect to CS in particular, I think it's different from all other school subjects in that it is and has always remained a predominantly self-taught subject. You don't really get "taught" to program.

(I think this may be part of the reason that while there are increasing proportions of women in all the other sciences, the proportion of women in CS is dropping. Girls are statistically better in all school subjects than boys, and tend to be more diligent at homework and so on. Math and science fall neatly into that paradigm; programming doesn't. You get no gold stars from grownups by programming. So a lot of people who have the problem-solving aptitude to be good programmers never actually get started -- because nobody TOLD them to.)

Undergrads do have serious information asymmetry problems. My best guess as to the reason is that they don't have a lot of interaction with a wide range of people in the working world. They don't really have any idea what it'll be like to be adults. Incidentally, I think this is a big part of the reason undergrads at good schools gravitate to such a small number of employers (investment banks, consulting firms, and famous programs like Teach for America). Kids simply don't know what's out there, and so they reach for known quantities. And one consequence of that ignorance is that they don't have a very clear idea of which skills are most marketable.

I'm not sure if this article is saying people with more CS theory and math do better than people with more work experience, or if it's saying it's hard for everyone to get hired.

Web apps are cheap to build, easy to learn how to build, fall into a predictable model for investors, and are equipped with lots of social institutions (like YC) encouraging people to build web startups.

So, all things being equal, we should expect more people to be creating web apps (compared to other kinds of technology) than their profitability or social value would warrant.

Framing makes a big difference. If there were an established culture and set of resources for engineering or biomedical startups, they might seem less daunting.

I know! I love it as a place to shop, would never try to make money as a seller. But if they wanted to change that, they could -- run calculations for sellers as to how much they should charge to be profitable, all the stuff that the average seller wouldn't think about.

Etsy is a form of "micro-work" but it's never been particularly good financially for people who sell stuff on Etsy. I wonder what they can do for connecting sellers to each other to pool supply costs and identify each seller's likely consumers.

Basically I agree that "when in doubt, learn science." In American society, we have more of the norm of "when in doubt, don't learn any science." And that's counterproductive.

In other countries, there are explicit and implicit reminders that science students are smarter than non-science students. In my dad's (European) high school, students were literally tracked from A to F; A was math, B was physics, biology and social sciences were somewhere in the middle, and vocational school was at the bottom. Harsh? Yeah. But it meant that generically smart and ambitious kids wound up in the hard sciences by default.

I'm not sure that it would be bad if the average size/age of businesses decreased. It seems to me that in larger businesses, while you do get advantages of scale, you also have problems related to size (bureaucracy, knowledge problems, etc.)

I'm so relieved to hear this from someone else.

About two years ago I read an article by Peter Thiel and it was kind of my "ask what you can do for your country" moment. That was when I realized that I didn't want to be an academic, that I believed people who built technology in the private sector were actually doing more good in the world. So I thought "I want to be an entrepreneur."

I've had to scale back a little as I did a couple of start-ups and learned how ignorant I was. I'm not ready to be CEO of anything. I'm not even sure I want to be; I want to write algorithms more than I want to manage a company. Honestly, right now, I want to work with people smarter than I am, and get better at what I do.

I'm done with the notion that you should always start a company right away in your early twenties, that you shouldn't have to take any time to pay your dues or build mastery. I don't want to work at a silly, unsuccessful start-up just for the sake of "being an entrepreneur." I want to get as good as I can at what I do. I want to work somewhere where I'm not the only one who's heard of an inner product. I want to do something substantive, and I'd rather do something substantive at 35 than something crappy at 23.

Link to LaSVM paper: jmlr.csail.mit.edu/papers/volume6/bordes05a/bordes05a.pdf Also a good overview of SVM techniques in general.

So...are you saying you need the dual formulation in order to allow a kernel model?

I'm not sure you couldn't solve this as an optical character recognition problem, with something like Isomap, which is invariant under orientation and scale changes. Break your image up into patches, reduce dimensionality on all of them, and find the location that way. I would predict that this is faster, at some cost in accuracy.

A lot of this is stuff I've been thinking implicitly for a while.

Success can exist on a continuum between competitive and niche.

The far end of "competitive" is being something like a concert violinist. Everyone knows that it's prestigious to be a concert violinist, millions of people want to, everyone plays the same audition pieces; you win if you're a hair's-breadth better than a million people doing the same thing. It's an unpleasant process, and the odds of success are long, but you know for certain what you'll have to do to win, and you know that if you get that coveted orchestra spot, you'll have reached the top of your profession.

The far end of "niche" is probably starting a company in a completely open industry. Nobody else is doing what you're doing; you're not competing directly with anyone; but you don't know what you'd have to do to succeed, you don't know if the niche you've chosen is any good, and you don't know how to judge how valuable it is to succeed in your niche. On the other hand, you don't have the demoralizing stress of trying to do the same thing as everybody else, infinitesimally better.

I've been trying to decide what to do with my life in this context. I'm a grad student in mathematics, with research interests related to machine learning. On the one end of the scale there's academia (which is at the extreme "competitive" side of the axis) and on the other side there's entrepreneurship (at the extreme "niche" side of the axis) and in between there's working as a data scientist at companies ranging from "three-month-old start-up" to "IBM." There's a tradeoff between the freedom of staking out your own territory, and the risk that if you're free to make things up, you'll get everything wrong.

(I've noticed that this is also true within science. In my computational vision research, there's a lot of "whitespace" for developing your own mathematical models; but most people, sooner or later, get proven wrong. In my harmonic analysis research there's a lot of classical theory already existing and the problems are better-defined; but it's correspondingly harder to make an original contribution.)

Where do people selling consumer electronics on Kickstarter find manufacturers and supplies? (My first guess would have been Alibaba, but is that feasible for someone really starting from zero?)

I agree that difficulty is a good lesson; but it doesn't seem to be working for a lot of people in the US. Many students give up on math early. That might not be economically sustainable. I don't think math education is optimized to make the most of intuition and motivation, and I really don't think experimenting with improvements is harmful.

I agree; the whole thing could be done with pictures rather than numbers. Lay out an actual grid, with objects -- seven ladybugs in the third row, fourth column, etc. When you subtract one row from another, you physically drag the ladybugs from one row over the other row, and when they 'match' they disappear. You can tap into physical intuition.