Today’s Fad

Posted on by David Wright

Here’s a quote from HN in response to the question: what’s the big deal with Machine Learning?

There’s this enormous focus on ‘web scale’ technologies. This focus necessarily invokes visualizing and making sense of terabytes and eventually even petabytes of data; conventional approaches would take thousands or millions of man hours to accomplish the same level of analysis that computers can perform in hours or days.

I totally agree. I’ve joined a few technology meetup groups here in NY and so far I’ve had interesting reactions to my field of expertise. I basically say my job is predictive models, but on small-mid-sized datasets. Cue disappointment.

Everyone is focused on predictive models that crunch BIG DATA. I’m taking a course on ML but I don’t do BIG DATA.

There are two kinds of big, you see. You can have a list of a billion addresses, but that doesn’t really qualify as BIG big.  People can get their heads around what to do with a billion addresses: what regressions to run, what information can be reasonably gleaned from analyzing it.

BIG big is different. BIG refers to a HUGE number of parameters that might or might not be meaningful. Think about some problems that are common applications of ML:

For highly dimensional problems, such as text classification (i.e., spam detection) or image classification (i.e., facial detection), it’s almost impossible to hard code an algorithm to accomplish its goal without using machine learning. It’s much easier to use a binary spam/not spam or face/not face labeling system that, given the attributes of the example, can learn which attributes beget that specific label. In other words, it’s much easier for a learning system to determine what variables are important in the ultimate classification than trying to model the “true” function that gives rise to the labeling.

BIG means you don’t really know what you should do with the data. You kinda know what answer you want, but you can’t really hold a thousand or ten thousand different parameters in your head long enough to specify some kind of regression.

Now think about technology trends today. Computing power, bandwidth and memory capacity are all now cheap enough that computers can handle BIG better than humans can. THAT’s interesting.

In my professional life (sadly or happily, depending on how fired up I am to do ML), I don’t tend to get that many parameters.  Insurance is rated on a very few rock solid inputs and the rest is just sniffing out when someone is trying to screw you over.

But I’m intrigued, nonetheless. Woe to the ambitious one who doesn’t keep a tab on the cutting edge.

Here’s a link to yet another ML class.

There Is No Middle

Posted on by David Wright

The Economist laments:

Imagine a presidential candidate next year who spelled out the need for deep future cuts in spending on entitlements and defence, as well as the need to raise some revenue (largely by getting rid of deductions); who explained that the pain would be applied only after the recovery was solidly in place; who avoided class or culture wars; who discussed school reform without fear of the Democrats’ paymasters in the teachers’ unions. Better still, imagine a new centrist block in Congress, which might give that candidate (or for that matter a President Obama or Romney) something to work with in 2013.

The political system is too open to so cleanly flip the bird to enfranchised minorities like teachers unions, farmers and everyone within earshot of the Medicare dinner bell.

THE problem today is that we aren’t as wealthy as we thought we were (and elect politicians whose job it is to preserve that illusion). There is a lot of debt that needs to go bad, which means gigantic capital losses. And that capital doesn’t ‘belong’ to bankers, folks.

It belongs to you. It’s your savings. It’s your pension. It’s your home equity.

It’s all well and good to idolize (in hindsight) historical figures that make good decisions in tough circumstances, but we quickly forget the ones that flub their moment of truth. There’s a reason the’re called “tough” decisions. It’s because you’re picking losers.

No politician in his right mind is going to tell voters: I’m going to destroy some of your wealth and keep destroying more until the system starts working again. I’ll let you know when I’m done.  Gosh, I hope I figure it out in time to stop before I go too far. Yes, I’m doing this when unemployment is 9%.

But that’s what has to happen. The most painless way is to do it with a burst of inflation. The most painful way is austerity. But it has to happen.

It’s no surprise WW2 happened after the last such episode.

Science Majors Change Their Minds Because They Run Out of Time

Posted on by David Wright

MR comments are pouring out.

Here’s the NYT.

And there are encouraging signs, with surveys showing the number of college freshmen interested in majoring in a STEM field on the rise.

But, it turns out, middle and high school students are having most of the fun, building their erector sets and dropping eggs into water to test the first law of motion. The excitement quickly fades as students brush up against the reality of what David E. Goldberg, an emeritus engineering professor, calls “the math-science death march.” Freshmen in college wade through a blizzard of calculus, physics and chemistry in lecture halls with hundreds of other students. And then many wash out.

I hope Robin Hanson chimes in on this. I bet he’d say that Universities award status, not education, so having high drop-out rates is what they’re selling.

If they were selling education, there would be many many more options for learning this material. And options for learning it later in life. Or over a longer period in life. My hypothesis is that the kids in University have too many other things they’d rather do then cram, which is distinctly unpleasant. So they don’t, and when they’re made to feel stupid, they crash out.

Milton Friedman, Ahead of His Time At 86?

Posted on by David Wright

Here’s Milton Friedman (via MR)

Friedman’s absence is mourned today because he was one of these awesome combinations of experience, accomplishment, intellect and brilliant communication.

The last few questions in the Q&A can probably be summarized as a prescient combination of all the sensible present-day views on the 08-12+ crises:

  • The Euro is Effed
  • Japan has had tight money for a long time
  • We know that about Japan because they have low interest rates, which are often a sign of recently tight money, except when they’re not.
  • By the way interest rates are a terrible ‘tool’ for managing the economy.
  • nominal aggregates should be monitored/targeted.

What strikes me most about Friedman’s analysis is his experience. He had lived through and studied just about every conceivable macroeconomic situation. What understanding he must have had.

He does make one point that he later appears to contradict (big quotes because, again, the dude was just so lucid):

Now, my preference, of course, would be to abolish the central bank altogether and to simply have a computer that would churn out—well, I have two variants of it. In one of them, I would freeze the amount of the highpowered money and let the market go. In the other, I would assist the market by printing out a specified amount of high-powered money every month or quarter and have a steady rate of monetary growth

Vs

In 1989, the Bank of Japan stepped on the brakes very hard and brought money supply down to negative rates for a while. The stock market broke. The economy went into a recession, and it’s been in a state of quasirecession ever since. Monetary growth has been too low. Now, the Bank of Japan’s argument is, “Oh well, we’ve got the interest rate down to zero; what more can we do?”

It’s very simple. They can buy long-term government securities, and they can keep buying them and providing high-powered money until the highpowered money starts getting the economy in an expansion. What Japan needs is a more expansive domestic monetary policy.

One interpretation of this is that he believes that his rules would have prevented the crisis today. The thing is that this view implies that the crisis was actually caused by tight money. Which is a bit controversial.

How would we have measured that tight money? This touches on his elaboration on the first point:

In recent years, that has not looked as good as it did much earlier, because the actual relationship in the world between monetary growth and inflation in the economy has become much worse in the last 10 or 20 years. But, that’s partly because there’s been so much financial innovation and adaptation and, ultimately, it is the money supply that rules the roost and that will determine what long-term inflation will be.

So, the problem with money supply targeting or, indeed, with any DERIVATIVE rule is that the market will find a way to innovate past your rule. Money, after all, isn’t what we care about. What we care about is GDP! And what we need to target some kind of aggregate, because that’s the only thing we can measure.

Enter the Sumner synthesis.

The Sumner synthesis says that we target what we care about, NGDP. Crucially, we also target the level. This makes up for the problem that the central bank isn’t particularly good at doing anything in two important ways.

First, levels means that in a liquidity trap the bank can wait. And wait. And wait.

Eventually we’re going to get out of the trap and when that happens, KA-BLAMMO, out comes the money.

My second point is related to the first. The central bank isn’t just ineffective in a liquidity trap, the central bank is kinda ineffective all the time.

Think about Friedman’s problem with the shadow banking system and their shadow monetary aggregates that we don’t know how to measure well. The central bank can’t control that money, so the central bank can’t loosen it up or reign it in at will. Let’s say the shadow banking system has taken the central bank from something like 50% effectiveness to 30% effectiveness.

Doesn’t matter with level targeting. Over a long enough time span, you’re going to hit your level right, even with the wobbliest of levers. The promise remains credible.

I Wish I Could Add Something

Posted on by David Wright

To Horce Dediu’s post where he argues that Siri might be the next revolutionary interface. Here are his points in its favor:

There are many things going for it:

  1. It’s not good enough
  2. There are many smart people who are disappointed by it
  3. Competitors are dismissive
  4. It does not need a traditional, expensive smartphone to run but it uses a combination of local and cloud computing to solve the user’s problem.
  5. It is, in a word, asymmetric.

He’s really thrashing his straw man with the first few points. My informal read of the commentariat suggest Siri’s boosters outnumber its haters by an order of magnitude. The 4th, though, is important

But read his whole post. His real point is this:

In 2007 something happened which changed the industry. It took a few years to even realize it was happening but by the time it was obvious, it had changed to such a degree that huge companies found themselves in financial distress.

Revolutions are hard to predict, even when they’re happening. And they’re slow, so don’t go getting all frustrated that Siri’s best case scenario is dominance in 3 years.

New York Tech Meetup Review

Posted on by David Wright

File this one under: things I ‘should’ do since I live in New York.

The evening is comprised of stuffing 700 people into an auditorium to watch a bunch of live tech (software and hardware) demos.

In the most Burroughs-esque way this is junk for innovation junkies.

You feel the pain of demos gone wrong, you cheer awesome implementation of awesome ideas. You can feel a jolt of charisma from some of the young entrepreneurs and you can smell the money backing the most polished products.

The audience is packed with all sorts of folks: computer nerds, unemployed equity analysts, employed equity analysts, friends of presenters, VCs and random enthusiasts like myself.

I sat next to an unemployed equity analyst with a CS background looking for startup ideas for himself. Interesting fellow and good for the odd sensible comment on the presentations.

I’m skipping the after-party tonight, but maybe I’ll make the time when I come back next month.

And come back I will.

A Model of Education

Posted on by David Wright

I’m heavily influenced by my own educational experience, obviously, but I think that the educational standards of the future should take heed of what CFA Charterholders and Actuaries need to do.

You need to do three things to enter into these clubs:

1. Have experience in the profession (vouched for by an insider). 2. Pass a bunch of (very) challenging exams.
3. Have a college degree.

My question is what the hell is the point of #3? All it does is restrict the pool of candidates to those than can afford to go to college.

The only valuable effect of #3 is that it sets an age floor for beginning the apprenticeship. And there should be a minimum age restriction of some form.

There shouldn’t be any 23 year old fully qualified actuaries. They aren’t mature or experienced enough to fulfill the role’s obligations.

Everyone focuses on the exams because they’re difficult, but the experience is much more important. It’s the apprenticeship that matters for performance.

Why can’t we have more independent exams to signal intelligence and drive and, from that pool of exam-takers, select apprentices? Then make the apprentices work for 10 years or something.

It’d be cheap, it’d be effective. It would make society better off.

Don’t Go To College

Posted on by David Wright

Sounds pretty controversial, non?

Here’s Alex Tabarrok:

The sluggish economy is tough on everyone but the students are also learning a hard lesson, going to college is not enough. You also have to study the right subjects. And American students are not studying the fields with the greatest economic potential.

Over the past 25 years the total number of students in college has increased by about 50 percent. But the number of students graduating with degrees in science, technology, engineering and math (the so-called STEM fields) has remained more or less constant. Moreover, many of today’s STEM graduates are foreign born and are taking their knowledge and skills back to their native countries…

More here.

Let’s face it. College has become a party for people who don’t really want to be there.

I feel like some day Colleges will be more like the music industry: the basic product is already free (have a read of the blogosphere or spend some time at the Khan Academy if you don’t believe me) for those who actually want to consume it.

As I commented on MR: what we need is a solid way of signalling domain competence without paying for a degree.

For most people a college degree is a WASTE OF MONEY.

Concave Loss Function

Posted on by David Wright

It’s all downside risk discussions today.

First there’s Corzine:

Izabella Kaminska has the wonky details of MF Global’s repo-to-maturity trade. It’s not easy to follow, but here’s the general gist. MF Global buys a bunch of European debt. The bank’sexplanation of the trade says that the purchases were “entered into repurchase and reverse repurchase transactions to maturity, which are accounted for as sales”. This is the repo-to-maturity trade.

In order to understand what that means, you first need to understand that banks like MF Global used to do nearly all their borrowing on an unsecured basis. But in recent years, that’s changed: nowadays, if you want to borrow billions of dollars for what MF Global calls “client facilitation and principal activities”, then you’re going to need to put up collateral.

So as soon as MF Global bought those bonds, it turned around and pledged them as collateral when it was borrowing money. That’s the repo.

Now here’s the trade: the rate at which it was borrowing money was lower than the coupon payments on the European sovereign bonds. And because this was a “repo-to-maturity”, MF Global was essentially locking in the difference as profit. It got to keep all the coupon payments, while it had to pay out something less than that in interest.

And here’s the money quote from Felix’s source:

Either way, a fall in the value of the bonds could create a major liquidity drain for MF Global. Though these sorts of liquidity risks should have been accounted for in VaR calculations. Much harder to anticipate would have been a complete disappearance of willing counterparties.

Facepalm!

Even better, Europe is EFFED:

Make no mistake about it, the decision to hold a “referendum” is a decision to turn down the deal altogether.

Why, you ask? Well follow the money! Greek banks go bankrupt when the bonds default, Greek pension funds go belly-up when the bonds default.

The Greek gov’t then needs to borrow tons more money to replenish the pension scheme, or maybe toss the old-timers into the street? It’s going to come down to that.

Naw, they’re going to exit the Euro. Gotta be a surprise or it won’t work. The contagion will rip through any other country that doesn’t get a blanket guarantee from the ECB. By contagion rip through, I mean massive shorting of government debt, bank debt and a ha-uuuuge increase in borrowing costs for weaklings, which may well drive them into default themselves.

But we’re not at the end game yet, kids. Perhaps still months away!

On Siri

Posted on by David Wright

Neat article. Everyone who uses Siri suggests that it’s awesome and can do things for you that you don’t feel like doing for yourself. I look forward to using something like this someday.

But I actually find that the simple tasks that Siri will be good at are things that are getting easier and easier all the time, anyway. Maybe finding a restaurant was a pain 20 years ago but how revolutionary is it that she can save me the few seconds it takes to google something?

Most people don’t live in Manhattan where they want to find a new restaurant every week. Most people stay where they live and are perfectly happy about that. Early adopters, worldwide jetsetters and technophiles are a very small minority, folks. Siri remains unproven.

When Siri can find me a new apartment or book me a holiday at my exact optimum combination of cheapness and niceness I’ll be impressed. Those are the things that feel slow and stressful and time consuming to me today.