pageok
pageok
pageok
The Rise of the Machines:

A bit over-the-top, but Richard Dooling's essay in today's NY Times ("The Rise of the Machines") has some ring of truth to it, to my eyes. It blames the financial crisis, in effect, on computers -- or, more precisely, on the networked machine intelligence without which the complex financial instruments now unraveling before our eyes could not exist:

"As the current financial crisis spreads (like a computer virus) on the earth's nervous system (the Internet), it's worth asking if we have somehow managed to colossally outsmart ourselves using computers. After all, the Wall Street titans loved swaps and derivatives because they were totally unregulated by humans. That left nobody but the machines in charge. . . .It was easy enough for us humans to understand a stick or a dollar bill when it was backed by something tangible somewhere, but only computers can understand and derive a correlation structure from observed collateralized debt obligation tranche spreads. Which leads us to the next question: Just how much of the world's financial stability now lies in the "hands" of computerized trading algorithms?"

There's something to it, I think. Nobody knows what to do about this crisis, at least in part because no one has the necessary information about the effects of small changes -- adding capital here, letting banks fail there -- in a complex interlocking network of financial instruments all linked to one another. I don't know if it's a case of the machines taking over (as Dooling suggests), but it's not like other crises we've seen before and will require some real innovative thinking to get it under control.

DiverDan (mail):
In considering this hypothesis, one would do well to remember that, although it was the computers that did the heavy computational lifting, it was people who wrote the programs, people who made the assumptions, people who input the data, and people who ultimately made the decisions (even if certain computers were programmed to automatically execute trades based upon certain decision rules, it was a human decision to let that happen). A computer can tell you that the odds of a fair coin being flipped and coming up heads ten times in a row are less than 8/100ths of one percent, but the computer doesn't make the underlying assumptions (i.e., that the coin is fair, that the tosses are sufficiently high and random in speed of spin so that the outcome isn't predictable by basic physics, etc.), and, importantly, the computer doesn't decide to bet the ranch against a very small, but not non-existent, probability.

The computers were nothing more than tools, and very useful tools. The real problem was human hubris - those very smart investment bankers with advanced mathematics degrees thought they were too smart to fail. Even the pricing mechanisms they used for derivatives like credit default swaps were based on probability models, but they forgot that probability models are only as good as the data and the underlying assumptions used to construct them. You can't accurately estimate the probability distribution of future mortgage defaults based upon historical data if your historical data is based upon one set of lending criteria (i.e., minimum down payments and stringent credit qualification), while the future you are trying to predict is based upon a completely different, much looser set of lending criteria.

I don't know that any innovative thinking is needed to get it under control. I can tell you one way that won't work is to have the government subsidize the highly risky bets by bailing out those who lost lots of money through sheer arrogance.
10.12.2008 2:53pm
LTEC (mail) (www):
1) Everything I've seen suggests that these particular problems are the predictable effects of bad Human decisions, not of unpredictable computer programs.

2) The reason that it is often hard to predict what the effects will be certain decisions that must (perhaps) be made now is that Human society is very complex. Computer programs might help with this problem, or might add to it, but they are not to blame for this complexity.

3) I believe there have been instances in the past where networked computer programs have unpredictably caused financial problems. It might happen in the future, and in a setting where shutting down computers would just make things worse. In this case we will simply reprogram the computers. The system is not Skynet and the unibomber is not John Connor, and I think the comparison is disgusting and dangerous.
10.12.2008 2:59pm
Paul Allen:
Is this man a luddite? (Sorry, just seemed like such an appropriate introduction given David's last post.)

But seriously, I don't think this guy really understands. "After all, the Wall Street titans loved swaps and derivatives because they were totally unregulated by humans." This statement is groundless, as is the next one: "but only computers can understand and derive a correlation structure from observed collateralized debt obligation tranche spreads"

None of that is true. Everything the computer does could be done by hand--laboriously. Most of these computers are running very well understood optimization algorithms. What's changed is that people used to have to do these things by hands using slide-rules, then calculators... but even with a calculator, much of the work was assisted with printed tabulations of statistical distributions, etc. Enter computers.

It became possible to recompute very quickly which made is possible to profit from transient pricing errors. He who finds the pricing mistake first, wins.

It also became possible to compute every more complex things quickly rather than laboriously.

Perhaps what he is reaching for is that only a small number of people understand the mathematics... and these people are almost certainly not the traders, turn bankers, turned CEO, turned members of the board.
10.12.2008 3:08pm
kietharch (mail):
I agree with DiverD and LTEC. Note that in all the "how we're going to fix this" and "we'll do it right from now on" Paulson, Bernanke and assorted economic experts have not said that they would do anything to curb exotic derivatives; it's like they think that is a part of the market economy. Why should this be so?

Paulson and Cox were even beaten back on shorting securities. Where in real life, can you "short" an asset? you can't short the value of an automobile, a house or farm. Yet it has become accepted wisdom that selling things you do not own is a part of the market and contributes to orderly markets.
10.12.2008 3:15pm
Norman Bates (mail):
keitharch

Where in real life, can you "short" an asset? you can't short the value of an automobile, a house or farm. Yet it has become accepted wisdom that selling things you do not own is a part of the market and contributes to orderly markets.
Perhaps you have never heard of commodities trading. In more primitive forms it has been around for several millenia. With the emergence of modern methods of grain mixing and storage in the US circa the mid 19th century it became the rationalized and standard way of dealing with the economic uncertainties of agriculture.
10.12.2008 3:33pm
Hoosier:
I'm with Dwight Schrute: When we have robots with artificial intelligence, make them plug-in. Then linit extension cord lengths to 6'. That way, they can turn against us, but they can't chase us.
10.12.2008 3:43pm
A. Zarkov (mail):
Blaming computers for the (now) worldwide financial crisis is a little like blaming the gun for a murder. While Dooling obviously doesn't understand what he's talking about, in an inadvertent way he's close to an insight. The financial services industry is the biggest user of computers in the country. For example something like 35% of Sun Microsystem's sales are to the finance industry. But computers mostly do mundane things like transmitting trades, reconciling accounts, order flow etc. Their use in derivative modeling is a small part of the overall application of computers in the financial services industry. It's the mathematicians that are responsible for financial derivatives like CDOs, not computers. The WSJ article How a Formula Ignited Market that Burned Some Big Investors discusses how David Li developed the theory behind CDOs. But Li warned his bosses that the theory is perfect and could backfire. Obviously they didn't listen.
10.12.2008 4:28pm
A. Zarkov (mail):
Whoops-- that sentence shoud read But Li warned his bosses that the theory is'nt perfect
10.12.2008 4:30pm
K:
The computers main effect was to permit much faster actions. This meant traders could work very fast and those that did not were laggards. Bosses do not like laggards.

But what if the laggard produced more profit that the speedy? Well, when things were booming and the computer recommendations seemed right the laggard simply couldn't. The speed of computers will now beat any human thought - if the computer program is right. e.g. chess.

So the computers were relied upon more and more and warnings from humans were disregarded. There was no money in caution.

Economists and market analysts at these firms made perhaps $500k and had no authority and brought no profit to the bosses. And who wants to listen to warnings? That isn't fun!

Traders made a million plus - often a very big plus - and that made tens and hundreds of millions for their bosses.
10.12.2008 5:17pm
Elliot123 (mail):
"Where in real life, can you "short" an asset?"

The stock market is real life. To short a stock:
1. Borrow the stock
2. Sell the borrowed stock to a buyer
3. Buy back that stock in the future
4. Return the stock to the lender.

If I owned an Escalade and thought gas prices were going up, I could
1. Sell the Escalade
2. Buy the same model, year, and milegage Escalade later for much less than I realized on the sale.

If I own a house now and am convinced prices will fall even further if Obama wins, I can
1. Sell the house
2. Buy a comparable house later for less than I realized on the sale.
10.12.2008 5:26pm
K:
second comment: DiverDan and others also mention the other problem. faith and hubris.

Economic models for the computer programs were extremely complex, devised by people of the highest prestige such as Nobel winners, and marketed as almost magic - no one believed they were actually magic - but they almost seemed to be.

The terrors of investing losses were gone forever. An infallible aide was being run by the smartest guys in the universe.
10.12.2008 5:27pm
Elliot123 (mail):
"but only computers can understand and derive a correlation structure from observed collateralized debt obligation tranche spreads."

Really? Who wrote the program? Would you say only a computer can understand 876,876 divided by 459?
10.12.2008 5:33pm
Fub:
Elliot123 wrote at 10.12.2008 4:33pm:
Would you say only a computer can understand 876,876 divided by 459?
Doesn't that depend on whether the division is floating point on an old Pentium?
10.12.2008 5:43pm
John Burgess (mail) (www):
Here's a quick and dirty solution to the problem of complexity of derivatives:

If the top 10 principle officers of a firm dealing in derivatives cannot explain exactly what that derivative is and how it works, in 200 words or less, the derivative cannot be sold, traded, bought, or otherwise handled by the firm.

The same 200-word rule applies to all Congressional committees required to oversee/regulate those derivatives.

No Tele-Prompters, no cheat sheets. Just live explanations.

It's immaterial whether a Nobel Prize-winning mathematician can explain it or whether the IT guru can explain how he modeled it. If those responsible cannot understand it fully, then it is not a product fit to be on the market.

This is different from products like HDTVs or a new hybrid car. Those products are complex, but evolutionary. Everyone can basically understand the mechanics behind them. No so with ephemeral derivatives.
10.12.2008 5:45pm
paul lukasiak (mail):
as someone else noted, its like blaming the death on the weapon. The blame lies with those who decided to use the machines to maximize their profits in the first place -- nevertheless, the nature of the weapon determines how easy it is to accomplish the task -- its a lot easier to kill someone with a .44 Magnum than with a 9mm pistol -- and you can kill a whole lot more people with a nuke than you can with an equivalent volume of hand grenades.

but the answer isn't to ban computerized trading, or exotic derivatives. Its to make computerized trading and exotic derivatives unprofitable by graduated taxation (higher taxes on more "complex" derivatives and on investments held for shorter periods of time, no taxes on "simple" investments when they are held for longer period of time.)
10.12.2008 6:21pm
David Warner:
Garbage in, garbage out. If you don't want your cat to explode, don't put it in the microwave. If you don't want your financial system to explode, don't pretend that south sea stock asset values will continue to appreciate forever.

Human error, not machine. Whether this is human or machine, is hard to say, however.
10.12.2008 7:20pm
Cornellian (mail):
Wow, so Skynet won't even need to find a way to activate the nukes. It will just take over the global financial system and wreak havoc. Can liquid metal terminators be far behind?
10.12.2008 7:23pm
kietharch (mail):
Norman Bates: thank you for your correction. I did not think of it that way but your are right.

The basis for commodity trading at the origin was to establish a market for crops and commodities that had yet to be produced. That is slightly different than going short or long on existing assets and it is easier for me to see the utility of that earlier arrangement.
10.12.2008 7:57pm
mls (www):
"If you don't want your cat to explode, don't put it in the microwave."

Sure, now you tell me.
10.12.2008 8:35pm
Paul Koski (mail):
I dare suggesting to have a closer look at research done in the field of accountancy. It may need to be curbed, similarly to genetics. Academic world HAS TO produce tons of master's, doctoral and other valuable papers to further academic careers. When these fall in hands of greedy managers and lawyers out there in board rooms, they mutate into "creative accounting" = plain cheating. Pity that noone listened to me back in the 90's, when I insisted that "creative accounting" is a fellony not a complement.
10.12.2008 8:37pm
Elliot123 (mail):
Keep in mind that the commodity may exist. After harvest, corn, for example, is all stored in elevators. It exists. The contracts for delivery before the next harvest are then based on future delivery of that existing commodity.

An example is trading a contract calling for March2009 delivery, and making that trade in November 2008. That corn is sitting in an elevator.

It is also possible to have a future contract for delivery of a bushel that has not yet been planted. For example, in November 2008, trade a contract for March2010 delivery. That corn hasn't been planted yet.
10.12.2008 8:41pm
Jim Rose (mail) (www):
After reading the article I feel a whole lot better about the Hadron Collider.
10.12.2008 9:59pm
PDXLawyer (mail):
I think one *can* blame the current global economic crisis on computers developed in the last few decades. After all, if computers didn't exist, financiers would have been constrained to avoid trading instruments and strategies which required too much computing power.

Similarly, one can blame the South Sea Bubble on the development of extremely seaworthy sailing ships during the few centuries immediately preceeding, which made global trade possible on a large scale.

In finance, as in every other field of human endeavor, a technical innovation usually becomes widespread *before* the academic or professional community develops any meaningful understanding of how it works in practice and how it fits into the larger scheme of things. Changes in institutions typically take even longer, and often come only when forced.
10.13.2008 12:09am
neurodoc:
Elliot123, if I believe that GM stock is going to decline still more in value, I can bet it will do so without ever owning GM stock; and if I am proven right, then by shorting GM, I will put money in my pocket. How can I do anything of the sort with Escalades without ever having to own one? (The car holds no appeal whatsoever for me.) If I believe that Escalades are going to decline in value, then I can sell one, but only if I first own one, and hope to be able to buy it or an identical one later at a lesser price. Alternatively, if I don't own one to start with, I can hold off on a purchase hoping to buy it for less at a later time. Either way, though, I wind up with a car I don't like and poorer the final cost of the car, not with the money I would have in my pocket from a successful short sale of stock.
10.13.2008 2:07am
Harry Eagar (mail):
Those nasty computers. Not programmed for moral risk, I guess.

When is the last time you heard the phrase 'triple witching hour'?

It seems we hardly needed all those sharpies after all, we could've outsourced everything to Bangalore. The Bangaloreans couldn't have done a worse job, and we've have saved money.
10.13.2008 2:48am
Arkady:
It might be instructive to read this piece by Bill Joy, who knows a little bit about computers and technology:


Why the future doesn't need us.

Our most powerful 21st-century technologies - robotics, genetic engineering, and nanotech - are threatening to make humans an endangered species.


Joy also references the passage from the Unabomber that Dooley references (it figured prominently in Ray Kurzweil's book, The Age of Spiritual Machines (1999)).
10.13.2008 8:24am
Elliot123 (mail):
"if I believe that GM stock is going to decline still more in value, I can bet it will do so without ever owning GM stock; and if I am proven right, then by shorting GM, I will put money in my pocket."

Somebody does own that stock. They loan it to you so you can sell it. Now you have a debt which must be repaid in stock. When you buy the stock, you pay that debt by giving the stock back to the lender.

You may not own the stock initially, but you are assuming a debt equal to the initial value of the stock.

I will be happy to make an agreement with you that lets you sell my Escalade, keep the revenue, and then give me a comparable Excalade in six months. During that period, you will pay me a monthly fee.

In six months, I have an Escalade back in my garage, I have all your monthly payments in my pocket, and I haven't had to look at the pig for six whole months.
10.13.2008 11:44am