Information Processing

Just another WordPress.com weblog

Archive for August 2005

Gecko and Van der Waals

with 3 comments

Apparently geckos have millions of tiny hairs on their feet (paws?) that adhere to almost any surface. As I recall, the individual hairs are actually using the Van der Waals force for adhesion. The photo is from this NYTimes article, which discusses attempts to produce artifical gecko paws for applied purposes. Imagine giant robots with sticky gecko paws 🙂

I really love the third photo on the right!

Written by infoproc

August 30, 2005 at 7:10 pm

Posted in physics

Startup update and patents

with 4 comments

We’re still in stealth mode, so no details. Our product testing is going well and the beta release on 1/1/2006 looks like a realistic target. I get to add patent filings to the list of other things I’m doing. In looking for prior art and related USPTO applications, I can’t help but think that (a) academic CS research can have a really low quality threshold and (b) the software patent situation is getting out of hand.

I say (a) because of all the nutty ideas I see academics publishing that will never work in the real world. If the basic idea were deep I could see pursuing it despite impracticality. But in these papers I see obvious, trivial ideas pursued (presumably consuming significant NSF, ONR, DARPA or other funding) in great detail. On the positive side, I suppose these lame projects still serve the purpose of training students. (BTW, I am referring here specifically to applied security research, where I think real world practicality is very relevant, not something more fundamental like algorithms.)

I say (b) because of all the silly things that have been granted patents. Officially, the threshold is that an idea be “non-obvious”, but many, such as the Amazon One-Click patent clearly fail that test. (Amazon has had a patent for many years now on a one-click process for online purchasing.) In security, inventors have been issued patents on ideas as vague as “method for detecting dangerous programs by matching code patterns against a database” or similar. Since companies have to protect themselves in this crazy environment, they have all become profligate submitters of applications to the USPTO. Microsoft alone has a target of 3000 patents per year! The whole system is clogged, and clogged with junk!

It is obvious why there are problems at the USPTO. It takes a very smart examiner, who keeps up with the rapid pace of technology development, to understand what is important or novel in a complex patent application. Anyone smart enough to do a good job is also smart enough to make several times as much on the other side working at an IP law firm. Also, it is not clear that there are any incentives in place to keep an examiner from having loose standards. A huge backlog of patents piling up on your desk ensures a negative merit review, but what counterbalancing mechanism is in place to discourage granting of bad patents? In rare cases, after years of litigation, a bad patent might be overturned, but are there any repercussions for the original examiner?

OK, enough ranting. As an amusing aside, this Business2.0 article lists a dozen or so ideas that VCs would like to fund. One of the ideas listed is something I previously worked on with a collaborator. We built a working prototype and shopped it to several funds before dropping it for a more promising idea (our present project). I guess we were talking to the wrong VCs! 🙂

Written by infoproc

August 30, 2005 at 4:02 pm

Posted in globalization, startups

Evolutionary theorist Robert Trivers

leave a comment »

Nice Guardian profile. Trivers linked the evolution of altruism to that of an ability to detect cheaters.

…There are less dramatic examples, however, which include sharing food, helping the sick, the very young, and the old, even when we are not related to them, and sharing tools and knowledge. All these are nearly universal human habits; in fact we describe societies where they don’t happen as inhuman.

This kindliness became part of human nature, Trivers argued, because kind instincts were rewarded and this happened because our ancestors lived sufficiently long lives in small stable groups to keep track of who owed whom favours. The great originality of the theory is not that it says that we are under certain circumstances naturally benevolent. Plenty of people had made that observation before. What no one had seen was that this benevolence requires a very strong sense of fairness if it is to become an established instinct. Fairness, or justice, has its roots for Trivers in the determination to see that other people are not cheating us, and taking favours without giving anything in return.

From abstract notions about the flow of genes he had come up with concrete and testable ideas about the ways our minds work; and it turned out to be demonstrably true that we find it much easier to solve logical puzzles if they are framed as if they are about cheating rather than an emotionally neutral subject, even though the two ways of putting the problem are logically equivalent.

The paper on reciprocal altruism, written before he had even gained a doctorate, has been enormously influential. Robin Dunbar, the professor of behavioural ecology at Liverpool University, says Trivers played a fundamentally important role in the development of modern evolutionary studies of behaviour and ecology. His four key early papers spawned (and continue to spawn) research in the study of both animals and humans. The importance of his contribution is beyond question. The modern field of behavioural ecology (the name under which sociobiology now travels) would simply not have been the same had he not written these papers.

Written by infoproc

August 28, 2005 at 6:18 am

Posted in Uncategorized

Dynamical hedging and Black-Scholes

leave a comment »

Derman and Taleb claim that you can derive Black-Scholes without assuming instantaneous replication of the option using cash and stock. (This assumption has some practical limitations in the real world.) I always thought the replication insight very important for justifying the use of the riskless rate to discount cash flows. As the authors note, formulae equivalent to B-S were derived by others such as Samuelson, but leaving the discount rate as an unknown parameter. Once you know the option can be perfectly hedged, it becomes straightforward to price it given any model in which you can compute the expected return.

For elementary discussion, see here.

Written by infoproc

August 27, 2005 at 5:29 pm

Posted in Uncategorized

Low returns next decade?

with 3 comments

Barron’s gives a nice analysis of historical earnings growth and P/E ratios for US equities. The inevitable conclusion? If history is any guide we are in for low returns over the next decade, with the central value for total returns at 6%, rather than the 10% we’ve become accustomed to. Under these assumptions, the equity risk premium will shrink considerably, and the probability that bonds beat equities becomes significant.

See earlier discussion here and here.

Earnings for the S&P 500 were $58.55 at the end of December 2004. The long-term rate of earnings growth is 6.1%. Compounding that for 10 years gives us an estimate of $105.85 in a decade. For the high and low estimates, we used the standard deviation of earnings growth of plus or minus 2.3 percentage points. Essentially, this means that two-thirds of 10-year periods will experience annual earnings growth ranging from 3.8% to 8.4%. We can then estimate that in 2014, earnings for the S&P 500 are likely to fall in the range of $85.02 to $131.17.

Over the past 55 years, the stock market has sold at an average of 16.4 times earnings, with a standard deviation of plus or minus 7.0 times earnings. This means that a reasonable range for multiples is 9.4 times to 23.4 times earnings. Using this methodology, we can estimate the likely range of values for the S&P 500 in 10 years.

The likely range for the S&P 500 Index in 2014 would be 761.59 to 3069.14. This equates to an average return of minus 1.7% per year to plus 12.2% per year. The base case scenario calls for the S&P 500 Index to be at 1735.94, a yearly 4.1% rate of appreciation. Adding dividend income of 1.9% to the appreciation yields a total return of 6.0% per year — something between minus 1.7% and plus 12.2% a year.

Written by infoproc

August 27, 2005 at 5:22 pm

Posted in finance

Foreign investors support US housing bubble

with one comment

…through their purchases of mortgage-backed securities. Note the Bank of China trader thinks the Fed will keep the bubble going for at least a few more (five?) years!

WSJ: Strong demand for mortgage-backed securities from investors world-wide is allowing American lenders to make more loans — and riskier ones — in a way that is helping prolong the boom in U.S. house prices.

The cash pouring in — not only from U.S. investors but increasingly from Europe and Asia — keeps stoking the housing market even as the Federal Reserve Board continues to raise interest rates, normally something that damps home prices. The market has shown a few signs of slowing recently, and talk of a bubble has grown louder, but prices continue to rise or remain at lofty levels as investors continue to gobble up mortgage-backed securities and banks keep lending.

“As the Fed has tightened, lenders have eased” terms for borrowers, says Mark Zandi, chief economist at Economy.com, a forecasting firm in West Chester, Pa.

Investment banks and other firms have been buying mortgage loans from lenders and packaging them into securities for sale to investors since the 1980s. But investor demand has surged in recent years, largely because in an era of low returns, mortgage-backed securities offer yield-starved investors much higher returns than government bonds.

U.S. lenders will make about $2.8 trillion in home-mortgage loans this year, according to the Mortgage Bankers Association. The MBA estimates that about 80% of these loans will end up in mortgage-backed securities. Mortgage-backed securities outstanding at the end of the first quarter totaled $4.61 trillion, up 61% since the end of 2000. In the same period, total Treasury securities outstanding grew 35% to $4.54 trillion.

Investors’ strong demand for mortgage debt, besides allowing lenders to offer many borrowers better terms, has also made it easier to offer mortgages to borrowers who might not easily qualify for a loan. The growth of the mortgage markets spreads the risks around. But some mortgage-industry analysts say lenders have become less stringent in their loan terms because they can sell almost any type of loan to those who package mortgage securities for investors.

“Loose lending standards are probably the single biggest thing fueling the speculative fever we have today” in housing, says Kenneth Rosen, an economist who is chairman of the Fisher Center for Real Estate at the University of California at Berkeley.

In a world of low interest rates, the market for mortgage securities is simply too big and profitable for many investors to ignore. Investors can earn about 5.5% on mortgage securities whose payments are guaranteed by Fannie Mae or Freddie Mac, government-sponsored companies. Those who can stomach greater risk can buy subprime mortgage securities, which come with no guarantee but can yield as much as 15%, according to Bear Stearns. By contrast, 10-year U.S. Treasurys yield about 4.2%; the equivalent government securities in Germany yield about 3.2% and in Japan 1.5%.

The buyers of mortgage-backed securities include U.S. pension funds, hedge funds and insurance companies. But overseas investors are the fastest-growing source of demand. The trade publication Inside MBS & ABS estimates that foreigners held $280 billion of U.S. mortgage securities at the end of 2004, or 6% of the total outstanding. The foreigners’ holdings rose 26% last year and have continued to bound ahead so far this year, Inside MBS & ABS says.

“There’s this insatiable appetite for mortgage-backed securities world-wide,” says Andrew Sciandra, a senior vice president at IndyMac Bancorp, a California thrift, who heads a team that creates those securities. In the past year, Mr. Sciandra has met with investors from places like Germany, France and Abu Dhabi. Asian investors now account for roughly 10% to 20% of mortgage securities sold by IndyMac.

For homeowners, the growing international demand for mortgages means it’s increasingly likely that the money they borrow to buy a home or refinance their mortgage is coming ultimately from outside the U.S. When Claude Gaty, a chef and co-owner of a bistro in Las Vegas, recently refinanced the mortgage on his four-bedroom Las Vegas home, the lender was IndyMac. But the bulk of the money came from investors in Asia.

IndyMac pooled Mr. Gaty’s loan with about 3,000 other mortgages that carry a fixed rate for the first three, five or seven years. Mr. Gaty is paying both principal and interest on his loan, but most of the loans in the pool are interest-only mortgages, which allow borrowers to pay no principal in the early years. When the $650 million offering of triple-A rated bonds backed by these mortgages came to market in June, it drew more than a dozen investors from Europe, Asia and the U.S., according to Deutsche Bank, which handled the deal. Such bonds typically yield 0.75 to 1.15 percentage point more than Treasurys, Deutsche Bank says.

The most recent entrant to the market is China. Its banks are rich with deposits from Chinese companies that earn dollars exporting to the U.S. Dollars have also been handed to some banks by the government in Beijing as part of its efforts to strengthen their balance sheets.

Until a few years ago, Chinese investors restricted U.S. investment mostly to Treasurys. Now, to boost their yields and because they consider the market safe, bankers from a number of institutions say they are devoting more of their portfolios to mortgage securities. Some bankers say their goal is to have 40% of their U.S. dollars in asset-backed securities.

China’s government also is testing U.S. mortgage investment. The country’s Bank of Communications, the only bank with a mandate to help manage China’s $700 billion of foreign-exchange reserves, has recently put a sliver of those reserves into mortgage-backed issues, according to a banker there. The State Administration of Foreign Exchange, the government agency in charge of the reserves, declined to comment.

Zhu Kai, who helps manage U.S. dollar investments at Bank of China, says in a rare interview that his mortgage-backed portfolio has “plenty of room to grow.” Mr. Zhu expresses confidence in the U.S. dollar and the health of the U.S. home market. Housing is so vital to the U.S. economy, Mr. Zhu and some of his counterparts at other Chinese banks reason, that U.S. authorities will prevent a bust.

Even the recent decision by the Chinese government to raise the value of its currency by about 2% isn’t likely to lead Chinese banks to shift their plans. “The timing may be a little bit surprising but we will not change our investment portfolio,” Mr. Zhu says.

While Asian investors have largely focused on triple-A-rated bonds, other investors are buying lower-rated debt. These bonds, which are created when bankers carve up pools of mortgages, offer higher yields, but also bear the first risk of losses should borrowers default. Investors who buy these bonds in effect set the standards for which mortgages are made by deciding how much extra yield they need to compensate for the added risks of lower-quality loans. They include real-estate investment trusts, hedge funds and investors from Europe.

Strong investor interest has also made loans available to borrowers with poor credit and many other people who might otherwise have trouble getting a mortgage. Subprime loans included in mortgage securities totaled $401.5 billion last year, nearly double the total for 2003, according to Standard & Poor’s. Meanwhile, loans with less than full documentation of the borrower’s income and assets accounted for 70% of mortgage securities rated by Standard & Poor’s in this year’s first half, double the level recorded in 2000.

“There’s no question that [lending] standards have loosened over the past couple of years,” says Arthur Frank, director of mortgage research at Nomura Securities International in New York. If house prices fall, “you may well have some pretty serious credit problems,” hurting holders of the lower-rated mortgage securities.

Mr. Zhu, the Chinese fund manager, is sanguine, for now. The U.S. housing market is “maybe losing a bit of steam,” Mr. Zhu says. “I think the monetary authorities, they don’t want this housing market to burst. I don’t think it is a bubble. But if things go on like this for another five years, it’s a different story.”

On a related note, the Economist can’t figure out why financial markets aren’t punishing economies for budget deficits. It is amazing how low real interest rates are now relative to the last 20-30 years.

…a more efficient international capital market is supposed to ensure that capital is allocated to the most productive use. Yet much of the recent inflow of foreign money into America is not financing productive investment, but a housing bubble and a consumer binge.

…The inevitable correction, when it comes, is likely to be all the more painful. When financial conditions tighten, investors are sure to become more discriminating. Sooner or later, the traffic lights will turn red.

Written by infoproc

August 24, 2005 at 4:35 am

Posted in finance, globalization

Man versus machine

leave a comment »

No, we’re not talking about Kasparov vs. Deep Blue, or the Turing test. We’re talking about bots invading online poker games!

While bots have been used to play the optimal strategy in other online card games, like blackjack, poker is a different animal. The biggest obstacles lie in the amount of information unavailable to the player and the need for the program to be able to employ a variety of strategies at different times, such as bluffing and laying traps for opponents, explained Billings, a doctoral student and master poker player.

“With chess – I don’t want to trivialize it – but it’s just a matter of calculation,” he said. “With poker, you really need to write a program that can think about the game and reason.”

The solution, in the case of the Vex Bot, was adding a layer of artificial intelligence over its ability to calculate probabilities.

“It will show you things that no human player has ever shown you before,” Billings said of the latest incarnation of the bot, which also has the ability to model its opponent’s behavior. “… One of the biggest advantages that programs have is that they have no fear, no shame. Humans can be intimidated. They will back off in the face of a very aggressive player. A bot will not. It has no compunction about doing whatever it will take to win. It will raise you with any two cards if it thinks that it has a very slight advantage based on your history. And it can induce a lot of anger and emotional upset. These things are ‘tilt monsters.’”

Rao, who served as one of the testers of Vex Bot, attests to its skills.

“It was a formidable foe,” he said of his initial encounter with the bot, before the addition of its new feature. “I can see that (the improved) bot, given enough hands, will become an absolute world beater.”

While the Vex Bot is undeniably at the head of its class, the mere existence of bots is a sensitive subject for operators of poker sites, all of which appear to have policies prohibiting their use.

…“If you’ve got a bot that can play 25 casinos, two tables apiece, even if you’re playing a (mid-level) $10/$20 game … that’s $1,000 an hour,” he said.

And if such a bot is created, how long will its author be able to keep it secret?

“It’s only a matter of time before a talented poker player who also happens to be a good developer decides she or he wants to be remembered as the author of the first bot that changed online poker forever,” an author who goes by the screen name “loic” lamented recently in the Twoplustwo.com poker forum.

More insidious, teams of bots playing at the same table and using out-of-band communication have an even stronger advantage. Even if most online games are not yet bots vs. humans, I imagine a lot of players are using poker software for strategy while playing online. (This is called “freestyle” play in chess.)

See related Wired article.

Written by infoproc

August 23, 2005 at 4:55 pm

Posted in Uncategorized