November 29, 2000

6:28 PM

"One of the most esteemed documents of modern paleontology is Stephen Jay Gould's doctoral thesis on shells. According to Gould, the fact that there are thousands of potential shell shapes in the world, but only a half dozen actual shell forms, is evidence of natural selection. Not so, says Wolfram. He's discovered a mathematical error in Gould's argument, and that, in fact, there are only six possible shell shapes, and all of them exist in the world. " A must-read article.
posted by costas (14 comments total)

Warning: said article contains nothing about Gould's thesis, the mathematical error, or Wolfram's discovery other than what's quoted in the extract above.

The article looks like some advance drumming for Wolfram's forthcoming book.
posted by grimmelm at 10:40 PM on November 29, 2000

Gould's thesis only gets a quick mention as an example of one of the theories Wolfram has debunked using his cellular automata theory. A New Kind of Science
is the name of his new book, which the article discusses in length. From the article:

"Within 50 years, more pieces of technology will be created on the basis of my science than on the basis of traditional science. People will learn about cellular automata before they learn about algebra." - Stephan Wolfram

I highly recommend you read this article, it sounds like he's up to something really big.
posted by stazen at 10:58 PM on November 29, 2000

It sounds like he's a megalomaniac.
posted by Steven Den Beste at 11:31 PM on November 29, 2000

I can top Wolfram. Within *ten* years, more writing will be based on this MeFi comment than all previous writings taken together. People will study this comment before they study Shakespeare.

It is the ultimate non-binary one-dimensional CA. Instructions: for each letter, if that letter is between two vowels, then make it an "e". If it is between two identical letters, make it the same as that letter. If it is between a vowel and a consonant, advance it one letter in the alphabet. If it is between two consonants, replace it with the letter two spaces before it in the alphabet. Repeat.

Woah. Trippy, huh?

(Seriously, if you are interested in the idea that natural selection has considerably less explanatory efficacy than is commonly believed, read the first few chapters of Origins of Order. I'll read Wolfram's book when it comes out, but I think Steven is closest to the mark here and I'd like to take that journalist out and give him a good shake.)
posted by sylloge at 12:26 AM on November 30, 2000

Reading over the Forbes article, I have this odd feeling that there's an ideological undercurrent to his interest in Wolfram's twist on the evolution of complexity. If you move Wolfram's ideas, oh, say, into the world of economics and simplify his message like mad, I can see the implications being a lot cooler (from a Forbes perspective) than Gould's annoying take on natural selection, with all those odd byproducts of circumstance and path dependence and lack of an overall "direction."
posted by grimmelm at 12:45 AM on November 30, 2000

Sorry, but the fact that something is desireable doesn't mean it's either possible or true.

Reality is what it is, not what we want it to be. The fact that his theory might be more pleasant (or "cooler") doesn't mean it's more likely to be valid or correct.

posted by Steven Den Beste at 1:30 AM on November 30, 2000

Speaking of "from a Forbes perspective," I always wonder when I hear things about theories for forecasting the stock market: if they worked, wouldn't they not work? That is, if a mathematical model predicted the stock market with accuracy, it would only work as long as only a few people knew about it; if every investor started using its advice, the assumptions built into the model about the way people invest would be obsolete and the model would break... right? You'd have to start building in factors for people using the model itself; a recursive formula, mebbee? But there always seems to be a way around it... In any case, seems to me that the search for a market forecaster is a science akin to alchemy: doomed to failure from the start. (But I'm no mathematician.)
posted by letourneau at 7:39 AM on November 30, 2000

Well predicting the stock market does work; I have friends who do exactly that. My company's software predicts retail sales --which is not quite as chaotic as the stock market, but for a large retailer, it could be.

From the little I can understand from the article, Wolfram's theory doesnot dispute chaos, but the origins of it...
posted by costas at 7:58 AM on November 30, 2000

Read Malkiel's A Random Walk Down Wall Street. The general idea is that all available knowledge is automatically priced into the market, making non-insider stock picking impossible. If there was a "working model" such as the one descibed above, it would get priced into the market, too.

80% of mutal funds, managed by "experts" with teams of researchers, underperform major indices. You can come up with a range of models that should work, and some that work generally, but they have little short term accuracy.
posted by tranquileye at 8:21 AM on November 30, 2000

That's what's known as the "efficient market theory", Tranquileye. EMT was developed by Eugene Fema, a professor at the University of Chicago (and a leading light of the libertarian-leaning, anti-Keynesian "Chicago school" of economics). It comes in various forms (weak, where all information about the stock is priced in; semistrong, where all public information about the company is also priced in; and strong, where all information, public or not, is priced in). Many economists don't accept EMT, particularly strong EMT.
posted by snarkout at 10:11 AM on November 30, 2000

Steven:
Sorry, but the fact that something is desireable doesn't mean it's either possible or true.

Yes, but the gorgeous thing is that those things which are true turn out to be desirable once you figure out how to look at them...

letorneau:
You'd have to start building in factors for people using the model itself; a recursive formula, mebbee?

Recursive formulae? Iterations based on the results of previous iterations? Sounds awfully nonlinear all of a sudden.

-Mars
posted by Mars Saxman at 12:00 PM on November 30, 2000

Well, stock market models do not have to be correct all the time: they just have to be correct 50+x% of the time, where x represents enough money for the fund to overcome overhead and keep investor faith.

Is that possible? yes; stock-market efficiency may or may not be an accurate model, but the important factor is not whether or not the stock market adjust for all information, but when it does. The idea is that a computational model, trained and fine-tuned over time --and adjusting to transient market conditions-- can sniff out inefficiencies in the market faster than the market can adjust for them. And yes, that's not only possible, it's done on a regular basis. Read this old Wired article on D.E. Shaw for one example.
posted by costas at 12:01 PM on November 30, 2000

Here's a simple and infexible version of wolframs code that I just hacked together in java.
Anyone want to hire me?;)
And if it doesn't seem to be working then tell me okay?
posted by davidgentle at 10:20 PM on November 30, 2000

Sorry. Not his code. My code. His idea.
posted by davidgentle at 10:20 PM on November 30, 2000