Meesh Pi. Click to enlargeA most peculiar feat was reported in yesterday's Philadelphia Inquirer: Marc Umile, a filing clerk from the Philly area, was able to recite the first 12,887 digits of π from memory - an American record. (The current verified world record is for 43,000 digits by Krishan Chahal of India.)
In an odd twist, Umile performed his prodigious feat at a law office, and not for the Guiness Book folks. In these litigious times, it is obviously prudent to be prepared for intellectual property infringement in any activity such as this, which in this case comes under the heading of π's and torts ...
While Umile's feat is incredible, I am more intrigued by the physical/mental issues involved in the data entry, storage and retrieval of these digits. The actual amount of data is not the issue - 12,000 is a very small number of single digits when compared to the potential of the human brain. Data entry is not hard to comprehend, either, with Umile spending two-plus years memorizing the digits. (I am not commenting on motivation or sanity here. See the Inquirer piece for this!)
How in the world are the values recalled/retrieved?
Umile was able to produce 1,000 digits at a time typing them on a keyboard. What's not mentioned is whether or not he can start at any number position in π and proceed from there. i.e. if he were asked for the 5000th digit, could he do it? Or does he have to start from the beginning (or perhaps the previous 1000-number cutoff). My guess is that it's the latter - Umile recalls a number stream in a serial fashion, and probably relies on some prior quasi-pattern to get him off and running.
(I am just speculating, of course, and may be relying too much on my experience with piano. I play a fair amount, but I don't play from sheet music - everything I play is embedded in a sort of "muscle memory" that allows me to play without thinking - and just play with feeling. As a result, if I mess up a note or chord I often have to backtrack to the beginning of the song, or to some natural break, in order to play the rest of the song through to the end.)
Now compare the Umile feet with the legend of Kim Peek - a savant who was the inspiration for Rain Man. Peek has memorized 12,000 books! To top that off, he knows every zipcode in the US, and supposedly can provide travel directions within any large US city, or between them. Again, the amount of information here, while sizeably larger than the first 12,000 digits of π, does not sound like it should tap the potential storage capability of the human brain. However, it is the entry and retrieval activity that is staggering.
Peek can memorize a page in 8-10 seconds! How else to load those 12,000 books?
It is the data retrieval that sets Peek apart from Umile. Peek has the ability to access the data in a random access fashion, a sort of Ram Man. In a 2005 Scientific American account titled Inside the Mind of a Savant by Darold Treffert and Daniel Christenson, Kim is described as being able to recall incredible details about the books he reads some 4 months later, and can name the page(s) on which the details are written, as well as other complete passages.
Peek has a severe brain abnormality, with no corpus callosum linking his left and right hemispheres. Treffert and Christensen speculate on how this defect may lead to some of Peek's abilities, and their article presents some provocative ideas about memory, savants, and brain.
The reason for this post, in addition to pointing out the differences between savant ability and the ability to memorize a huge number of digits π, is to focus on the unique way that models and understanding need to be developed in brain studies. The approach is quite different from the typical inductive/deductive approach more associated with the basic sciences: brain research requires pathologies in order to understand normal behavior. It is like saying that the only way to understand how a pendulum works is to investigate pendulums that behave abnormally!
I don't know any Pendulums Behaving Badly, but it sure sounds like a good name for a band.
The message is clear, though - without RamMan and others with similar abilities/developmental issues, our normal is too normal to ever yield the understanding, prediction, and ultimate clinical applications to keep it normal.
Although I'm reasonably sure that Marc Umile, while wishing that he could memorize much more, much faster, would choose a corpus callosum over a colossal corpus of records.
For more on the π records, see the PI World Ranking List for rules, regulations, and how to enter.