Ti Point Tork

Nine to Noon: 16 July 2009

Listen to my 16 July 2009 appearance on Radio New Zealand’s Nine to Noon show. I spoke about Science Foo Camp which was at the Google campus last weekend: discovering new science from huge amounts of data, hormonal traders, personal genomics, and open publishing.

Below are my notes. I will update this post with links to audio when Radio New Zealand post it. Correction on the air I said the hormonal trader paper was published in PLoS but it was actually in the Proceedings of the National Academy of Sciences.

Links

1. Theodore Gray author of Mad Science
2. Hormonal Traders
3. Google Translate
4. PlosOne – Public Library of Science
5. Universal Scaling Laws in Biology — my hero’s paper
6. 23 And Me – personal genomics company

Science Foo Camp

Weekend, at Google, ~150 people, no schedule. Food and drink laid on. Mixture of people: astronomers, biologists, chemists, computer scientists, bloggers, journalists, psychologists, economists. “real invisibility cloak”. Nobel prize winners. Nature, O’Reilly, Google sponsors.

Data-driven Science

Overarching theme is nominally “data-driven science”, idea that from lots of data you can use computers to find patterns and thus discover new scientific properties and laws. Biologists do this all the time. They’d be stuffed without computers. Not to write their reports, but to analyse their data: you copy a piece of DNA into a squillion copies, split a those into millions of random chunks, wash the chunks over a grid each bit of which sticks to a different known segment of DNA, so then by looking at what stuck you can figure out what bits are in there and then the algorithm reassembles the bits into a coherent whole. It’s been this way for ages, and it’s time-consuming and painful.

Astronomy works the same way: humans don’t look at every inch of the sky now. Instead, algorithms scour the huge amounts of information to find patterns and surprises.

Science, of course, hasn’t quite caught up. This inverts the way science works. Used to “hypothesis driven science”. This is “data driven science”. “He writes software” is the kiss of death in science. As one attendee said, “half my graduates can’t decide on any one day whether they’re computer scientists or astronomers”.

Obviously useful to Google: they specialise in doing clever things with lots of data. Google Translate, for example, translates between a dozen or so languages. They didn’t build syntax models and teach it the rules of grammar for English, French, Spanish, etc. Instead, they set their software loose on documents with paired texts in the languages and built up the models and conversions. The result is something that won a Defence Department contest for translating between English and Chinese and Arabic.

Hormonal Traders

There were a few economists present. One, John Coates from Cambridge, studied the hormone levels in traders. The hormones are cortisol and testosterone. Testosterone, released when they make a profit, increased by 74% during a 6-day winning streak. Cortisol, associated with the volatility of the market, went through even more extreme swings. if testosterone persists, it might explain some of the bull market foolishness: while prices go up, you experience testosterone, which makes you take risks even as the market tanks. 260 traders, 4 female in trading floor; sample size = 17 representative of whole.

(show a man a picture of attractive woman and he makes poor decisions)

Personal Genomics

Ability to make sense of your own genome. USD$399 to 23andme.com gets you swab, mail back and they analyse your genome. Looking for SNPs — points where your genome differs in important ways from ‘the average’. Get a report with everything from “greater risk of breast cancer” to “caffeine has less of an effect to you” and ancestry (“you have Namibian DNA!”). They also conduct research …. ask questions and hope that when they know your DNA they can make connections between your answers and your DNA. Founder = google founder’s wife.

Open Publishing

Nature Magazine was a sponsor. Nature makes lots of $ charging research institutions for subscriptions to their journals. However, science depends on reach. Publishing going through a rough time, as are newspapers. Public Library of Science … open access peer-reviewed journals. Nature working on online videos, wikis, classrooms, and more trying to make sure they’re part of the future and not left in the past.

Open notebooks an interesting part of this: reproducability important in science. Papers publish findings, but rarely all the data. But can’t reproduce the analysis of the data or find competing explanations without access to the data. Fear of being scooped in further research. Scientists are people too. Increasing movement, analogous to open source, to release the data for scientific experiments in science. Then … how to preserve it, license it, attribute it, etc.

Mad Science

Theodore Gray. Made icecream with liquid nitrogen. Has a book of “fun things that you probably shouldn’t do”. graysci.com. The book is fun of great stuff: “white sun” from phosphorous, exploding soap bubbles (hydrogen gas), making a levitating magnet, and more. Explains the science (writes for Popular Science), great photos, fantastic coffeetable book.

My Hero

Geoffrey West wrote a great paper relating lifespan, heart rate, and mass of the organism in a rule that holds across six orders of magnitude: from mitochrondria through to elephants. All comes down to the physics of pushing liquid through capillaries. He figured this out, wrote it up, and blew my mind when I read the paper a decade ago. I was tickled to meet him.