Tuesday, February 23, 2010

Data-Driven, rather than Hypothesis-Driven

Article about Google's reliance on data at Wired.com

Instead of using a semantic framework to build up a theory of language, Google mines its massive trove of data to find contextual word associations.

As Google crawled and archived billions of documents and Web pages, it analyzed what words were close to each other... "Today, if you type 'Gandhi bio,' we know that bio means biography," Singhal says. "And if you type 'bio warfare,' it means biological."

Want to introduce a new feature? Forget focus groups or relying on management to make decisions, run experiments on actual users!

But Google also has a larger army of testers — its billions of users, virtually all of whom are unwittingly participating in its constant quality experiments. Every time engineers want to test a tweak, they run the new algorithm on a tiny percentage of random users, letting the rest of the site’s searchers serve as a massive control group.

Blog post about data-driven versus hypothesis-driven science

The new data-driven approach suggests that we collect data first, then see what it tells us.

More info can be seen at a previous blog post.

Compressed Sensing

Article at Wired.com

Compressed sensing works something like this: You’ve got a picture — of a kidney, of the president, doesn’t matter. The picture is made of 1 million pixels. In traditional imaging, that’s a million measurements you have to make. In compressed sensing, you measure only a small fraction — say, 100,000 pixels randomly selected from various parts of the image. From that starting point there is a gigantic, effectively infinite number of ways the remaining 900,000 pixels could be filled in.

The key to finding the single correct representation is a notion called sparsity, a mathematical way of describing an image’s complexity, or lack thereof. A picture made up of a few simple, understandable elements — like solid blocks of color or wiggly lines — is sparse; a screenful of random, chaotic dots is not. It turns out that out of all the bazillion possible reconstructions, the simplest, or sparsest, image is almost always the right one or very close to it.

This question really highlighted the utility of compressed sensing:

Digital cameras, he explains, gather huge amounts of information and then compress the images. But compression, at least if CS is available, is a gigantic waste. If your camera is going to record a vast amount of data only to throw away 90 percent of it when you compress, why not just save battery power and memory and record 90 percent less data in the first place?

"Good" Beats "Innovative" Nearly Every Time

Article at BusinessWeek by Scott Berkun

But it's clear most companies fail not because of their lack of inventiveness; it's their lack of basic competence. Most leaders fail to prevent bureaucracy, hubris, and too many cooks from killing good ideas before they ever get a chance to make it out the door, resulting in the mediocrity we know too well.

Most projects aimed at innovation fail because creators become distracted by their egos from the true goal: to solve real problems for real people. (emphasis added)

Wednesday, February 10, 2010

The Way [Paul English] Works

Article at Inc.com

It's like hot potato. Except I take it seriously. When the phone rings, I literally jump over the desks just so I can get to the phone before anyone else. I love talking to customers, even angry ones. I learn a lot from them about how to make the site easier to use. When the call's over, I'll say, "If you have any follow-up questions, my name is Paul English; I'm the co-founder of the company." I'll give out my personal cell-phone number. Only one out of 20 people might actually call, but they're blown away when I do that.

I'm not attempting to excerpt-summarize this article - I think the above excerpt gives a flavor of the Paul English's approach. The article is well worth reading in its entirety.

Does Slow Growth Equal Slow Death?

Joel Spolsky article at Inc.com

If real life is anything like the Civilization series by Sid Meier, I would argue - "Yes!" This one quote from a Civ4 strategy guide sums it up best:
And in this, you’ll see a hint of what’s necessary to win on higher difficulties: more aggression.

Back to Joel Spolsky:
I have always believed that there is a natural, organic rate at which a business should grow, and that if we expanded too fast, the wheels would come flying off.

Uh-oh. Are we actually losing our market leadership position because we're careful?

It's entirely possible. Think of it this way: If you're growing at 50 percent a year, and your competitor is growing at 100 percent a year, it takes only eight years before your competitor is 10 times bigger than you. And when it's 10 times bigger than you, it can buy 10 times as much advertising and do 10 times as many projects and have meetings with 10 times as many customers. And you begin to disappear.

Cultivate Teams, Not Ideas

Blog post at CodingHorror

I wouldn't call ideas worthless, per se, but it's clear that ideas alone are a hollow sort of currency. Success is rarely determined by the quality of your ideas. But it is frequently determined by the quality of your execution. So instead of worrying about whether the Next Big Idea you're all working on is sufficiently brilliant, worry about how well you're executing.

This is all your app is: a collection of tiny details.
- Wil Shipley

If you give a good idea to a mediocre group, they'll screw it up. If you give a mediocre idea to a good group, they'll fix it. Or they'll throw it away and come up with something else.
- Edwin Catmull (Pixar)

Tuesday, February 2, 2010

Slime Mold vs Traffic Planners

Freakonomics article
MSNBC article

A very elegant experimental setup:

The scientists placed food deposits (oat flakes) in a pattern that mimicked the distribution of population in the greater Tokyo area. They also discouraged mold growth in areas corresponding to obstacles like ocean and mountains by placing light sources (mold’s sworn enemy) in these spots. The researchers then introduced a single deposit of the mold on their mock central Tokyo and let the slime do its thing.

The result? The mold formed a network that closely mimicked the actual Tokyo railway map. In terms of efficiency and fault tolerance, the mold performed about the same as the real Tokyo system, and it did so at a slightly lower cost.


Organic development can complement the planning efforts of a central intelligence. Planners see the big picture, but may have limited information about the small details. Organic planning accumulates the collective wisdom of myriad individuals who each know only a very small part of the picture, but know their part very well.

An interesting thought experiment: think about how this describes how a company functions. Are there any lessons that companies (or any other group effort) can learn from slime molds?