Technology applies survival of fittest to product evolution

By Simson Garfinkel
Special to The Seattle Times

E-mail this article Print this article

A 2-year-old Cambridge, Mass., startup has figured out how to harness the power of natural selection — by putting genes into felt-tipped markers.

Crayola, the art-supplies company, hopes to use this technology to develop a new generation of markers. It's better art supplies through selective breeding, or something like that.

The whole process seems like a mixture of Gregor Mendel and Dr. Frankenstein. Crayola will cross a population of top-selling felt-tip markers with a population of pens that have new genes — that is, new ideas dreamed up by Crayola's designers. These new pens might have a different kind of cap or different taper in the barrel.

The genetically modified pens will be evaluated by Crayola's customers and segregated into two groups — those customers like, and those they don't.

The pens nobody likes will be sacrificed, but pens that are loved will be allowed to breed and have children of their own. The whole process then will be repeated.

Mathematical model

Of course, Affinnova, the Cambridge startup, hasn't really figured out a way to put genetic information into inanimate objects. At least, not the kind of genes made out of nucleic acids.

What Affinnova has created instead is a mathematical model of Crayola markers in which key aspects of the marker's visual appearance are dictated by eight numeric parameters — the "genes."

To breed two markers, Affinnova's software combines two parents and creates a range of children that have some characteristics from each parent — and perhaps a random "mutation" thrown in every now and then for good measure. The software uses these parameters to draw a digital picture of a marker with those characteristics, without forcing Crayola to spend tens of thousands of dollars setting up a production run at one of its factories.

Finally, the images will be downloaded to the Web browsers of people who have signed up to be part of Crayola's focus group. These people vote on which markers they like and which they don't.

Affinnova has come up with a revolutionary new twist on the 1980s concept of "genetic programming." Back then, a bunch of computer scientists realized the principles behind natural selection could be easily applied to computer design. The scientists treated the actual machine code of a program as its genetic blueprint.

By cross-breeding the programs that performed well at a particular task, then simulating thousands of generations, researchers were able to evolve programs that could sort numbers, fly airplanes and even make money on the stock market.

But despite the technique's apparent success, genetic programming never really caught on because evolution is a very messy business: The programs that genetic programming evolved did a good job, but they were filled with apparently arbitrary code that made them extremely difficult for humans to analyze and figure out precisely why they worked.

Limits of gene technique

An autopilot evolved with genetic-programming techniques might do a better job flying a simulated passenger jet than one that had been painstakingly developed by a team of programmers, but there was no way the Federal Aviation Administration was going to certify such a program for controlling an actual airplane with actual passengers.

Likewise, many investment houses felt the same way. With evolution writing the code, there was no one to blame if the software lost a fortune instead of making one.

These old arguments simply don't apply to Affinnova, because the company's software helps designs evolve, rather than design actual programs. That's not much of a limitation, as it turns out. The software can be used to design a better breadbox, create a new national print-advertising campaign, or even create furniture.

With a large-enough gene pool and user base, Affinnova's technology will even find small market niches that could be missed by traditional focus groups.

Not surprisingly, this radical new use of genetic programming comes from a radical source — the biotech industry.

Affinnova's founder and chief executive officer, Dr. Noubar Afeyan, spent much of the past 18 years building biotech companies, most recently Celera Genomics, which played a critical role in the sequencing of the human genome.

"I've spoken with a lot of people who say, 'Isn't this the way that products are made today?' " Afeyan says.

And it's true: Today's market is an evolutionary process. For most products such as sneakers or felt-tipped markers, producing each generation is tremendously slow and expensive. With the Affinnova technology, new products can be evolved in minutes.

The technology is fun to use. You sit down at a computer and get to play the role of product manager, deciding which designs should be developed and which dropped. At the end of the process, your computer screen shows a product that you genuinely like — and perhaps a product that you wouldn't have thought up if you had worked from a blank sheet of paper.

One day, Afeyan says, customers will be able to develop their own products, then just click the "buy" button. But making that happen will take more than fancy computer graphics.

Simson Garfinkel is a technology journalist and author based in New England.

Copyright © 2002 The Seattle Times Company