Skip to Content
Uncategorized

The Art of the Possible

Can Eric Bonabeau’s Hunch Engine expand your mind?
September 1, 2006

I am a low S trapped in a high S’s body.

Illustration by Marc Rosenthal

S is for steadiness in the DISC personality-profiling system, outlined in 1928 by psychologist William Moulton Marston (also the creator of Wonder Woman). Assessing colleagues and customers according to their need for dominance, influence, steadiness, and compliance is a skill still taught in many corporations. My outwardly high S means that if you ask me whether I would like to vacation in Italy or India, I will say Italy, because I’ve already been there, and I know what it’s like. If I were on Deal or No Deal with Howie Mandel, I’d choose the cash I knew I’d won over the next briefcase every time.

Yet in my secret heart I am a low S. The best things in my life have turned up by surprise. If I let my boyfriend plan our vacations, we end up in exotic places like Shanghai or Belize, and I have a great time. I love new things when I encounter them; I just can’t remember that beforehand.

So when I heard about the Hunch Engine, my reaction was predictably dismissive. I’d been assigned to write about the design and search tool for this special TR35 issue of Technology Review because it had just been unveiled as the latest brainchild of theoretical physicist Eric Bonabeau, whom the magazine had named four years earlier as one of our top young innovators (see “TR100/2002,” June 2002). In 2000, Bonabeau had founded a company called Icosystem to commercialize his ideas about the “swarm intelligence” that emerges from systems, such as ant colonies, whose individual parts are not themselves intelligent.

Multimedia

  • Video: Evolutionary Design

According to the company’s publicity materials, the Hunch Engine is software that uses evolutionary algorithms to breed solutions to science, engineering, business, or design problems–solutions that no human mind could have predicted. Icosystem claims that evolutionary algorithms expose ideas to a kind of natural selection, allowing users to “reach beyond the limits of their imagination.” But the notion that serendipity might produce better results than thinking and planning left me suspicious. That was before I heard about the French letter carriers.

In June, I spent an afternoon with Bonabeau at his company’s headquarters in Cambridge, MA. Bonabeau, who has degrees from Paris-Sud University, the École Polytechnique, and the École Nationale Supérieure des Télécommunications and was a research fellow at the Santa Fe Institute, is a very low S–and an almost negligible C. Over tea, we talked about the menu at El Bulli, a notoriously experimental restaurant north of Barcelona where reservations must ordinarily be made a year in advance but where Bonabeau had, the week before, won a table after a relentless series of e-mails to the ­maître d’. It sounded like exactly the kind of place I would never try on my own but that I in fact enjoy.

Designing efficient mail-delivery routes, Bonabeau explained to me after our tea, is an age-old headache for post-office bureaucrats. The route that’s shortest in distance or fastest in time may be unfeasible for the most illogical of reasons, such as the belief among postal carriers that it’s inefficient for any two mail routes to cross. Or the mailman may simply dislike his new route. A large percentage of the mail-route reorganizations attempted by post offices around the world lead to labor-management tensions and strikes, Bonabeau says.

In 2004, when the Belgian postal department turned to optimization software from the field of logistics, the result was no better. The software produced a set of routes that should have taken the department’s employees 5 percent less time to complete. Unfortunately, Bonabeau says, “the mailmen hated it, and they dragged their feet,” taking 30 percent longer to complete the new routes. “There are failed implementations of optimization software all over the world in postal organizations–because it’s not an optimization problem at all,” he says.

It is impossible, Bonabeau is saying, to design a web of postal routes around a city that perfectly fits the sensibilities of both managers and mailmen. There are only better solutions and worse solutions. It’s even difficult to say what makes one solution better than the next, since “sensibilities” can’t always be systematically predicted or quantified. One postal worker may prefer a longer route to a shorter one, if it goes by the patisserie where the owner greets him every day by name. One worker may prefer to walk downhill all day because of his fallen arches, while another likes a steep uphill grade because he’s training for a triathlon.

No software can take all these things into account. So programmers shouldn’t even try, argues Bonabeau. Instead, they should look to nature, in which random mutation and natural selection produce the features that fit an organism to its environment. Evolutionary algorithms start with a few solutions and a set of constraints, then breed generation after generation of new solutions by recombining selected elements from the previous generations (see “Technology Design or Evolution?” July/August 2006). In order to steer the solutions in a certain direction, the algorithms need only be told which solutions in each generation to recombine. The choices can be made automatically, or they can be made by humans.

And that is where a program like the Hunch Engine comes in. Last year, the French postal agency La Poste hired Icosystem to spend a week combing the kinks out of the route system in a small city two hours north of Paris. On the first day, Bonabeau recounts, Icosystem handed a set of six route options to each of the city’s 40 mail carriers. The carriers handed in their ranked choices the next morning, and these were fed into the Hunch Engine. The program randomly recombined the routes to generate new ones, then picked out the ones with the highest-ranked “parents.” That night, the carriers ranked the new routes, and the cycle continued. After a few days of this, a persistent web of routes began to emerge; the program had homed in on a solution. And when the new routes were put into practice, the letter carriers said they were happier. Or less malheureux, as the case may be.

The astonishing thing, says Bonabeau, is that “the mailmen never say why they like this route or that one.” Bonabeau believes that programs based on the Hunch Engine can help people with a particular class of problems: the kind where they have a hard time expressing exactly what they want, but they know a solution when they see it.

Bonabeau is careful to emphasize that evolutionary algorithms are nothing new. Computer scientists have been experimenting with them since the 1980s, and Icosystem has been selling industrial design tools based on them for several years. One tool, for example, helps pharmaceutical researchers breed biological molecules that are likely to interact favorably with receptors in the body.

What’s new is the use of evolutionary algorithms in programs that laypeople might use to invent things. A simple demonstration on Icosystems’ website, for instance, asks a user to select a few initial designs for ­Mondrianesque wallpaper or bathroom tiles; the designs’ evolution can then be directed toward the pattern the user likes best. The first standalone commercial service based on the Hunch Engine will debut this fall, when Icosystems launches an online company-naming service. Bonabeau says that for $15 or so, the naming engine will let a user recombine random phonemes and filter the resulting names until something pleasing, inoffensive, and non-trademarked emerges.

I’d been intrigued by the mailman story, but here I balked. The idea that a beautiful name might emerge from a mindless program offended me. Surely, I told Bonabeau, the quality of the solutions the Hunch Engine can generate must be limited by the quality of the real-world rules programmers are able to write into their software. These rules, I argued, must often be too numerous or subtle to capture. What kind of world would it be, I asked, if we let chance and mindless algorithms substitute for human creativity?

It was, of course, my outer high S speaking: I want to know where we’re going and how we’re getting there. But as usual, I was cutting myself off from exciting possibilities.

There’s no need to capture all the rules, Bonabeau replied. “Your question is really about the design space,” he said. “How can you make sure it’s rich enough that you’ll have a chance of finding something interesting? Well, there are cases where the design space is well defined. A good example is cooking, where, with five ingredients and five ways of combining them, I can already create more dishes than you can count. In those cases the Hunch Engine helps you navigate the design space. You can never be sure you’re exploring the space in an exhaustive manner. But you’re navigating it in a way that is a lot smarter than a random walk and a lot more empowering than being forced into a solution.”

Left on my own, Bonabeau might as well have been telling me, I tend to putter in one corner of my life’s design space. I am afraid to trade the route I know for the new one that’s probably better. But the Hunch Engine may give us inner low Ss a taste of the wider possibilities–and a reminder that in the end, life will find its own way.

Wade Roush is a senior writer at Technology Review.

The Hunch Engine
Icosystem
Cambridge, MA
www.icosystem.com/hunch.htm

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.