Ants, autism and technology

I once interviewed an autistic programmer who said she could detect flaws in software at a glance by spotting irregularities in coding patterns. She said she could help remote clients debug programs she hadn't seen in years by displaying a "printout" of the source code in her mind.

And Temple Grandin, the autistic "Anthropologist on Mars" in Oliver Sacks' book, An Anthropologist on Mars: Seven Paradoxical Tales (Vintage Books, 1996), in vivid detail compares her thinking to computing. "All of my memories are stored as images," she told me. "I can go and look at these pictures like Web pages on the Internet."

These women have severe limitations stemming from the neurological disorder known as autism, but they also have certain mental abilities that far exceed those of most other people. How is this possible?

They say it's because they think like computers. And they're not using computer as a casual metaphor, the way people sometimes refer to computers as "electronic brains." They know something about computers, and they insist their brains work the same way.

Comparing brains to computers -- and vice versa -- turns out to be of more than academic interest and can produce tangible benefits for the corporation.

Recently, Computerworld's Future Watch section [Technology, June 18] looked at what companies such as General Motors Co. are learning from ant brains. There's a growing field in computer science that deals with "agent-swarm" or "ant-swarm" intelligence. Rather than relying on complex, centralized logic, systems that mimic ant behavior use many small, autonomous software agents. Acting on the simplest of rules, these agents together can solve problems that are enormously complex when viewed as a whole.

Higher-level, or "swarm," intelligence emerges from those rudimentary rules in ways that would be difficult or impossible to preprogram into conventional software.

Tucker Balch, a robotics researcher at Carnegie Mellon University in Pittsburgh, has an ant colony in his office. He says the ants (which you can see at www.cs.cmu.edu/~trb) amount to a usable "library" of logic. "It takes two scientists to watch one ant," he explains. "One watches and announces what it's doing, and the other takes notes."

Now, imagine you're writing software to control the food-gathering behavior of a 1,000-member ant colony. You could write a hugely complex piece of software that would centrally direct the behavior of every ant over time, trying to allow for every possible interaction and contingency.

Or you could just give each ant a tiny program to run in its head and allow it to act on its own. According to researcher H. Van Dyke Paranuk, ants have the following rules programmed into their wetware (brains):

  • Avoid obstacles.
  • Wander randomly, but with a preference for the direction of ant pheromones (scents laid down by ants passing along the same path earlier).
  • When holding food, drop food pheromones and wander in the direction of the nest pheromone.
  • When at food and not holding any, pick it up.
  • When at the nest and holding food, drop it.

Following these few simple rules, ants over time will find the shortest route to food and will efficiently return it to their nests.

GM's Saturn subsidiary uses such autonomous-agent concepts to control the flow of part assemblies. Instead of each factory machine's actions being centrally programmed and unvarying, Saturn's machines can sense the type and state of a part and can adapt their actions on the fly.

Some factory workers don't like those machines because their behavior can't be predicted, and Saturn went so far as to install dummy switches to give workers the illusion that they had some control. Workers don't like computers to act too brainlike, it seems. I'll admit I'm uncomfortable seeing what the primitive ant can accomplish.

Neurologists perhaps could learn more about autism by studying computer science. And maybe computer scientists could build better computers, software and human/computer interfaces by studying high-functioning autistics like Grandin and the autistic programmer, just as they're building better systems by watching ants.

In September, Computerworld's Future Watch section will run a story on brain/computer interfaces. It's not about how brains mimic computers or vice versa; rather, it's about how the two increasingly will be directly connected. Don't miss it.

Related:

Copyright © 2001 IDG Communications, Inc.

  
Shop Tech Products at Amazon