Andrew Adamatzky is a professor in Unconventional Computing at the University of the West of England, and throughout his career he has indeed taken an unconventional approach to computing. Instead of servers and microchips, he uses a single-celled slime mold. The brainless, seemingly unintelligent organism (Physarum polycephalum) has been harnessed to transfer specific colors between foods dyed with food coloring, move a small boat through a gel medium and even solve mazes.
His latest project, though, is perhaps the most unconventional of all. Over the past several years, he and Andrew Ilachinski of the Center for Naval Analyses have used the slime mold to do something astoundingly complicated: design plans for national highway systems. And each time, within days, the mold created routes that are remarkably similar to actual systems designed by human engineers.
The slime mold, it turns out, is specifically evolved to do one thing very well: efficiently transport nutrients from one location to another. As the pair of researchers explained in a New York Times op-ed this past weekend, the forest-dwelling organism forages for microscopic nutrient particles by sending out protoplasmic tubes of slime and maintaining the links between these food sources as efficiently as possible.
So Adamatzky, Ilachinski and a team of colleagues decided to use this ability to determine exactly which routes would be most logical to build if one were designing, say, the U.S. Interstate Highway System from scratch. As detailed in an article that will soon appear in the journal Complex Systems, the team replicated the United States for the mold by overlaying a agar gel dish shaped like the country on top of a map and placing a food source (rolled oats) in each of the 20 most populous metropolitan areas. They repeated the experiment for 13 other geographical areas, including Brazil, Africa and Germany, and replicated it several times for each map.
After placing the oats, they let the slime mold spread naturally from the largest city or capital, and observed what routes it determined were most efficient for transporting the nutrients across the country. As depicted in the video above (showing one of the experimental trials for Canada) and the image to the right (showing the results of a trial for the United States), the slime mold repeatedly created routes that were strikingly similar to the ones laid out by decades—and sometimes centuries—of human engineering.
“Physarum is renowned for building optimal transport networks, which minimize distance of cytoplasmic transfer but also span as much sources of nutrients as possible,” Adamatzky told Wired last year. “Ideally, human-built roads should fulfill the same criteria.”
Indeed, it seems that the U.S. Interstate Highway System does fulfill the same criteria, as the mold created routes that match the majority of interstates. In nearly every trial, the mold grew links that correlate with Route 95 from New York to Boston and Route 45 from Dallas to Houston; In most trials, the mold closely replicated highways that span the major cities of the southwest (Denver, Albuquerque, Phoenix and Los Angeles) and the eastern seaboard (Route 95 all the way from Boston to Jacksonville).
The mold’s designs correlate even more closely with Belgium, Canada and China’s highway systems, suggesting that those are more efficient in terms of minimizing travel distance between population centers and spanning as many densely populated areas as possible.
Why do the mold’s and humankind’s route creations match so closely? The authors speculate that, because many early roads were determined based on prehistoric human footpaths and animal trails, and many modern highways are in turn based on these early roads, our design process is really not so different from the slime mold’s: using trial and error to find the most convenient paths for travel over time.
The experiments are fascinating—and maybe a little creepy—in the way they demonstrate that seemingly unintelligent life forms can perform extremely complicated tasks. But they also hint at potential applications in the real world. Adamatzky seeks to devise means of problem-solving that are cheaper and simpler than silicon-based computing, and the mold has already been used to solve a number of arcane spatial mathematical problems. The mold requires relatively little expertise or laboratory resources to use, and it is a more sustainable computing option than traditional electronic circuitry.
One practical application that immediately comes to mind is using the mold to analyze which routes would be most efficient to build for countries that don’t yet have developed national highway systems. They could also be used to efficiently model ideal pathways on a much smaller scale, such as a college campus or public park.
Regardless of what we might end up using it for, one thing is already clear: the brainless slime mold is much smarter than we think.