Poor flatworm. It may not be the most complex creation of all time, but it can do a few things that us humans wish we could do: If it doesn’t get eaten, it can practically live forever, and if you slice one in half, it can regenerate itself in a wondrous way.
Naturally, humans have inflicted all kinds of cruelties on the flatworm in the hopes of unlocking its secrets. Cut off a flatworm’s head, and its head grows back. Cut one into pieces, and a new flatworm regenerates from each piece. Zap a flatworm with radiation, and it repairs itself back to health. The mystery of the flatworm’s durability has bugged evolutionary biologists for over a century—and worm after worm has been cut up with little success. Even the promisingly-titled International Flatworm Conference hasn’t been able to change that.
Most recently, researchers at Tufts University in Massachusetts wanted to find out what form the regenerated pieces of tissue take, or to be more precise: What exactly has to happen in the cells for two halves of what was once a single flatworm to become two complete, independent worms. So, as demonstrated in a paper published in PLOS Computational Biology, the researchers put an artificial intelligence to work on the problem—and not only did their machine solve the mystery, but it only took 42 hours. Take that, humans.
The AI relied on an algorithm that continuously simulated and modified possible flatworm gene networks—much too complex and time consuming a task for humans. After three days, the computer found a model of the flatworm’s gene network that fit with the results of over a centuries worth of studies. “One of the most remarkable aspects of the project was that the model it found was not a hopelessly-tangled network that no human could actually understand, but a reasonably simple model that people can readily comprehend,” co-author Michael Levin told Wired.
The discovery "represents the most comprehensive model of planarian regeneration found to date," he also said.
The cut-up organism’s data gets compared to many of the AI-simulated gene networks. It’s too time-consuming for humans, but the computer found a match in 3 days that fit into all previous studies. Image: Tufts University/PLOS Computational Biology
Does this mean AI is about to replace all the inquiring minds of science? We’re not that far yet. AI still requires humans to teach them their own language, after all. Nonetheless, the success of this experiment demonstrates the potential of artificial intelligence in totally new fields: "The invention of models to explain what nature is doing is the most creative thing scientists do—it's not just statistics or number crunching, this is the heart and soul of the scientific enterprise," Levin told LiveScience. "Here, the computer really did give back more than what was put in.”
The fact that an artificial intelligence was confronted with a scientific inquiry and independently worked out a theory to solve it without human assistance, instead of just digesting data, is no small feat. Researchers are celebrating the new model as a giant breakthrough in both biology and the research of artificial intelligence. And all this without ever cutting up a single worm.