Networks in natural ecosystems can be used to improve information processing systems in artificial intelligence (AI) according to new research at Japan’s Kyoto University.
We’re surrounded by such networks, from the food web to complex underground communication networks among plants and fungi.
But what’s that got to do with a machine learning algorithm?
Computers are essentially information processors. They do this through algorithms – taking information in and spitting it back out according to what the algorithm directs the computer’s software to do.
It should come as no surprise that, as with a great many things (flying, magnetic navigation, for example), nature got there first.
WATCH: For Australia’s research sector to stay competitive, we need a huge transformation in AI capability
Researchers at Kyoto University have demonstrated the computational power of ecosystems. And they believe it will help in rapidly developing AI technologies.
In their simulations, the scientists showed that ecological networks, like those between predator and prey organisms, can process information efficiently. This makes them strong candidates for use as computational resources.
The approach has been dubbed “ecological reservoir computing,” according to ecologist Masayuki Ushio PhD, who is now a principal investigator at the Hong Kong University of Science and Technology.
Two types of ecological network were studied to see if they possess “computational power.”
One was artificial. This “in silico ecological reservoir computing” models hypothetical ecosystem dynamics to simulate the response of the system.
The other, “real-time ecological reservoir computing”, measured the population dynamics of the single-celled organism Tetrahymena thermophila. For this test, the researchers adjusted the temperature of the environment in which the organisms were living – the temperature being the input data – and counted the number of cells present – the output data.
Read more: Unique turtles, crocodiles are more vulnerable to extinction, threatening important ecological niches
The study confirmed that the single-celled organism population was able to make predictions into the near future to affect the overall ecology (number of cells).
Hypothetical models “also suggest that there might be a link between high biodiversity and high computational power, shedding light on new values of previously unknown biodiversity,” adds Ushio. “A direct relationship between a community’s diversity and computational capability may enhance its biodiversity quotient [the number of species in a given area].”
“Our new computing method might lead to the invention of novel types of computers. Also, in developing a way to measure the information processing capacity of a natural ecosystem, we may find clues to how ecosystem dynamics are maintained,” says Ushio.
The study is published in Royal Society Open Science.