Artificial intelligence is being used to predict our interests — from which movies we like to what groceries we need. And now researchers are using the tech to predict which species interact and better understand how extinction works.
“We're pretty much in a mass extinction event, and a lot of these extinctions are mediated through species interactions,” said John Llewelyn, a research fellow at Flinders University in Australia who worked on the study. “[A certain] species going extinct can … cause an extinction cascade. So species interactions are really important in that way.”
Researchers used machine learning, which uses AI to create models. They have a list of all the species and their traits, and it goes through decision trees that are like a large, repeated game of 20 questions. For a fox, it will ask questions such as “Is it over 10 pounds?”
After the AI classified the species, researchers tested its knowledge and found that it could accurately predict the interactions of mammals and birds. That had “never been done before,” according to Llewelyn.
Once the AI learns and categorizes a species as a predator or prey, researchers can try various scenarios and interactions to see how it might respond. For example, what would a fox eat if it moved to a different habitat, or if its rabbit prey died out?
“It works well even when there's no interaction data for the species we want to make predictions for, which is really handy if we're dealing with rare species, which are the ones that are most vulnerable to extinction,” Llewelyn added.
This technology could be helpful for Mountain West states, as natural disasters and climate change increase the probability of species moving from one habitat to another.
Llewelyn’s team is currently learning about past extinctions and plans to use the AI simulation for current and future extinctions soon.
Scientists have manually recorded species interactions over the years, but Llewelyn said that’s only a tiny fraction of the interactions that actually occur.
“Turns out we don't have good food webs for pretty much every ecosystem on the planet, we're missing data,” he said. “[So there’s] less confidence in our understanding of how extinction cascades could unfold.”
Llewelyn said this technology can help in the race against time to protect the hundreds of thousands of species facing extinction.
“People could go out and try and physically record all of those interactions, but it's very time consuming and very costly,” he said. “AI makes it quicker [and] more time efficient so we can take action when we need to.”
But AI is not perfect, and biased data could influence results. Llewelyn said his team had to use data on mammals with lots of predator-prey interaction records to avoid feeding the model incorrect or biased data. Additionally, this model was only used on mammals, birds and fish; researchers believe it could be more accurate by adding data on reptiles and amphibians.
Llewelyn hopes this can be a tool that helps scientists conserve species diversity.
“It's really important just because of the number of species we're losing at the moment,” he said. “We need whatever we possibly can to try and curb the extinctions.”
This story was produced by the Mountain West News Bureau, a collaboration between Wyoming Public Media, Nevada Public Radio, Boise State Public Radio in Idaho, KUNR in Nevada, KUNC in Colorado and KANW in New Mexico, with support from affiliate stations across the region. Funding for the Mountain West News Bureau is provided in part by the Corporation for Public Broadcasting.
