Many people have their doubts surrounding artificial intelligence. New technology is scary, and the incredibly rapid growth in which AI has seem to undergo throughout the past few months certainly has its concerns. Since these concerns are definitely valid, it’s nice to see some positives created by AI, like UCLA’s use of the technology to predict landslides.
According to phys.org, researches have worked with deep neural networks (DNNs) for several years to develop landslide prediction models. Variables related to the creation of landslides, like terrain shape, soil type, rainfall, earthquakes, and much more, as well as historical data relating to past landslides, are fed into these DNNs, allowing them to learn and predict future landslides with quite a high level of accuracy.
Those systems, unfortunately, fail to “show their work”, making it impossible to actually understand what factors have the highest influence in sparking the natural disasters. UCLA doctoral student Kevin Shao and postdoctoral researcher Khalid Youssef (co-first authors of the journal paper) have teamed up with UCLA associate professor Seulgi Moon and professor Louis Bouchard to develop a new AI model that could solve this problem.
Their new superposable neural network (SNN) allows separate layers of the network to run alongside each other, rather than constantly feeding into each other like the DNN. When information relating to landslides in the Himalayans was fed into the new AI model, the SNN was able to stand up to the accuracy of the DNN while still allowing researchers to point to specific variables that seemed to have a larger effect on landslides than others.
“Similar to how autopsies are required to determine the cause of death, identifying the exact trigger for a landslide will always require field measurements and historical records of soil, hydrologic and climate conditions, such as rainfall amount and intensity, which can be hard to obtain in remote places like the Himalayas. Nonetheless, our new AI prediction model can identify key variables and quantify their contributions to landslide susceptibility.” – Seulgi Moon
Featured Image Credit: Taicheesy, CC BY-SA 4.0, via Wikimedia Commons
