Rates of suicide have been steadily on the rise over the past 20 years, increasing by more than 30% in half of the states . It is a problem on a desperate scale, and mental health professionals are working hard to turn the tide. They may soon have help from an unlikely quarter: the field of machine intelligence.

One artificial intelligence project, given thousands of patient records, was able to create a model that predicted with 95% accuracy whether a patient would commit suicide—and did it years before the attempt occurred. The therapeutic applications are promising.

“Part of the endgame of this is to develop an app that could be integrated into the electronic health record,” says Jordan Smoller, one of the researchers responsible for the project and director of the Psychiatric and Neurodevelopmental Genetics Unit at Massachusetts General Hospital. Equipped with this information “a clinician seeing a patient could be alerted to some degree of higher-risk.”

Smoller explores the frontiers of AI research in psychiatry, then the podcast talks with Kenton White, Chief Data Scientist at Advanced Symbolics. White is working with the Canadian government to discover populations at higher risk of suicide, based on an AI assessment of discussions on social media.

Listen here, or subscribe to the Proto podcast on iTunes or Stitcher.