Google AI Can Reportedly Predict a Patient’s Time of Death

Artificial intelligence is expected to do many amazing and previously unimaginable things once the technology has advanced far enough. However, it would seem that Google’s AI has gotten to a level of advancement where it can actually predict a patient’s time of death. At least, this is what a new paper that presents the technology’s feature is saying.

Published in Nature, the paper explains how the AI is supposedly able to make certain predictions by examining mountains of health documents. Through deep learning, the AI is able to analyze the data it is fed to create a predictive model that encompasses everything from the length of stay to when a patient will die.

“Deep learning models achieved high accuracy for tasks such as predicting: in-hospital mortality (area under the receiver operator curve [AUROC] across sites 0.93–0.94), 30-day unplanned readmission (AUROC 0.75–0.76), prolonged length of stay (AUROC 0.85–0.86), and all of a patient’s final discharge diagnoses (frequency-weighted AUROC 0.90). These models outperformed traditional, clinically used predictive models in all cases,” the study’s Abstract reads.

In order to make these predictions, however, the AI had to analyze a considerable amount of data. They even included handwritten notes by health professionals. As a result, the neural network had enough information to make a reasonably accurate probabilistic conclusion.

As Futurism notes, hospitals could make use of this information to significantly improve patient care. Patients who are at high risk of death due to preventable factors can be attended to. The same goes for shortening hospital stays to reduce expenses.

The flipside of this encouraging development is how it could afford Google a monopoly on the healthcare industry. Some health professionals are also concerned that such breakthroughs could lead to an overreliance on machines to provide medical information and advice. However, there is an argument to be made that AI capabilities have advanced so much that such concerns are now moot.