Google’s latest algorithm does not predict what you are looking for but when you will die

Google has created a tool that could predict how long a patient would have to stay in a hospital, the chances of it having to be readmitted, and even the likelihood that they might die soon.

It is an algorithm that exploits the potential of neural networks in the field of health care. Neural networks are models that basically seek to mimic the functioning of the human brain, that a computer is able to combine parameters to predict results, and also learn and improve in the process.

This form of artificial intelligence that Google has already used before to defeat its own reCaptcha, improve the voice recognition in Android, save electricity in its data centers, or recognize street numbers in streets View, for the first time opens the Doors to a new market,  it’s also a breakthrough and that could help doctors and health personnel in a very tangible way.

Google versus Hospital computers

google doctor

A woman with terminal breast cancer was admitted to a hospital, had fluid in her lungs when she was seen by two doctors and had a radiological study. The hospital computers read his vital signs and estimated that he had a 9.3% chance of dying while he was in.

Then it was the turn of the Google algorithm. It was able to combine more than 175,000 points of interest before delivering its assessment and determined the risk by 19.9%. The woman died in a few days.

Because Google’s predictive system was able to analyze previously out-of-reach data, in PDFs or in old record notes, it combined much more information than any existing technique and in a much faster and more accurate way. He even showed exactly what records led to that conclusion.

The predictive system was developed by the Google Brain team, the part specializing in health and biological sciences, along with a professor associated with Stanford University. The study was published in Nature and explains the use of predictive models and deep learning in electronic medical records to identify relevant information within patients ‘ medical histories.

The predictive model of Google is faster than any other system, and able to analyze many more data than before in the medical records of patients.

Google’s next step is to move its predictive system to clinics, and also work on various artificial intelligence tools that can predict symptoms and illness with a high level of accuracy.

Of course, there is both optimism and concern. Google has a good time making deals with several universities in exchange for anonymous information on millions of patients, and are looking for more contributors, for some is just too much information for a company that already knows enough about us, information they can use to capitalize.

While Google and its partners claim that their data is anonymous, secure and used with patient consent, they have already been found on other occasions by analyzing public medical records without patients being used, and some believe that Google will have more difficulty treating data rigorously when it expands to smaller hospitals.

Leave a Reply

Your email address will not be published. Required fields are marked *