A recent headline from MIT Technology Review grabs our attention: “Hundreds of AI tools have been built to catch covid. None of them helped.” The article summarizes the findings of studies that reviewed hundreds of AI tools intended to help with Covid: Diagnose patients; predict how sick they would get; triage resources and more.
The key finding: The tools were useless because they lacked good data. As one researcher concluded: “This pandemic was a big test for AI and medicine. But I don’t think we passed that test.”
“This pandemic was a big test for AI and medicine. But I don’t think we passed that test.”
Laure Wynants, Maastricht University, the Netherlands
Many of the problems found by the researchers are linked to poor quality data. With over 27B faxes (yes, faxes!) sent each year in the US for healthcare insurance claims, lab test results, physician referrals and more, it is not surprising that basic data flows don’t work well. And that researchers could not easily access the data trapped in a fax.
In addition, researchers tried to splice together smaller data sets, as their AI training requires mounds of data to initialize. The splicing led to duplicates and often the researchers trained and tested on the same data set, leading to an overconfident prediction of accuracy.
Here at GLYNT we empathize enormously. Our AI transforms documents, including faxes, into data. We know how much time and sweat it took to build our system. Not to be done on a moment’s notice at the start of a global pandemic.
But our AI has two features which make it a great partner for your AI. First, GLYNT does not need much training data. Take a look at this video, where we train on just two documents. This means if you are sitting on a pile of faxes, PDFs, or scans of paper, you can GLYNTify them and get the data your AI needs.
Second, GLYNT is highly accurate and has the built-in tools needed to produce 100% accurate data. So your healthcare AI can rely on GLYNT.
GLYNT has the same value-add in other industries too. Do you know anyone who wants AI that produces less than 100% accurate data? Do you know anyone that wants to collect 1000s and 1000s of documents just to get started? We haven’t met them yet!
Talk to GLYNT about how our AI can make your AI better.
