Current limitations to AI use in medicine

Taken From CNBC's Article: Tempus: Health start-up backed by Groupon co-founder joins the fight against Covid-19 by Daniel Bukszpan

[Read Full Article Here]

Dr. Jose Morey, Chief Medical Innovation Officer for Liberty BioSecurity and advisor for MIT Solve and NASA iTech, said that one problem in relying on AI in the case of a new virus is that there isn’t enough data at the moment.
“You have to have large data sets to be able to train, test, and validate,” he said. “When you have something new and novel like this, they just don’t exist. ... It will come eventually, but not yet.”

He added that the technology is also only as good as the data, and currently, there are systemic flaws in the way medical data is gathered and recorded. For example, he said that data in medical centers could be “dirty.”

“This means that the data is not structured to be plugged into the math that is AI,’” he said. He added that health-care data is also frequently “siloed,” or separated into unconnected compartments. For AI to do its job, it requires large amounts of interconnected data, which Morey said doesn’t exist in the U.S. health-care system.
“Health-care entities, such as device manufacturers, electronic medical record vendors, and hospitals, are not incentivized to share data amongst each other,” he said. “This is why there are so many AI companies doing incremental algorithm and application development.”

​[Read Full Article Here]