Is it possible to regulate AI? If so, how would the U.S. Food and Drug Administration (FDA) do it?
John Sotos, a cardiologist, medical director at Intel and member of the Air National Guard, answers those questions in a guest blog post in the Wall Street Journal.
His take on regulating AI is the FDA cannot test all possible failure scenarios. But, he writes, the FDA can ensure that the system require the physician or radiologist to rate the accuracy of each AI suggestion. Furthermore, each rating would be sent to the FDA, where each AI system would be tallied to show its accuracy. Those ratings would be published online by product name.
Sotos then goes on to discuss the ramifications of such regulation.
Read the rest and make your own judgment: