ARTIFICIAL INTELLIGENCE
THOUGHTS ON THE INTELLIGENT USE OF TECHNOLOGY IN MEDICINE Benjamin Rogers , MD , MS
Articles discussing artificial intelligence ( AI ) in medicine range from urging caution to suggesting it will change practitioner ’ s lives for the better . In practice , it can be challenging to know where the line that separates beneficial and detrimental exists . I cannot claim to have that answer , but I would like to propose that this fact , the uncertainty , makes artificial intelligence just like every other tool medicine has incorporated into its toolbelt or left by the wayside . It is precisely the fact that medical practitioners are accustomed to critically thinking about new technologies , rigorously testing them , improving them and making dynamic decisions regarding implementation that makes medicine the perfect place for advanced computational methods to best tested for real world functionality . Beyond the altruism , the rate of publications only continues to rise ( Figure 1 ) and the proposed clinical applications increase , suggesting that one way or another , technology ’ s reach will likely find us all . But what is required of us as practitioners to ensure our patients receive better care with the help of technology ?
At the heart of any conversation regarding AI use in medicine is the issue of reliability . Intuitively , the optimal scenario would involve a technology able to never do worse than a human , at minimum ; one that could improve our own capacities would be even better . We already have tools that meet this criterion . The readiest example is the calculator , which far surpasses any human I ’ ve ever met . Clearly calculators raise few mental red flags , but what separates the calculator from the machine learning algorithm that is spotting cancer on radiology images ? The crux of the issue is that at some point , the data that calculators use to answer any mathematical problem they encounter was hard coded into their machinery . A human made every decision the calculator makes ; the rules are the rules , calculator error = human error . Artificial intelligence crosses that threshold ; it asks the computer to use the data it receives and create . Herein lies the rub : every aspect of the computer ’ s decision making is no longer controlled .
Most of us are comfortable being beaten at chess by a computer . Many of us would be much less at ease with letting a computer fly the airplane we are in without a human in the cockpit as well . So , too , most physicians are more likely to employ technology where the risk to the patient is low , but hard pressed to let a computer make lifeand-death decisions . My experience has been that the delineation is rarely so black and white . I would like to touch briefly on the concept
14 LOUISVILLE MEDICINE