ARTIFICIAL INTELLIGENCE
ARTIFICIAL INTELLIGENCE IN EM : MERGING WHITE COATS WITH BLACK BOXES
Martin Huecker , MD , & Jacob Shreffler , PhD
Artificial intelligence ( AI ) refers to a computer system ’ s ability to perform sophisticated tasks ( visual perception , decision-making , speech pattern recognition ) that normally require human intelligence . AI lives at the intersection of statistics , which determine relationships from data , and computer science , which focuses on algorithms to make predictions . AI purists refer to the study of “ any system that perceives its environment and takes actions that maximize its chance of achieving its goals .” Thus , you may be interacting with AI if a machine does not “ learn :” web search engines ( e . g ., Google ), recommendation systems ( YouTube , Amazon and Netflix ), speech recognition ( Siri and Alexa ), and self-driving cars ( e . g ., Tesla ).
Major subfields of AI include machine learning ( ML ), deep learning ( DL ) and natural language processing ( NLP ). Machine learning refers to a computer system improving its functioning , i . e . learning , with continued exposure to data . Deep learning also involves the review of inputs and outputs but uses a complex network of nodes that evolves as its exposure increases . DL is inherently a black box , with opaque decision logic that presents legal implications : who is at fault when the system fails ?
ML and DL do not simply perform tasks or generate data , they learn by creating and testing assumptions based on large data sets . Taken to the logical extreme , some foresee computers eventually acquiring consciousness , albeit artificial .
For people resistant to change , the good news is that we already use machine learning in medicine – have you ordered a CBC or an ECG recently ? Health care represents the largest market for AI investment across all sectors . The AMA prefers the term “ augmented intelligence ” to emphasize the role of computers to assist us , not replace us . AI exists to help physicians provide optimal patient care . Consider valuing AI as you would your stethoscope or medications for sedation .
When people compare health care to aviation , we ER insiders envision a doctor trying to land a plane full of delirious passengers , with no seatbelts ( or perhaps soft restraints ), who forget or lie about medicines they have taken , who are not in the best possible shape when they come to us and it ’ s always , since COVID-19 , when we do not have a full flight crew . The controlled chaos of the Emergency Department ( ED ) encapsulates this situation more accurately than any other place in medicine . When we walk into the ED patient ’ s room , hundreds of data points strike us within seconds , and we have minutes or even seconds to make decisions .
After many years of pattern recognition , we strive for the skill to immediately halve the dichotomy of sick or not sick . All of the computing power available in 2022 cannot arm a robot to recognize patterns the way an ER doctor can . We meet patients for the very first time , potentially on their worst day , while juggling 15 other active patients , understaffed and sleep deprived . ER docs are part of the small remaining tribe of generalists in medicine ( family medicine , internal medicine , pediatrics ). Like those professionals who
6 LOUISVILLE MEDICINE