	Over the past twenty years, health records have moved from paper to Electronic Health Records 
(EHRs).  If a patient suffers a medical emergency away from home, the hospital they are taken to can 
request their information from their usual health care provider.  If a patient is given a drug in the 
emergency room and then taken to an operating room, the anesthetist can be warned before 
administering a dangerous overdose of the drug.  But most importantly, EHRs have created the large 
dataset needed to fuel research advancements, the data needed to create a breakthrough in patientcare 
through AI.  However, the authors of the Stanford One Hundred Year Study on Artificial Intelligence 
quickly dismiss EHRs, blaming clunky user interfaces complete with annoying popups that physicians 
routinely dismiss.  This is an unfair criticism; it is hard to fathom that the systems are clunkier and more 
annoying than the thick folders of patient data that were passed around before EHRs.  In reality, EHRs 
are the backbone of nearly every healthcare benefit they predict AI will provide in the next fifteen years.
	The paper predicts advancements from AIs ability to mine outcomes from millions of patient 
clinical records.  Well, there would be almost nothing to mine if it were not for EHRs.  EHRs collect all of 
the data and store it in a database that makes data mining possible.  In fact, the paper specifically 
mentions the water crisis in Flint, and that crisis was actually first identified from a report built on EHR 
data.  The predictive models being built to prevent future water contaminations and other healthcare 
crises are likely heavily reliant on EHR data.
	Next, the paper foresees a breakthrough in healthcare image recognition by AI, specifically for 
performing diagnoses. While the technology to perform a diagnosis from an image is not from EHRs, 
EHRs are partly responsible for its advancement.  EHRs have driven the digitalization of images so that 
they can be easily accessible from the system, which created the dataset necessary to develop AI image 
recognition.  The paper states an obstacle to image recognition development is that only very large 
organizations can develop it because of the barriers to data sharing in healthcare.  However, EHRs and 
third-party organizations are currently building research databases comprised of the data from their 
community members.  These efforts will make it possible for this technology, and others, to develop 
outside of a large organization.  EHRs have undoubtedly made a large impact in healthcare image 
recognition.
	Finally, the paper mentions the benefit of mobile applications, from a new generation of tech 
savvy clinicians utilizing specialized mobile applications to the population as a whole employing mobile 
applications to do things like find similar health populations or tell them when to exercise.  EHRs are 
largely responsible for this benefit.  EHRs provide their own mobile applications and offer APIs that 
allows other applications to access the EHR database.  The young clinicians and the rest of the 
population are likely either directly using an EHR application or using an application that take advantage 
of an EHR database. EHRs are a critical part of the healthcare benefits from mobile applications.
	While some criticism of the user interface of EHRs is fair, the paper ignores the large benefits 
that the systems currently provide and the fundamental role that EHR data will provide in the 
advancement of healthcare.  Already, EHRs have brought AI into nearly every part of hospitals, and the 
future of EHRs and AI will be intertwined in the next fifteen years.  The authors should have credited 
EHRs for advancing AI, not blamed them for holding it back.
