Robot doctors, automated hospitals, and so many probes, oh my! Artificial intelligence (AI) is everywhere nowadays, and it can be pretty intimidating. And I don’t mean intimidating like Ultron or Terminator. From a technical standpoint, AI seems like magic—like making golems from clay. But at its core, AI is an attempt to model machines after human intuition, and then transcend human capabilities.
In healthcare, AI has the potential to increase automation. For example, computers are beginning to diagnose eye conditions with a similar accuracy to physicians. But paradoxically, AI could also lead to more humanistic medicine. With the rate of advances in AI, it’s hard to keep track of what’s science—and what’s science fiction. Below, the basics to help you discern.
Scanning the Landscape: Types of AI
In the 1950s, the fathers of the field Marvin Minksy and John McCarthy described artificial intelligence with large brushstrokes. They defined AI as any task performed by a program, which if performed by a human, would require intelligence. That’s a pretty big umbrella statement! What exactly is intelligence? And what tasks require it?
Perhaps this ambiguity is why today’s definitions separate artificial intelligence into two categories: general AI and narrow AI.
General AI is the type that runs rampant in popular culture. Some examples include SkyNet in The Terminator, Ultron in Avengers: Age of Ultron, or maybe even Mary Shelley’s monster in Frankenstein. This is the form of artificial intelligence that attempts to mimic the flexibility and adaptability of human intellect. In other words, general AI gives a program a wide range of specialties and the ability to carry out completely different tasks. This type of AI mostly remains in the realm of science fiction instead of actual science.
Narrow AI, as the name suggests, takes a much more specialized approach to learning programs. Instead of trying to encapsulate all of human intelligence, narrow AI applications are designed to perform a small number of similar tasks. This form of AI permeates our everyday lives—constantly running in the background to direct our morning commutes and determine what shows up on our newsfeeds.
Examples of narrow AI in practice also include fraud monitoring systems, automated customer service systems, and even programs that find the cheapest rates for travel and lodging. In fact, narrow AI is having a tremendous impact on healthcare practices and costs today.
Narrowing It Down: AI and Healthcare
When you think of AI and healthcare, you probably conjure up images of android doctors and fully automated hospitals. And that is definitely one way that researchers hope to eventually implement AI. However, this vision still seems to be quite a few years away and pending some pretty heated debates about healthcare providers’ roles in healing. Like other aspects of AI implementation, the straight and narrow seems to be the most productive option—at least for right now.
Robotic Helping Hands
Fully sentient robots aren’t around yet, but robotics is still the narrow AI application with the most bang for the buck. The potential annual value of robotic-assisted surgery is projected to be around $40 billion by 2026, according to the Harvard Business Review. Robotic-assisted surgery doesn’t mean that a robot is replacing a surgeon. Rather, robotics support a surgeon to help her or him better manipulate surgical devices.
For example, in orthopedic surgery, a form of AI-assisted robotics analyzes data from pre-op medical records to physically guide the surgeon’s instrument in real-time during a procedure. Patients who have surgery done this way have reduced healing times, spend less time in the hospital, and have fewer post-op complications. In turn, this results in fewer expenses and greater profit for hospitals.
But not all AI applications—or even many of them—have anything to do with robotics. Most AI programs focus on data management and monitoring.
Virtual nursing assistant applications, for instance, provide quick answers to online queries. These apps use large amounts of data including user feedback to keep tabs on clients’ conditions. They frequently ask if patients are feeling okay and/or keeping up with their prescribed health regimen.
Virtual personal health coaches follow a similar logic. By drawing upon a patient’s electronic medical records, these programs use machine learning to build a personalized relationship with the client. The applications gradually figure out when a patient will be most receptive to learning about a health condition or how best to manage care.
Wouldn’t it be nice if we could just step into a scanner and find out what ails us? No blood tests, no pap smears, no prostate exams. What if just a quick glance by a machine’s eye could tell us everything we needed to know? It sounds far-fetched, and most folks wouldn’t be comfortable with being diagnosed by a machine with no human interaction whatsoever. And although we’re already seeing great potential for this AI application, that hesitancy is one reason that AI diagnostics are still in their infancy.
In recent years, AI algorithms have performed the same as or better than human healthcare professionals at diagnosing some problems. For example, one AI algorithm detected skin cancers at the same level of exactitude as dermatologists. In another instance, a company used a deep learning program to eavesdrop on human dispatchers as they took emergency calls. The algorithm analyzed what a person said, their tone of voice, and the background noise and detected cardiac arrests with a 20 percent more accuracy than humans.
Recently, other researchers have attempted to extend these diagnostic advancements to image analysis. They’re hoping that AI could be the next big thing in non-invasive diagnosis. The technology could also lessen healthcare professionals’ workload in analyzing very minute details for current imaging technologies.
The most unnoticed but pervasive uses of AI arise in a similarly overlooked sector of the healthcare workforce: secretarial labor. Did you know that over half of a nurse’s workload and nearly a fifth of a physician’s workload is devoted to activities outside of patient care? Between documentation, editing the patient record, and completing other regulatory paperwork, it’s now wonder that healthcare professionals today sometimes seem out of touch with patients.
AI has the potential to make healthcare more—well, human. It’s strange to think that bringing artificial intelligence into the hospital will make it more humanistic, but it makes sense when we realize all that goes on behind the scenes. AI-based technologies such as voice-to-text transcriptions could both shorten hospital wait times and improve the patients’ quality of care.
AI: To Robotics and Beyond
Artificial intelligence is so often portrayed as a dystopic technology that is bound to completely annihilate or replace humankind. But the reality is that the technology is currently poised to augment rather than replace us. In healthcare, AI has many possible uses to help both patients and providers. Artificial intelligence therefore might not only make hospitals more automated, but it could also make healing more human.
Feature Image Source: Getty/chombosan