The Need For Humans Will Remain
Artificial intelligence is already revolutionizing health care, and changes are happening fast. Understandably, the disruption is causing questions and even concern. But the truth is, although the way medicine is practiced will necessarily change, the need for human practitioners isn’t going anywhere.
The Roots of Concern
Artificial intelligence has a number of advantages over human intelligence.
It’s fast, for one thing. AI can store more data, access it faster and with perfect recall, and synthesize it with breathtaking efficiency. Its image recognition capabilities are at the next level, as well, making it potentially game changing as a diagnostic tool. When it comes to diagnosing breast cancer and skin cancer, for example, AI systems performed better in trials than human doctors and radiologists, picking up a greater number of cases and reducing both false positives and false negatives. Oxford University is currently trialing AI diagnosis for prostate cancer diagnosis, and researchers are hopeful for similar successes.
AI’s ability to rapidly synthesize and analyze diverse sets of data means that AI systems are very, very good at prediction. In fact, in 2019, an AI system developed by the Canadian company Blue Dot, detected a cluster of unusual pneumonia cases near a market in Wuhan China more than a week before the World Health Organization, working with traditional epidemiological models, issued a warning for the novel COVID-19 virus.
Some of the most important developments happening right now are in machine learning. Artificial intelligence has been present in health care in various forms since the 1970s, but recent developments in big data, computer processing power, and algorithm development have meant a giant leap forward in AI systems’ ability to teach themselves. Some experts predict that AI will reach a point of singularity–where it equals, then surpasses human intelligence–within the next decade.
One couldn’t blame a casual observer for being concerned.
A Bit of Perspective
Disruptive events always cause concern, and change is, and always has been inevitable. But before panicking, it’s important to adjust one’s perspective.
The 2009 introduction of Grammarly, an AI-enabled cloud-based writing assistant, was revolutionary in its own way. Suddenly, in addition to checking spelling, writers could self-edit for grammar, language use, and register, as well as check for plagiarism. Grammarly uses AI and natural language processing to train itself to adapt not only to the nuances of human language usage, but to the nuances of individual users’ language use.
It’s a powerful tool, but in the more than ten years since its introduction, it hasn’t replaced human editors. Why not? One reason is that an AI system is only as good as the corpus it trains on. AI training on a corpus of imperfect user-generated writing will necessarily generate errors–sometimes quite amusing ones.
Another reason is that, unlike an algorithm, human behaviors–like language use–are sometimes unpredictable, sometimes contrary, and are liable to go off-script. It takes a human editor to recognize sarcasm, semantically significant alternate spellings, and slang. In these cases, the AI-generated alternate suggestions are often laughably inappropriate.
So, while Grammarly is now a mighty arrow in the editor’s quiver, its output will always require final human evaluation.
So, it is with AI in healthcare settings. Poorly designed algorithms or biased data sets can result in imperfect learning and errors in performance. Moreover, artificial intelligence cannot replicate the therapeutic relationship between practitioner and patient, which is so vital to effective treatment.
AI will never replace medical personnel, but it can be an incredible assistant.
AI Will Enhance Human Practice
As part of a human practitioner’s toolkit, artificial intelligence has the potential to turbocharge healthcare in a number of different ways.
Improving Efficiency
AI systems’ data-crunching virtuosity can help both healthcare organizations and individual practitioners to make the most of their data. Imagine a given practitioner being able to instantly bring vast bodies of knowledge to bear on a given case. Combining and analyzing disparate data can also help practitioners to identify problems they might have missed or predict issues that might arise.
At the organizational level, harnessing AI’s analytical abilities can increase the efficiency of operational workflows and processes, helping the practice to make the most of its resources.
Relieving Administrative Burden
AI’s ability to store, recall, and synthesize information can go a long way toward helping practitioners and administrators struggling under the burden of excessive paperwork. From optimizing research to highlighting patterns and trends a human may miss, AI has the potential to do the footwork for human practitioners, freeing them to provide patient-centered care. Products like IBM’s Watson Health are already doing all of these things and more.
Optimizing Patient Experience
Properly trained AI can also help to reduce health inequity. Unconscious racial, gender, age, and other biases demonstrably affect both research and care, resulting in poorer outcomes for many patients. Machine analysis of patient data removes human bias from the equation, paving the way for more objective diagnoses and interpretation of research results.
Artificial intelligence can help human practitioners to tailor treatment programs to the needs of individual patients. Just as Grammarly can help to individualize self-editing, AI can combine large datasets with machine learning to provide individualized insights and experiences for patients. By combining data sets, for example patient records, with databases of caseworkers and counselors, AI can help practitioners to put together more complete plans of care and treatment for their patients.
Error Detection
Medical errors result in an estimated 200,000 deaths per year and an estimated annual cost of $1.9 billion. Artificial intelligence systems could potentially reduce both of these figures by spotting errors, or even predicting them before they happen.
Toward the Future
Excitement over AI’s potential to take healthcare to the next level must be tempered by a healthy respect for risks and unintended consequences. Big data means big responsibility, especially when it comes to patient privacy. This new, powerful tool must be used ethically, responsibly, and for the benefit of patients as well as healthcare and insurance companies.