AI and Medicine: Promising Future or Dangerous Mistake?

/in Blog/by
By Ben Rozner, NCL Summer Communications Intern
​
Artificial intelligence has encroached upon every aspect of people’s daily lives; health and wellness is no exception. While it can provide numerous benefits, including the ability to access straightforward medical advice without an appointment quickly, AI is far too prone to mistakes to be fully trusted with your health. Here are five things you should keep in mind when using AI to self-diagnose:
  1. AI is often wrong or deliberately misleading: AI accounts on TikTok that pose as doctors have been gaining millions of views recently, posing a “huge danger to public safety,” according to numerous experts. AI-generated physicians have peddled disproven myths about numerous activities causing cancer and, in some cases, fabricated medical products. Alex Ruani, health misinformation researcher at UCL, called these false claims an “industrialized exploitation of trust,” putting profits and engagement above people’s safety and health.
  2. AI may ignore emergency symptoms: While doctors are trained to ask specific questions that help patients recognize symptoms they might not otherwise mention, AI chatbots focus only on the prompt they are given.
  3. AI does not know your medical history. It does not know what medications you take, what surgeries you have had, or how your body works. AI can be used as a starting point, but it should not be the source of your official diagnoses. Generally, AI can provide surface-level education that a doctor can later clarify and work with.
  4. AI chatbots are not required to protect your personal information: Doctors must strictly follow HIPAA regarding patient confidentiality; AI chatbots such as ChatGPT do not have such restrictions. For example, a newer version of ChatGPT specifically designed to help people navigate health-related questions collects users’ personal information to make more accurate diagnoses and recommendations. While this may mitigate some of the other issues AI faces when giving medical advice, it also puts users at greater risk. Users often upload their entire medical history to these chatbots, meaning that a single cyberattack could expose their most sensitive information to bad actors online.
  5. Your doctor knows best: As mentioned above. At the same time, an AI chatbot can provide helpful answers to basic questions; you should always check with a real healthcare provider before making major medical decisions. Doctors spend years learning how to diagnose and treat patients on a case-by-case basis. AI chatbots cannot give you the same personalized care and attention you need in a medical emergency.
The influence of AI on medicine does not stop at online inquiries. Artificial intelligence is already in your doctor’s office, and it often proves more burdensome than helpful. Doctors have found that AI-drafted patient charts are also riddled with mistakes, forcing them to revise each document manually. According to Sarah Preum, a corresponding author of a recent study on this issue, “AI can sound like a doctor but not think like one.” Importantly, the study found that an AI trained to adapt to individual physicians’ communication styles is more accurate and requires less editing. These AI models could save doctors precious time, but must be under constant scrutiny to ensure that they are safe and unbiased.
AI can provide numerous benefits to the medical sector. A study from Johns Hopkins found that algorithms designed to spot nearly imperceptible variations in medical images may revolutionize cancer screening. Predictive models could warn doctors before a medical emergency happens. AI has already helped researchers identify and create new drugs, vastly reducing the time it takes to develop new vaccines. And while chatbots are not reliable for self-diagnosis, they can serve as virtual assistants, providing helpful instructions to patients navigating a range of medical issues. There are undoubtedly more benefits still to be discovered, yet whether they outweigh the risks will likely remain heavily debated.
​There is no single right way to use AI regarding health and wellness. Whether it is doctors using AI chatbots to streamline the communications process or patients asking important questions they may be too embarrassed to ask their physician, AI can and does help everyone involved. Yet we must not ignore its pitfalls, as the risk AI poses to patient safety and privacy is immense. A single data leak or repeated instances of algorithm bias could put lives at risk in a field that already carries a great deal of peril. If medical professionals fully adopt AI systems, they must be consistently monitored to ensure maximum security for both patients and physicians.